Read the full guide.

Read the full guide.

Read the full guide.

Read the full guide.

Read the full guide.

Read the full guide.

Read the full guide.

Read the full guide.

Read the full guide.

Read the full guide.

Read the full guide.

Read the full guide.

Read the full guide.

Read the full guide.

Read the full guide.

Read the full guide.

Read the full guide.

Read the full guide.

Read the full guide.

Read the full guide.

Read the full guide.

Read the full guide.

Read the full guide.

Read the full guide.

Read the full guide.

Read the full guide.

Read the full guide.

Read the full guide.

Read the full guide.

Read the full guide.

Read the full guide.

Read the full guide.

Read the full guide.

Read the full guide.

Read the full guide.

Read the full guide.

Read the full guide.

Read the full guide.

Read the full guide.

Read the full guide.

Read the full guide.

Read the full guide, with the calculator and the 5 step self check.

Read the full guide, with the calculator and the 5 step self check.

Read the full guide, with the calculator and the 5 step self check.

Read the full guide, with the calculator and the 5 step self check.

Read the full guide, with the calculator and the 5 step self check.

Read the full guide, with the calculator and the 5 step self check.

Read the full guide, with the calculator and the 5 step self check.

Read the full guide, with the calculator and the 5 step self check.

Read the full guide, with the calculator and the 5 step self check.

Read the full guide, with the calculator and the 5 step self check.

Read the full guide, with the calculator and the 5 step self check.

Read the full guide, with the calculator and the 5 step self check.

Read the full guide, with the calculator and the 5 step self check.

Read the full guide, with the calculator and the 5 step self check.

Read the full guide, with the calculator and the 5 step self check.

Read the full guide, with the calculator and the 5 step self check.

Read the full guide, with the calculator and the 5 step self check.

Read the full guide, with the calculator and the 5 step self check.

Read the full guide, with the calculator and the 5 step self check.

Read the full guide, with the calculator and the 5 step self check.

Read the full guide, with the calculator and the 5 step self check.

Read the full guide, with the calculator and the 5 step self check.

Read the full guide, with the calculator and the 5 step self check.

Read the full guide, with the calculator and the 5 step self check.

Read the full guide, with the calculator and the 5 step self check.

Read the full guide, with the calculator and the 5 step self check.

Read the full guide, with the calculator and the 5 step self check.

Read the full guide, with the calculator and the 5 step self check.

Read the full guide, with the calculator and the 5 step self check.

Read the full guide, with the calculator and the 5 step self check.

Read the full guide, with the calculator and the 5 step self check.

Why a kept chat with a dated yes is worth more than one sale, Meta’s rules for WhatsApp, Messenger and Instagram DMs in one table, a “What is one saved chat worth?” calculator and a “Can I message them now?” checker. Read the full article.

We asked the AI behind ChatGPT 100 buying questions Indians ask, 60 about D2C products and 40 about courses and coaches, with web search on. It searched the web in all 100 answers, cited ranking or comparison lists in 56, and cited the top pick’s own site twice as often as the last brand named. Read the six findings, a 30 day GEO plan, a free AI visibility score and the raw data in the full article.

Read the full article

Read the full study, the charts and the 5 leak self check.

People used to find a brand by searching and clicking. Now many of them ask. They ask Google, they ask ChatGPT, they ask a WhatsApp group of friends. And the answer they get is built from what other people said about you in public.

The short answer: your community is now part of your search ranking. Search and AI trust brands that real people talk about, in places anyone can read.

I have worked in growth for 8+ years, with 25+ brands, since 2017, and run DM automation since 2018. The brands that keep showing up are rarely the loudest. They are the ones whose customers answer questions for them.

Here is why community now shapes search and AI answers, the flywheel that makes it work, 7 steps, and a 30 day plan for founders and for creators. There is one step most founders skip, and it is the one AI answers notice most. We get to it at step 6.

Why do Google and ChatGPT trust brands with a community?

Start with Google’s own words. Its guide to helpful content says it looks for experience, expertise, authority and trust. It adds one line every founder should read twice: trust is the most important of the four.

Google also asks if your content shows real experience, the kind that comes from using a product yourself. A forum post from a real buyer is exactly that.

Google has backed this with action. In November 2023 it added support for forum markup, for search features built to show views from forums and communities. In February 2024 it expanded its deal with Reddit, saying people increasingly come to Google to find helpful Reddit content, like product tips. And for AI Overviews and AI Mode, Google says the same SEO basics apply. No secret trick.

AI answers follow the same pattern. One study tracked 680 million citations in AI answers from August 2024 to June 2025. In Google AI Overviews, Reddit, YouTube, Quora and LinkedIn led the top 10 cited sites. In Perplexity, Reddit alone was 46.7% of its top 10 citations. In ChatGPT, Wikipedia led and Reddit came next.

Another study looked at 75,000 brands. Mentions of a brand across the web lined up with AI Overview visibility far more closely than backlinks did. The score was 0.664 for mentions and 0.218 for backlinks. Brands in the top group for mentions earned up to 10 times more AI mentions than the next group. And 26% of the brands had no AI Overview mentions at all.

These are links between numbers, not proof of cause. But the direction is clear. Being talked about in public now counts.

How does the community flywheel work?

Think of a busy tea stall near a college. Nobody puts up a big board for it. Students tell new students. The owner knows them by name and answers their questions. When someone asks where to get good chai, ten people point the same way.

Search and AI now work like those ten people. They repeat what the crowd says.

The flywheel has five parts. Members ask you questions. You answer in public, where others can read. Happy members mention you and leave reviews. Google and AI answers see those mentions and trust you more. New people find you, join, and ask their own questions. Then it turns again, a little faster each time.

What does a silent brand lose?

Arjun sells cold pressed coconut oil from Kochi. Good product, steady Meta ads, a nice website. One Sunday his niece in Bengaluru asked ChatGPT which coconut oil to buy for cooking and hair.

It named a smaller brand from Coimbatore. It quoted a Reddit thread where three buyers praised it. Arjun had never heard of that thread.

He looked closer. The rival founder ran a WhatsApp community of a few hundred buyers. She answered oil questions on Reddit and Quora under her own name. Her Google profile had a long list of reviews, each with a reply. His had 11, and he had replied to none.

Arjun had spent his money on being seen. She had spent her time on being talked about. So before he spends another rupee on ads, Arjun works through 7 steps.

1. Pick one home for your people

You need one place where buyers and fans can gather and talk to you. In India that usually means a WhatsApp Community, a Telegram channel, or an Instagram broadcast channel.

Pick the one where your buyers already spend time. Do not open all three. An empty group feels worse than no group.

Do this: Create one group. Write a short welcome reply that says what the group is for, the rules, and how often you post. Invite your last 50 happy buyers or fans by hand.

2. Turn every repeated question into a public answer

Your DMs hold your richest content. The same 10 questions come again and again. “Is it safe for kids?” “COD hai?” “Which plan should I pick?”

Each answer sitting in a private chat helps one person. The same answer on a public page helps everyone who searches for it, and gives Google and AI something to read.

Do this: List the 10 questions you get most. Answer each in plain words on one FAQ page on your site. Use the exact words your buyers use.

3. Answer where your buyers already ask

Your buyers already ask questions on Reddit India subs, Quora and in LinkedIn posts. Those are the places AI answers read most.

The rule is simple. Help before you sell, and never push. Say who you are. Give a full answer. Do not drop a link unless the group allows it and it truly helps.

Do this: Find 3 fresh questions in your field this week, on a Reddit India sub or Quora. Write a helpful answer to each under your real name. Read each group’s rules before you post.

4. Ask for reviews at the happy moment

A review is a public mention with your name on it. Google says positive reviews and helpful replies can help a business stand out on its profile.

The trick is timing. Ask when the buyer is happiest, right after a good result, not right after payment.

Do this: Pick your happy moment, like the day after delivery or the end of week one in a course. Send one short message asking for a Google review, with the link. Reply to every review, good or bad, within two days.

5. Let members talk where others can see

A private group is warm, but search cannot read it. The magic happens when member stories come out into the open.

Share a member’s win as a post, with their permission. Invite a happy buyer to write a short story you can share. Host a monthly live where members ask questions, then post the top answers.

Do this: Each week, take one good member question or win from the group. With permission, turn it into a public post on LinkedIn, Instagram or your blog.

6. Make your name easy to mention

This is the step most founders skip. If people call you three different names, search and AI cannot add up your mentions.

Use one brand name everywhere. Keep the same spelling on your site, your Google profile, your social bios and your marketplace pages. Write a clear about page that says who you are, where you are and what you sell. Link your social profiles from your site.

Do this: Search your brand name today. List every spelling you find. Fix your own profiles so they all match, and link them from your site.

7. Track mentions, not just traffic

Clicks tell you who came. Mentions tell you who is talking. You need both.

Ask every new buyer one question: “Where did you hear about us?” Watch searches for your brand name. Google Search Console now has a branded queries filter that splits searches for your name from the rest.

Do this: Add the “where did you hear about us” question to your order form or welcome message. Check your branded searches once a month. Ask ChatGPT and Google the 5 questions your buyers ask, and note who gets named.

Your 30 day community plan

You do not need a big team. You need 30 minutes a day and a plan. Here is one for founders and one for creators.

If you are a creator, your followers are rented. A community is how you start to own your audience, and we wrote a whole guide on how Indian creators build an audience they own.

What happened to Arjun?

Arjun started small. He opened one WhatsApp Community and invited 60 past buyers by hand. He wrote a welcome reply and posted one cooking tip a week.

He wrote down the 12 questions buyers asked most, and put the answers on one FAQ page. He answered two coconut oil questions a week on Reddit and Quora, under his own name. He asked for a Google review the day after delivery, and replied to every one.

Weeks later, a stranger in a Reddit thread said, “Try the Kochi one, the founder answers you himself.” Arjun had never met that person.

No ad bought that sentence. His buyers wrote it for him. It is not a promise of a ranking. It is how the loop starts to turn.

Which step should you start with?

Start with step 2 today. Write down your 10 most asked questions and answer them on one page. It costs nothing and you already know the answers.

Then pick your home in step 1 and invite your happiest buyers. Add step 4 the same week. Steps 3, 5, 6 and 7 build on top, a little each week.

The numbers in one place

Your community strength score

What should you remember?

• Google says trust matters most, and it rewards real experience, like a buyer’s own story.
• AI answers lean on forums. Reddit, YouTube, Quora and LinkedIn led Google AI Overview citations in one study of 680 million citations.
• Brand mentions tracked AI Overview visibility about three times more closely than backlinks in a study of 75,000 brands.
• A private group is warm but invisible. Bring answers and stories out where search can read them.
• Start with your 10 most asked questions. Answer them in public this week.

Sources (checked 23 October 2026)

• Google Search Central: Creating helpful, reliable content for people (experience, expertise, authority and trust, “trust is most important”, real experience from using a product)
• Google Search Central: AI features and your website (the same SEO basics apply to AI Overviews and AI Mode)
• Google: An expanded partnership with Reddit, 22 February 2024 (people increasingly use Google to find helpful Reddit content, like product recommendations)
• Google Search Central: New in structured data, discussion forum and profile page markup, 27 November 2023 (search features built to show views from forums and communities)
• Google Business Profile Help: Tips to improve your local ranking on Google (positive reviews and helpful replies can help your business stand out)
• AI Platform Citation Patterns, June 2025, updated August 2025 (680 million citations, August 2024 to June 2025, top 10 cited sources per AI platform)
• An Analysis of AI Overview Brand Visibility Factors (75K Brands Studied), 26 May 2025 (brand web mentions 0.664, backlinks 0.218, 26% of brands with no AI Overview mentions)
• We Studied the Impact of AI Search on SEO Traffic, 21 July 2025 (Quora and Reddit the two most cited sites in Google AI Overviews in that study)
• Google Search Central: Introducing the branded queries filter in Search Console (available to all eligible sites since 11 March 2026)

Arjun, his rival and their numbers are made up to show the method. They are not results from a real brand.

This article first appeared on madhurtyagi.sites.tab.bot, written by Madhur Tyagi, The Human Algorithm.

A webinar is an online class you teach live. It is free or cheap, like the 99 rupee class many Indian coaches run. At the end you invite people to your paid program.

I have worked with 15+ coaches and creators, and I have built DM automation since 2018. The pattern is almost always the same. The class goes well. The chat fills with fire emojis. Then the price slide comes, the chat goes quiet, and the coach decides webinars “do not work”.

They usually do work. The sale leaks out in the hours after you log off.

Here is where it leaks, why it happens, and seven fixes with lines you can copy.

Where does your webinar lose buyers?

Ritu has one big wish. She wants 30 working moms in her 12 week Fit at 40 program, so she can leave her 5 AM shift at a gym. She has the sign ups. So where do the buyers go?

Do the count she did. Open your last webinar and write down four numbers. How many signed up? How many came live? How many watched the replay? How many bought?

Your show up rate is the share of sign ups who actually come. For Ritu it was 71 of 212, about one in three. Put your own numbers in below. Nothing leaves your phone.

When Ritu typed hers in, the biggest group was not the people who skipped. It was the 68 moms who came, stayed till the end and did not buy. They liked her. They just never heard from her again.

Why does this happen at all?

Think of a chai stall at a busy bus stand. A man stops, smells the chai and asks, “Kitna?” The chaiwala is busy and does not answer. The bus comes, and the man is gone. He did want the chai. Nobody answered him in time.

A webinar works the same way. At the end of your class, people are warm. They trust you. They also have one doubt left. Will this work for me? Can I pay for it? Will I find the time?

A doubt like this is called an objection. It is the reason a person gives for not buying yet.

That warm feeling fades fast. By Monday the kids have school, the boss has a deadline, and your class is one of fifty things on their phone.

A Harvard Business Review study shows how fast interest cools. It looked at 1.25 million sales leads. Firms that tried to reach a lead within an hour were nearly seven times as likely to have a real talk with the decision maker. That is compared with firms that tried even an hour later.

So the job after a webinar is simple. Reach people while they are warm, and answer the doubt before it turns into “later”. It is the same rule as in my 7 step funnel leak audit. A funnel is the path from stranger to buyer: ad, sign up, class, payment. Like a kitchen funnel, it is wide at the top and narrow at the bottom.

What does a quiet Monday cost you?

Every person who came live gave you 90 minutes, a little money and some trust. You paid for the ad that brought them. When nobody follows up, you do not just miss a sale. You lose a person who was ready, and you pay again to find someone new.

Ritu worked it out on the back of a gym form. Her ad money had bought 212 sign ups. Her silence after the class gave most of them away.

1. Do people get reminders they actually see?

The leak: people sign up on Wednesday for a Sunday class. By Sunday they forget. One email sits under forty others and never gets opened.

Ritu sent one email, on Wednesday. Many of her sign ups never saw it.

The 10 minute test: sign up to your own webinar with a spare number and email. Count the reminders you get, and see where they land.

The fix: send short reminders where people read them. Send one the day before, one an hour before, and one when you go live. Use WhatsApp for people who said yes to updates there, and email for the rest. Keep each one to two lines, with the link.

Copy this (1 hour before):

“Hi [name], we go live in 1 hour. Today I will show you [one clear result]. Your link: [link]. See you at 7!”

Running ads to fill the class? Check that you are not paying for fake leads from Meta ads before you blame the reminders.

2. Do people have a reason to come live?

The leak: if the class is the same on replay, people think, “I will watch it later.” Later rarely comes.

Ritu’s sign up page said “Free replay for all” in big letters. So many moms planned to watch later, and never did.

The 10 minute test: read your sign up page and your reminders. Is there one thing people get when they come live, and miss on the replay?

The fix: give an honest live reason. Answer questions by name in the last 15 minutes. Look at three people’s plans on screen. Say it in every reminder.

Copy this (in the reminder):

“Come live at 7. In the last 15 minutes I will answer your questions one by one, by name.”

3. Does your class answer the real doubt before the price?

The leak: most coaches teach well and then show the price. The doubts in people’s heads never get said out loud. So they stay, and they win.

Ritu’s moms had one big doubt. “I have tried diets before. Will this work for a mother with no free time?” She never said it, so nobody heard her answer.

The 10 minute test: read your last webinar chat and the DMs that came after it. Write down the top three doubts people typed.

The fix: before the price, name each doubt yourself. Then answer it with a real story or a real day in the program. Think of a cricket captain who sets the field before the bowler runs in. You plan for the shot before it comes.

Copy this (before the price slide):

“Some of you are thinking, ‘I have no time for this.’ Fair. Here is what one day in this program looks like: [three short lines].”

4. Is your offer clear, with one easy way to pay?

The leak: the price slide shows three plans, two bonuses and a link to a long form. A confused mind says no.

A low ticket offer is a small price, like the 99 rupee class. A high ticket program costs a lot more, often a few months of coaching. Your webinar is the bridge between the two, so the bridge must be easy to cross.

Ritu showed three plans and a bank transfer note. The chat filled with “Which one should I take?”

The 10 minute test: show your price slide to a friend for ten seconds. Can they say what they get, what it costs and how to pay?

The fix: make one main offer, said in one sentence. Give one payment link or UPI ID. UPI is the pay by phone system most Indians use every day. Show it on screen, then send it in the chat and on WhatsApp.

Copy this (offer line):

“Here is everything in one line: [program] for [weeks], with [what they get]. Pay here: [link]. Any questions? Reply to this message and I will answer you myself.”

5. Do you follow up within an hour of the live?

The leak: the class ends at 8:30. You are tired, so you log off. The warm crowd goes to sleep and wakes up cold.

Ritu sent her follow up on Tuesday afternoon. By then, most of her 71 moms had moved on.

The same Harvard Business Review article found many firms are slow. In an audit of 2,241 US companies, 37% replied to a new lead within an hour. 23% never replied at all.

The 10 minute test: check what time your last follow up went out after the live.

The fix: write the follow up before the webinar starts. Send it within the hour after it ends, from a person, to everyone who came. Send a different short message to people who signed up but did not come.

Copy this (people who came):

“Hi [name], thank you for staying till the end today! You asked a great question about [topic]. Do you want me to tell you honestly if the program fits you? Just reply yes.”

Copy this (people who did not come):

“Hi [name], we missed you today. The replay is here till [day, time]: [link]. Short on time? Watch the part at [minute] about [topic].”

6. Is there a replay plan with an honest end date?

The leak: a replay is the recording of your live class. Many coaches send it days later, or leave it open forever. With no end date, “I will watch it later” turns into never.

That matters, because the replay is not a leftover. One study of over a million webinar sign ups found about 6 in 10 people watched, live or on replay. And 42% of all views came from the replay.

Ritu sent her replay on Wednesday with no end date. Almost nobody watched it.

The 10 minute test: when did your last replay link go out, and when does it close?

The fix: send the replay the same night. Give it an honest end date, like 48 hours, and keep it. If you say Tuesday 9 PM, take it down at 9 PM. A fake deadline breaks trust for good.

Copy this (replay message):

“The replay stays up till Tuesday, 9 PM. After that I take it down, because [real reason, like the batch starts Wednesday].”

7. Does a person reply to every price question in DMs?

The leak: after the class, people do not ask in the open chat. They ask in private. A DM, or direct message, is a private chat on Instagram. On WhatsApp it is just a message: “Ma’am, fee kitna hai?” If nobody answers that night, the moment is gone.

This is what Ritu found on Monday morning. 38 messages, most of them some form of “fee kitna hai?”, all unread. These were the buyers. They had raised their hands, and nobody saw.

The 10 minute test: open your DMs and WhatsApp after your last webinar. Count the price questions with no reply, or a reply after 12 hours.

The fix: give one person the job of replying to every DM and WhatsApp question for 24 hours after the live. Reply with the price, what it includes and one question back. A question back starts a real talk, so the buyer is not left alone with a number.

Copy this (price reply):

“Hi [name], the program is [price] for [weeks]. It includes [3 things]. Can I ask what you want to change most in the next 3 months? Then I can tell you honestly if it fits.”

More lines like this are in my guide on turning Instagram DMs into coaching clients. Software can help sort the messages, but keep the money talk human. Here is what to keep human when AI helps. Want a trained person to run these replies with you? See how the Revenue Desk works.

Which fix should you start with?

Ritu did not fix all seven at once. For her next Sunday class she did fix 5 and fix 7, because they cost nothing but time. Her sister sat with her phone and replied to every DM that night. Ritu sent the thank you message at 8:40.

By Tuesday, the moms who had asked “fee kitna hai?” were in her new batch group on WhatsApp. A month later, Ritu gave notice at the gym. Her 5 AM alarm is now for her own class.

Ritu is made up, and your numbers will be different. The order is the useful part. Start where your count points:

• Lots of sign ups, few came live: fix 1 and fix 2.
• Many came, few bought: fix 3 and fix 4.
• People asked, then went quiet: fix 5 and fix 7. They cost nothing, and you can start this Sunday.
• Most people watch later: fix 6.

Thinking of hiring someone to run all this? Read agency vs freelancer vs AI growth partner before you pay anyone.

Want to see where else your leads leak, beyond the webinar? The free Growth Leak Check takes 2 minutes and shows your top 3 leaks, each with one fix. It works for coaches who sell batches and programs.

Your webinar follow up checklist

What should you remember?

• Most webinar sales are lost after the live, not during it.
• Count four numbers after every webinar: signed up, came live, watched replay, bought.
• Remind people where they read, and give them an honest reason to come live.
• Say the real doubt out loud before the price, then make paying one easy step.
• Follow up within the hour, and treat every “fee kitna hai?” as a raised hand.
• Give the replay an end date, and keep it.

Sources (checked 21 October 2026)

• Harvard Business Review: The Short Life of Online Sales Leads, March 2011 (the 1.25 million lead study, and the audit of 2,241 companies: 37% replied within an hour, 23% never). The page is behind a paywall, so I read it in full on the Internet Archive copy.
• Contrast: Webinar Benchmarks from 1M+ Registrants, 2026 data (61% of registrants watched live or on demand, 42% of views on demand)

This article was published on madhurtyagi.sites.tab.bot, written by Madhur Tyagi, The Human Algorithm.

Pooja had one dream. She wanted her Aai’s mango pickle on kitchen shelves across India, not just in Kolhapur.

So she left her job, built a small website and ran ads. Diwali came, and the orders poured in. Most buyers picked cash on delivery. She packed every box herself, late into the night.

Then the parcels started coming back. Box after box. By the end of the month, a whole shelf in her godown held jars nobody wanted.

She had paid to ship each box twice, and sold nothing.

I have run performance marketing for 8+ years, for 25+ brands since 2017. I see Pooja’s shelf in many D2C brands. The good news: most of it is fixable. Not with a new courier. With a few small habits.

What is RTO, in plain words?

RTO means return to origin. A parcel leaves your warehouse, does not get delivered, and travels back to you.

You pay the courier to send it. You often pay again to get it back. The product sits in a truck for days. Sometimes it comes back damaged or stale.

COD means cash on delivery. The buyer pays the delivery person at the door, not online.

Most RTO lives in COD orders. Shipway’s ShipNotes report for FY25 says under 2% of prepaid orders were returned. For non prepaid orders, the number was nearly 26%. That is about one parcel in four.

Why do COD orders come back at all?

Think of a chai stall. You pay when the cup is in your hand. The stall never has a “return”.

A COD order is the opposite. The buyer says yes today, with zero money down. The real decision happens days later, at the door. In those days, a lot can change.

So a COD order comes back for four simple reasons:

1. They could not be found. Half an address, a wrong pin code, a phone that does not ring.
2. They changed their mind. It was a late night tap. Or a cousin in the family WhatsApp group found it cheaper.
3. They were not home. Or nobody had cash ready when the parcel came.
4. It took too long. The wish faded before the parcel arrived.

Every fix below closes one of these four gaps. When Pooja finally called 20 buyers whose boxes came back, she heard all four.

What does each returned parcel cost you?

A return is not one loss. It is four.

You lose the shipping to the buyer and back. You lose the ad money you spent to get that order. You lose the time spent packing it. And the stock is stuck while it travels, so you cannot sell it to someone who wants it.

Most founders count just the reverse shipping. That is like counting runs and ignoring the balls faced.

Put your own numbers in below. Nothing is sent anywhere.

When Pooja did this sum, she stopped thinking of RTO as bad luck. It was a line on her bill, every single month.

1. Is the address complete before you ship?

Two of Pooja’s returned boxes went to “Near temple, Shivaji Nagar”. There are many temples. There are many Shivaji Nagars.

Why it leaks: your kirana uncle knows your house. A courier in a new city does not. A short address form saves the buyer five seconds and costs you a parcel.

The test: open your last 30 COD orders. Count the ones with no house number or no landmark.

The fix: make house or flat number, street, landmark and pin code required fields. Check the pin code matches the city. Do not ship an order with a blank landmark until someone confirms it.

Copy this (note under the address box):

“Please add your house number and a landmark. It helps our delivery partner find you in one go.”

2. Do you check every COD order on WhatsApp before packing?

Pooja’s turn came on a quiet Tuesday. She messaged ten new COD buyers before packing. Seven said yes. Two fixed their address. One said, “Sorry, ordered by mistake.” That box never left her table.

Why it leaks: an order placed at midnight is a wish, not a promise. A short message turns the wish into a real yes, while you can still stop the parcel.

The test: for one week, count how many COD orders you pack without any contact with the buyer.

The fix: send every COD buyer a short WhatsApp message soon after they order. Ask them to reply with one number. Pack the clear yes orders. Call the ones who do not reply, before the day’s dispatch.

Copy this (order check message):

“Hi [name], this is [your name] from [brand]. Your order for [product] is ready to pack, cash on delivery, to [area] [pin code]. Reply 1 to confirm or 2 to change the address.”

Copy this (if they do not reply in a few hours):

“Hi [name], just checking before we send your [product]. Should we pack it for you today?”

A person can send these. Tools can help too. Our page on WhatsApp automation for D2C brands shows how a draft goes out with a person in charge. For more on where an agent fits, see how Indian D2C brands use AI agents.

3. Do you treat risky orders differently?

Pooja saw a pattern. Many returns were new buyers, big orders, from a few pin codes.

Why it leaks: not every order has the same risk. A buyer who already paid you twice is not the same as a stranger ordering a large gift hamper at 2 AM.

The same ShipNotes report found return rates on COD orders ranged from 18% in Vadodara to 35% in Patna. Where your buyer lives changes the odds.

The test: sort your last three months of returns by pin code. Are a few pin codes carrying most of them?

The fix: make simple rules. A new buyer on COD gets the WhatsApp check every time. A large COD order gets a call before dispatch. A pin code with many returns gets one extra check, or online payment, until it improves.

Copy this (call script for a large COD order):

“Hi [name], I am calling from [brand] about your order of [product]. We want to make sure it reaches you. Is this address right, and will someone be home with cash on [day]?”

4. Is paying online the easier choice?

Pooja noticed something else. Her prepaid buyers almost never sent anything back.

Why it leaks: money paid is a decision made. A buyer who prepaid has already said yes with their wallet. So the more of your orders are prepaid, the smaller your RTO.

Unicommerce’s India D2C Report 2026 shows the gap clearly. In the festive quarter, COD orders came back at 58%. Prepaid orders came back at under 15%. The report says the brands that brought RTO down to 21% by March all had a reason for buyers to prepay. Others were still near 39%.

The test: check what share of your orders are prepaid. Then look at your checkout. Is there any reason to choose online payment?

The fix: you do not have to stop COD. Make prepaid the easy, sweet choice. A small cashback, a free sample, faster dispatch for prepaid orders. Say it clearly at checkout.

Copy this (line at checkout):

“Pay online and we pack your order today, with a small gift inside.”

5. How fast does your parcel reach the buyer?

During the Diwali rush, Pooja shipped late. Some boxes took a week. By then, the buyer had moved on.

Why it leaks: a wish has a shelf life, like milk. The longer the wait, the more time a buyer has to change their mind.

ShipNotes found this in the data. COD orders attempted within one to two days were returned 22% of the time. At three to five days, it was 27%. After more than five days, it rose to 35%.

The test: for your last 50 orders, check the gap between order and the day delivery was tried.

The fix: dispatch within a day. Tell the buyer when to expect the parcel. Send the tracking link on the day it ships.

Copy this (dispatch message):

“Hi [name], your [product] left our warehouse today. It should reach you by [day]. Please keep [amount] ready if you chose cash on delivery.”

6. Do you call when a delivery fails?

One buyer told Pooja, “The delivery boy came when I was at work. Then the parcel just went back.”

Why it leaks: when a delivery fails, the courier marks an NDR. NDR means non delivery report. If nobody acts on it quickly, the parcel starts its trip back to you.

The test: check your courier panel today. How many NDRs are more than a day old with no action?

The fix: treat every NDR like a missed call from a buyer. Message or call within 24 hours. Fix the address, or lock a better time for the next try.

Copy this (NDR message):

“Hi [name], our delivery partner could not reach you today. When should they come back, and is this address right? Reply here and we will sort it.”

7. Do you know why each parcel came back?

At the end of the month, Pooja made a simple sheet. One row for every return. One word for the reason.

Why it leaks: if you do not know why parcels come back, you fix the wrong thing. It is like a captain changing the bowler when the problem is the field.

The test: can you say, right now, the top reason for your returns last month?

The fix: tag every return with one reason: address, not reachable, refused, not home or late. Every week, sort the tags by pin code, product and courier. Change one rule based on what you see.

Copy this (weekly question for your team):

“Which reason sent back the most parcels this week, and what one rule do we change?”

Which fix should you start with?

Pooja did not fix everything in one week. She started with fix 2, the WhatsApp check, because it cost nothing. Then she made the address form strict. By Holi, the shelf in her godown was almost empty. Last week, a jar of Aai’s pickle reached a buyer in Guwahati. It was paid before it left Kolhapur.

Your reason tags tell you where to start:

• Mostly “not found”: fix 1, the address form, and fix 6, the NDR call.
• Mostly “refused”: fix 2, the WhatsApp check, and fix 4, the prepaid nudge.
• Mostly “not home”: fix 5, the dispatch message with a delivery day.
• No idea: fix 7. Start tagging today.

RTO feels like a courier problem. Most of the time, it is a conversation that never happened before the parcel left. The same thing happens with leads. Our funnel leak audit walks through it step by step.

Want to find the leaks before the parcel stage too? The free Growth Leak Check takes 2 minutes and shows your top 3 leaks, each with one fix.

Your RTO checklist

What should you remember?

• RTO is a parcel that comes back unsold. It lives mostly in COD orders.
• A COD order is a yes with zero money down. The real decision happens at the door.
• Four gaps cause most returns: not found, changed mind, not home, too slow.
• A WhatsApp check before packing costs nothing and stops many returns.
• Make prepaid the easy choice. You do not have to stop COD.
• Ship within a day, and call on every failed delivery within 24 hours.
• Tag every return with a reason. Fix the biggest reason before the rest.
• Getting junk orders from ads too? Read our guide to fake leads from Meta ads.

Sources (checked 21 October 2026)

• MediaBrief: ShipNotes reveals 26% RTO rate on COD orders across India (Shipway ShipNotes FY25: under 2% of prepaid orders returned, nearly 26% of non prepaid, city range 18% Vadodara to 35% Patna, 22% / 27% / 35% by days to delivery attempt)
• Unicommerce: India D2C Report 2026 (built on 410M+ shipments across 6,000+ brands: COD 58% vs prepaid under 15% in the festive quarter, brands at 21% RTO by March vs others near 39%, prepaid incentive)
• Amazon Shipping India: RTO explained (what RTO and NDR mean, the main causes, acting on an NDR within 24 hours)

This article was published on madhurtyagi.sites.tab.bot, written by Madhur Tyagi, The Human Algorithm.

Here is her week. An agency runs her ads. A freelancer named Rohit built her WhatsApp replies. An intern sends the follow ups. Each hire made sense on the day she made it. Nobody planned for the gap between them. The gap is where the sale lives.

I see this setup all the time in growing Indian businesses. I have spent 8+ years in performance marketing, with 25+ brands since 2017. I have worked as a freelancer. I have done white label work for an agency, which means their name was on it and my work was inside. Today I work as a growth partner at The Human Algorithm. So I have no reason to tell you one of these is always right. Each one fits a different stage.

We will follow Kavya from that Monday to a happier one. On the way you get a plain table, seven questions to ask and a 5 question chooser. The answer might not be me.

Up next At a family dinner, Kavya's cousin asks her a question she cannot answer.

What is an AI growth partner?

Sunday dinner. Kavya's cousin from Bengaluru says, "Why don't you get an AI growth partner?" Everyone nods. Nobody at the table knows what it means. Kavya does not either.

Start with why the role exists at all. A sale is a chain. An ad brings a lead. The lead gets a reply. The reply gets a follow up. The follow up gets a payment. Each link can be done by a different person, and the chain breaks where two people meet.

An AI growth partner is one senior person, sometimes with a small desk behind them, who owns that whole chain, from the ad to the payment. They work inside your own accounts. They use AI to do the repeat work faster. And a human says yes to every decision about money.

Think of the kirana uncle near your home. He stocks the shelf, knows your name and keeps your udhaar in his notebook. One person owns the whole sale. A growth partner brings that back to a business that has grown too big for one notebook.

The difference from an agency is ownership. An agency is usually hired for a channel, like ads or content. A growth partner is hired for the result of the whole chain, and you can see which ad brought which sale.

The word "AI" matters less than people think. Stanford's 2026 AI Index reports that organisational use of AI reached 88%. Agencies and freelancers use it too. What changes is who checks the AI's work and who answers for the sale.

The term is new and used loosely. This is how I use it at The Human Algorithm.

Up next Back home, Kavya writes down what each of her three helpers really does all day.

What does each one actually do?

That night, Kavya opens her notes app. She writes one line for each helper. You can do the same.

A freelancer is one skilled person you hire for a task. A video editor, an ads specialist, someone to build a WhatsApp flow. Rohit is one. Good freelancers are fast, direct and fair on price, like the chai wala who makes one thing very well. You talk to the person doing the work. The catch: you are the manager. You write the brief, check the work and connect it to everything else.

India has a lot of this talent. NITI Aayog estimated 77 lakh gig workers in 2020 to 21 and expects 2.35 crore by 2029 to 30. That count includes delivery and ride work. NITI puts about 22% of gig work in high skilled jobs. Finding a freelancer is easy. Finding one who fits your business takes testing.

An agency is a team. Strategy, ads, design and reports, often under one account manager. That is the person who talks to you, not the person who runs your ads. Good agencies bring process, backup when someone is on leave, and the power to run a channel at volume. The catch: the scope, which is the list of work in the contract, ends where the contract says it ends. Your WhatsApp replies are rarely inside it.

An AI growth partner is closer to a part time growth head. One person owns the plan and the numbers. They work across ads, chats, follow ups and tracking, and use AI for drafts, sorting and reports. The catch: they can take few clients at a time. They are a poor fit if you need one task done cheaply.

Kavya looks at her list. Three good helpers. Not one of them has the word "sale" in their job.

Up next She puts the three side by side, and one row makes her stop.

How do they compare side by side?

Two notes on this table. The pay row shows how each is usually priced, not what you will pay. For sourced agency and freelancer rates in India, read what Meta ads agencies charge in India. And in all three columns, the ad account, the WhatsApp number and the data should be in your name. If anyone says otherwise, that is your answer.

Up next Kavya reads the “who owns the sale” row again. Then she finds a chat from last Thursday.

Why does "nobody owns the sale" cost so much?

Last Thursday at 9:12 PM, a woman from Indore messaged Kavya's page. She wanted kurtas for her daughter's wedding. The agency's ad had done its job. Rohit's welcome reply went out. Then nothing. The intern had gone home. The agency thought it was Kavya's lead. Kavya thought the intern had it.

Kavya finds the chat on Friday morning, with her chai going cold. The last message says, "No problem, found another shop." She reads it twice.

You know this feeling from your WhatsApp family group. A message lands. Everyone sees it. Nobody replies, because everyone thinks someone else will. Or think of cricket. A high catch goes up between three fielders. Each one waits for the other. The ball drops. When a reply is everyone's job, it is nobody's job.

This is old team math from Fred Brooks's book The Mythical Man Month. Every person you add brings more handoffs than you expect. Four people touch Kavya's lead: her, the agency, Rohit and the intern. Four people have six lines between them where a message can drop.

The most expensive drop is the reply. A Harvard Business Review study tested 2,241 US companies with an online lead. 37% replied within an hour, and 23% never replied at all. That data is from the US in 2011, and I see the same pattern in Indian chats today.

To count how many of your own leads go quiet, use the 7 step funnel leak audit. And if many of your leads are fake to begin with, fix that before you hire anyone. Here is how to spot fake leads from Meta ads.

The fix is not always a new hire. It is one name written next to each step. Who replies. Who follows up. Who checks which ad brought the sale.

Up next So whose name should go there? Kavya starts with the helper she knows well.

When is a freelancer the right hire?

Rohit built her WhatsApp replies in a week, on a clear brief, and they work. That is what a freelancer does well. Kavya does not need to replace him. She needs to stop asking him to own things that were never his job.

Pick a freelancer when:

• You can describe the task in two lines and judge the result yourself
• The task has a clear end, like 10 reels, a landing page or one WhatsApp flow
• You have a few hours a week to brief, review and connect the work

Copy this brief:

"I need [task] by [date]. Done means [what I can check]. Here is one example I like. Please send a rough draft by [date], and tell me what you need from me."

Up next Next, Kavya calls her agency and asks them the 9 PM question.

When is an agency the right hire?

Pick an agency when:

• One channel needs steady volume, like ads across many products or content every day
• You want a team, so nothing stops when one person is on leave
• You already have someone in house who owns the leads after the click

Copy this question for the pitch meeting:

"Who exactly will run my account each day, and can I talk to that person directly? And when a lead comes in at 9 PM, whose job is the reply?"

Kavya's agency is honest. "Replies are outside our scope," they say. That is fair. They are good at ads, and ads are what she pays them for. A good agency will answer both questions clearly. Just know the answer before you sign.

Up next So the ads are fine, and Rohit's replies are fine. What Kavya is missing is a person.

When is an AI growth partner the right hire?

This is the turn in Kavya's story. Read the list below and you will see her in every line.

Pick an AI growth partner when:

• Leads come in, but sales do not follow, and nobody can say where they drop
• You sell through chats, like Instagram DMs and WhatsApp, not just a checkout
• You want one person who answers for the numbers, not three reports

This is how I work with a D2C fragrance brand, which means a brand that sells straight to buyers online. I run it end to end: content, lead generation, follow up systems and the website. The result there is 3.5x+ ROAS on Meta held for 6+ months, with 35% lower cost per lead. ROAS means how much sales money comes back for each rupee spent on ads. Read the case study. Results vary by business.

The AI part is real but quiet. AI drafts replies, sorts leads and builds the weekly report. A person checks the work, and nothing about price or payment goes out without the owner's yes. If you want to see that kind of setup, here is how THA Agent Desk runs it. To see which jobs D2C brands give to AI and which they keep human, read how Indian D2C brands use AI agents.

Copy this question for any growth partner:

"Which three numbers will you report to me every week, and which one are you accountable for?"

Up next Before Kavya pays anyone, she sends seven questions on WhatsApp.

What should you ask before you hire anyone?

Kavya sends these on WhatsApp before any meeting. The answers tell you more than any pitch deck.

1. Whose name will the ad account, the WhatsApp number and the customer data be in?
2. Who replies to a new lead at 9 PM, and who follows up on day 3?
3. Which numbers will you report every week, and where will I see them?
4. What exactly is in the scope, in writing, and what is not?
5. How do you use AI, and who checks its work before it reaches a customer?
6. Who approves any message about price, discount or refund?
7. If we stop working together, what do I keep?

A fair answer to number 7 is "everything". If the answer is anything else, ask why.

Up next Kavya has her answer now. Five quick taps will give you yours.

Which one should you hire? Take the 5 question chooser

Can you mix a freelancer, an agency and a growth partner?

Yes, and many founders should. A mix works when one person owns the path and the others own tasks inside it. Two setups that work:

• Growth partner plus freelancers: the partner owns the plan and the numbers, and freelancers make reels or design on clear briefs.
• Agency plus one owner in house: the agency runs ads at volume, and one person on your team owns every lead after the click.

The setup that fails is the one from the top of this page: three hires, three reports, no owner.

So what does Kavya do? She does not fire anyone. She keeps the agency for ads and Rohit for small builds. Then she gives one person a single job: own every lead from the click to the payment, and tell her every Monday how many became sales. Same helpers, one owner.

A few weeks later, at 9:40 PM, a message comes in from Indore. It is the same woman. Now her niece is getting married. This time someone is paid to notice. A person replies that night. Kavya wakes up to an order and a voice note: "Your shop is so sweet, beta." She smiles over her chai. It feels like her nani's shop again.

That is the whole lesson in one line. A sale needs an owner more than it needs another helper.

Not sure where your own leads drop? Do the free Growth Leak Check before you hire anyone. It takes 2 minutes.

Your hiring checklist

Kavya kept this list on her fridge before she hired. Tick yours, then save it for your next meeting.

Sources (checked 21 October 2026)

• Stanford HAI: The 2026 AI Index Report (organisational adoption of AI reached 88%)
• NITI Aayog: India's Booming Gig and Platform Economy, policy brief, June 2022 (77 lakh gig workers in 2020 to 21, 2.35 crore expected by 2029 to 30, about 22% of gig work in high skilled jobs)
• Harvard Business Review: The Short Life of Online Sales Leads, March 2011 (2,241 companies tested, 37% replied within an hour, 23% never replied)
• Frederick P. Brooks Jr., The Mythical Man Month, 1975 (team size and communication lines)

This article was published on madhurtyagi.sites.tab.bot, written by Madhur Tyagi, The Human Algorithm.

Most D2C founders I talk to have stopped asking “should we use AI?”. They ask a sharper question now: “Which jobs can I hand over without my customers feeling it?”

That is the right question. An AI agent is good at repeat work with clear rules. It is weak at judgment, feelings and money. Get that split wrong and you save an hour but lose a customer.

I have run ads and lead systems for 25+ brands since 2017, and DM automation since 2018. Here is how I would split the work for an Indian D2C brand in 2026: five jobs an agent can take, the lines that stay human, and a 7 day way to start.

What does an AI agent actually do for a D2C brand?

An AI agent is software that reads a customer message, checks your own data and then drafts a reply or takes one small step. It might look up an order, check stock or read back an address. When it is not sure, it asks a person.

Think of it as a new hire on day one with a very good rulebook. It knows every policy you wrote down. It does not know what you would do when a loyal customer is upset.

Customers already expect this kind of help. In a Gartner survey of 3,566 customers run in February and March 2026, 58% of those who use GenAI had already used it to get a task done for them, not just to get an answer.

The same survey carries the warning. 87% said a company using GenAI for service must give them a way to reach a human. Keep both numbers in mind for the five jobs below.

Can an agent answer product questions at 2 AM?

The job: “Does it last all day?” “Is it okay for oily skin?” “When will it reach Lucknow?” These questions arrive late at night, and the same 30 come back again and again.

Some brands do not even give the question a place to land. When we opened 100 Indian D2C homepages on a phone on 19 October 2026, 17 had no WhatsApp link, no chat window and no signup field.

How an agent helps: you write a facts sheet once. Notes for each product, delivery days by region, return rules, COD charges. The agent answers from that sheet and nothing else. It says it is an AI assistant. When a question is not on the sheet, it says so and tells the buyer when a person will reply.

Keep human: anything about health, allergies or skin reactions. A wrong answer there is not a lost sale. It is a hurt customer.

Copy this (instruction for the agent):

“Answer from the facts sheet below and nothing else. Say you are an AI assistant in your opening line. If the answer is not on the sheet, say a person will reply by [time]. Never guess about skin, health or allergies.”

Copy this (reply the buyer sees):

“Hi! I am the [Brand] assistant, an AI. [Product] lasts about [hours from your sheet] on skin. Want me to share the travel size so you can try it?”

Should an agent confirm cash on delivery orders?

The job: a COD order arrives with half an address and a number that may be wrong. You ship it, it comes back, and you pay the courier both ways.

The numbers are not small. Unicommerce’s India D2C Report 2026, built on 410M+ shipments across 6,000+ brands, puts return to origin at 39.2% at the festive peak. The brands that brought it down made a few changes together, and one of them was checking the address before dispatch.

How an agent helps: it messages every COD buyer soon after the order, reads back the item, the amount and the pin code, and asks for one reply. A clear yes goes to packing. A changed address or a vague answer goes to a person.

Keep human: cancelling an order or marking a buyer as fake. The agent flags it. A person decides.

Copy this:

“Hi [name], this is the [Brand] order assistant, an AI. You ordered [product] for ₹[amount], cash on delivery, to pin code [pin]. Reply 1 to confirm or 2 to change the address.”

How should an agent handle abandoned carts?

The job: Baymard Institute averages 50 studies and puts the cart abandonment rate at 70.22%. Seven in ten carts are left behind.

Many brands answer with a discount blast. That teaches buyers to wait for the coupon.

How an agent helps: one reminder, a few hours later, that answers the doubt buyers usually have at that step (size, delivery days, COD, returns) and offers help instead of a discount. If the buyer replies, the agent answers from the facts sheet.

Keep human: discounts. Any coupon, offer or price change waits for your yes. Our own THA Agent Desk runs on the same rule: every price waits for a person.

Copy this:

“Hi [name], you left [product] in your cart. Most buyers ask about [size, fit or delivery] at this step, so here it is: [one line answer]. Need help choosing? Just reply here.”

Can an agent ask for reviews and reorders?

The job: after delivery, nobody asks how it went. Happy buyers forget to leave a review. Unhappy ones go to your Instagram comments.

How an agent helps: a few days after delivery, one short message asks how it is going. A happy reply gets the review link. When a pack is likely to run out, based on its size, one refill reminder goes out.

Keep human: every unhappy reply. The handoff is what makes people willing to use the agent at all. Salesforce found that 45% of consumers are more likely to use an AI agent when there is a clear way to reach a person.

Copy this:

“Hi [name], your [product] reached you on [day]. How is it so far? Reply 1 if you love it or 2 if something is off. If it is 2, a person from our team replies today.”

Can an agent run your weekly numbers?

The job: every Monday someone opens the store dashboard, the courier panel, the ads account and the WhatsApp inbox, and copies numbers into a sheet. It takes an hour and is out of date by Tuesday.

How an agent helps: it pulls the numbers into one short note: orders, COD share, RTO, ad spend and the questions customers asked most. Then it points out what changed. You decide what to do about it.

Keep human: the decision. An agent can tell you RTO went up in one city. It cannot tell you whether to stop COD there, because that depends on what you know about your buyers there.

Copy this (instruction for the agent):

“From this week’s order and ad sheets, list: total orders, COD share, RTO count, ad spend, the 5 questions customers asked most, and the 3 numbers that changed most since last week. Do not guess. If a number is missing, write ‘missing’.”

What should a D2C brand always keep human?

Here is my list. Keep it next to your yes rules.

1. Refunds, discounts and price exceptions. Money going out needs a person’s yes.
2. Complaints and angry customers. They want to be heard before they want a fix.
3. Health, skin and allergy questions. A guess here can hurt someone.
4. New offers and your brand voice. The agent copies your voice. It should not invent it.
5. Deals with creators, wholesalers and partners. Relationships are not tickets.
6. Cancelling an order or blocking a buyer. The agent flags. A person decides.

Two more rules sit on top of the list. Every agent message says it is AI, because nearly 75% of consumers told Salesforce they want to know when they are talking to an AI agent. And every chat shows a way to reach a person. Gartner’s 2026 survey found customers “will push back when AI acts as a barrier to reaching a human.”

An agent should take work off your team, not put a wall between you and your buyers.

This is the same lens behind our work with a D2C fragrance brand: 3.5x+ ROAS on Meta for 6+ months, with 35% lower CPL. That result came from ads and lead work, not from an AI agent. It taught me what this article is about. The ad brings the buyer. What happens in the chat after the click decides whether she stays.

How do you start in 7 days without losing trust?

Pick one job. Not five. Here is the week I would run.

• Day 1: choose the job. COD confirmation or product questions are the easiest to check.
• Day 2: write the facts sheet and the yes rules: what the agent may say, and what must wait for a person.
• Day 3: write 20 tricky test messages, in Hinglish too. “Bhai price kam hoga?” “COD pe extra charge hai?” “Gift wrap milega?”
• Day 4 and 5: run the tests and read every answer. When an answer is wrong, fix the sheet, not the single reply.
• Day 6: go live for that one job, with the AI label and a clear way to reach a person.
• Day 7: read the opening 50 real chats yourself. Then decide on the next job.

Want someone to build that one job with you? WhatsApp automation for D2C brands shows how we set up one flow, and you approve every word.

Back to Meera and her 2:07 AM buyer in Indore. With job one in place, the buyer gets an honest answer from Meera’s facts sheet, marked as AI, before she opens another brand’s page. A refund question that lands at 3 AM waits for Meera’s yes at 9. She wakes up to an order and one decision, not a lost sale.

Not sure which job to hand over? Leaks in replies and follow up usually sit closest to the money. The free Growth Leak Check asks 7 quick questions in 2 minutes and shows your top 3 leaks, each with one fix.

AI or human? Sort your own tasks

Your checklist before an agent talks to customers

• Every agent message says it is AI.
• The agent answers from a facts sheet you wrote and checked.
• Prices, discounts and refunds wait for a person’s yes.
• Complaints, health and skin questions go to a person.
• Every chat shows a way to reach a person.
• You tested 20 tricky messages, Hinglish included, before going live.
• You read the opening 50 real chats yourself.
• One job at a time. The next job waits until this one holds.

Sources (checked 21 October 2026)

• Gartner: Survey Finds 87% of Customers Say Companies Using GenAI for Customer Service Must Provide Access to a Human Agent, 4 August 2026 (survey of 3,566 B2B and B2C customers, February and March 2026: 87% want access to a human, 58% of GenAI users had it complete a task for them, the “barrier to reaching a human” line)
• Salesforce: State of the AI Connected Customer research, 31 October 2024 (nearly 75% of consumers want to know if they are talking to an AI agent, 45% more likely to use one with a clear escalation path)
• Unicommerce: India D2C Report 2026 (410M+ shipments, 6,000+ brands, 39.2% RTO at the festive peak, address verification before dispatch as one of the changes. Figures read from the report page, not the full PDF.)
• Baymard Institute: Cart abandonment rate statistics (70.22% average across 50 studies)
• The Human Algorithm: 100 Indian D2C homepages checked on a phone, 19 October 2026 (17 of 100 had no WhatsApp link, no chat window and no signup field)

This article was published on madhurtyagi.sites.tab.bot, written by Madhur Tyagi, The Human Algorithm.

“Meta is sending me fake leads.” I hear it from founders, coaches and D2C brands every week, usually with a screenshot of a lead sheet full of switched off numbers.

The anger makes sense. You paid for those leads. But in most accounts I open, “fake” is four different problems that look the same in a spreadsheet: bots, people who tapped by mistake, people who typed a wrong number on purpose, and real buyers who went cold while nobody called.

Each kind has a different fix. Treat them all as “fake” and you will fix the wrong thing.

I have run performance marketing for 8+ years and worked with 25+ brands since 2017, across ₹1Cr+ in ad spend for 9+ clients. Here is how I sort the junk from the buyers, and the six fixes I use, in the order I use them.

Are your leads fake, or just cold?

Before you change a single setting, find out which kind of bad lead you have. Pull your last 30 leads and put each one in a bucket:

1. Bot or spam: gibberish name, a number that is not a real mobile, the same details sent many times.
2. Accidental tap: a real person and a real number, but they say they never filled a form.
3. Wrong on purpose: a real name with a made up number, often on forms that offer something free.
4. Real but cold: they did fill the form, but you reached them a day or two later and they lost interest.

The 10 minute test: count each bucket. Most founders expect bucket 1 to win. In my experience, buckets 2 and 4 are often bigger, and cheaper to fix.

Copy this (for whoever calls your leads):

“After every call, mark the lead B, A, W or C: bot, accidental, wrong number or cold. Leave it blank if it was real.”

That one column will tell you more than any dashboard.

When Priya did this, the surprise was bucket 4. Most of her “fake” leads were real people. She had been calling them two days late.

Why does Meta send fake leads at all?

There is a popular theory that Meta adds fake leads to make your results look better. I have never seen evidence for it, and you do not need it to explain what happens.

The simpler answer is in your campaign settings. If your performance goal is “Maximize number of leads”, Meta’s own help page says your ad is shown to the people “most likely to share their contact information with you”. Not most likely to buy. Most likely to fill a form.

Then the form makes it easy. Meta can prefill name, email and phone number from a person’s Facebook or Instagram profile. Good for real buyers. It also means a curious thumb becomes a “lead” in two taps, and that person may truly not remember it.

So the fixes below do two jobs. They make the form a little harder to submit by accident, and they teach Meta what a buyer looks like.

1. Is your form too easy to submit?

The leak: the default form type, “More volume”, is built to get as many submissions as possible on a phone. Meta says so plainly. You get volume, and some of it is people who were half interested at most.

The 10 minute test: open your live form on your own phone. Count the taps from ad to submit. If it takes two taps and asks you nothing new, it is too easy.

The fix: switch the form type to Higher intent. Meta adds a review screen where people confirm their details, plus a line under the contact fields that says your business may follow up. Meta describes these as features that help you get more intentional submissions. Expect fewer leads and a higher cost per lead. Judge it on cost per real conversation instead.

Use the intro section too. Tell people what happens after they submit.

Copy this (form intro):

“Fill this in and our team will call you on this number to share prices and slots. Please enter a number you pick up.”

2. Do your leads verify their phone number?

The leak: anyone, or anything, can type ten digits into a form. Bots and people chasing a freebie both do it.

The 10 minute test: in your bucket count, add up B and W. If they are more than a few of your 30, you need this fix.

The fix: turn on SMS verification. With it, a person must enter a one time password (OTP) sent to their mobile before the form submits. Meta’s help page says the feature filters out “fake leads such as spam bots and bad actors”, and a landline cannot pass it. The feature is not offered in a few countries such as Pakistan and Sri Lanka, so check the box appears in your account.

The option appears when your form has a required phone number field (the normal one, not work phone).

Copy this (form intro line):

“We will send a code to your mobile to confirm it is you. It takes 10 seconds.”

3. Does your form ask one question a buyer can answer?

The leak: a form with just name and number tells you nothing about intent. Every lead looks equal, so your team calls the tyre kickers with the same energy as the buyers.

The 10 minute test: look at your form. Is there a single question where the answer would change who you call back today?

The fix: add one multiple choice question about timeline, budget or fit. Meta’s own guidance says more multiple choice questions usually bring fewer submissions but better quality leads, and it suggests asking about the purchase timeline. One or two questions is enough.

Copy this (pick one):

“When do you want to start? This week / This month / Just looking”

“What is your budget? Under ₹[amount] / ₹[amount] to ₹[amount] / Above ₹[amount]”

“Which city are you in? [City 1] / [City 2] / Somewhere else”

Call “This week” leads before anyone else.

4. Are you teaching Meta to find buyers or form fillers?

The leak: Meta learns from what you count as success. If you count every form as a win, it finds more people who fill forms.

The 10 minute test: in Ads Manager, open your ad set and check the performance goal. If it says “Maximize number of leads”, you are paying for form fills.

The fix: tell Meta which leads turned into conversations and sales. Meta offers a “Maximize number of qualified leads” goal: you connect your CRM so Meta learns which leads went further (since April 2026, new campaigns need a Conversions API connection for it). By Meta’s own numbers, lead ads that used the CRM connection with the qualified leads goal saw 21% lower cost per quality lead than ads that used “Maximize number of leads”.

Small setup? Start simpler. Keep the quality column, and build lookalike audiences from people who bought, not people who filled the form. Meta recommends that too.

Copy this (note for your ads person):

“Optimise for leads that became conversations, not for form fills. Use our buyer list for lookalikes, not the lead list.”

5. Are you calling every lead within an hour?

The leak: this is the one that makes real leads look fake. Someone fills your form on Sunday night. You call on Tuesday. They have seen five other ads since then. “I never filled any form” is often just “I forgot you”. That was Priya’s biggest bucket.

A Harvard Business Review study audited 2,241 US companies with a test lead. 37% replied within an hour, and 23% never replied at all. In a separate study of 1.25 million leads, firms that tried to reach a lead within an hour were nearly seven times as likely to have a real conversation with a decision maker as firms that tried even an hour later.

The 10 minute test: in your sheet, check the gap between when each lead came in and when someone called. Average it.

The fix: schedule your ads for hours when someone can respond. Meta itself suggests ad scheduling so your team is ready when leads arrive. Then send a short WhatsApp message from a person right after the form comes in, and call within the hour.

Copy this (WhatsApp, sent by a person):

“Hi [name], this is [your name] from [brand]. You asked about [offer] on Instagram a few minutes ago. Is now a good time for a 2 minute call, or should I share the details here?”

This message also reminds them who you are, so your call is not a stranger’s.

6. Do you score every lead by the ad it came from?

The leak: you judge ads on cost per lead. An ad with cheap leads looks like a winner, even if half its leads are junk.

The 10 minute test: split your quality column by ad. Does one ad or one placement bring most of the B and W leads?

The fix: every week, work out cost per real conversation for each ad: spend divided by the leads your team actually spoke to. Pause the ad with the worst number, even if its cost per lead looks great. If a placement keeps sending junk, test removing it.

Copy this (weekly review question):

“Which ad gave us the most real conversations for the money this week?”

Judging ads on real conversations instead of form count is the same lens behind our work with a D2C fragrance brand: 3.5x+ ROAS on Meta for 6+ months, with 35% lower CPL.

Which fix should you start with?

Priya did not fix all six in one week. She started with fix 5, because it cost nothing, then added fix 3. Her team now calls the “This week” leads before lunch, and her Friday review names the ad that brings buyers. Same budget, fewer wasted calls, and she stopped blaming Meta.

Your bucket count tells you where to start:

• Mostly bots or wrong numbers: fix 2, SMS verification.
• Mostly “I never filled any form”: fix 1 and fix 3, a higher intent form with one question.
• Mostly cold: fix 5. It costs nothing and you can start today.
• Leads fine, sales not: fix 4 and fix 6, so Meta and your team chase buyers.

Fake leads are rarely a Meta problem alone. They are half form and half follow up, and both halves are yours to fix.

Want to see where else your leads leak after the form? The free Growth Leak Check takes 2 minutes and shows your top 3 leaks, each with one fix.

Your fake lead checklist

Sources (checked 21 October 2026)

• Meta Business Help Center: About form types for lead ads (More volume vs Higher intent, the review screen)
• Meta Business Help Center: SMS verification for lead ads (filters spam bots and bad actors, landlines cannot verify, 90 days, countries where it is not offered)
• Meta Business Help Center: About performance goals for lead ads (what “Maximize number of leads” optimises for, the 21% lower cost per quality lead figure, the April 2026 change)
• Meta Business Help Center: About prefill questions (details taken from a Facebook or Instagram profile)
• Meta Business Help Center: Tips to create lead ads (multiple choice questions and quality, ad scheduling, lookalikes from customers)
• Harvard Business Review: The Short Life of Online Sales Leads, March 2011 (2,241 companies, 37% within an hour, 23% never, the 1.25 million lead study)

This article was published on madhurtyagi.sites.tab.bot, written by Madhur Tyagi, The Human Algorithm.

Neha makes face serums in her flat in Pune. Small batches. Pretty labels. She pays for her Meta ads from her savings.

She has one dream. Move the brand out of her bedroom and into a small studio on FC Road.

Her ads work. The messages come in. “Price?” “Is it good for oily skin?” “Do you deliver to Nagpur?”

But the sales do not come. So every week Neha does what most founders do. She blames the ads. New creative. New audience. Maybe a new agency.

I hear her story every week, from founders, coaches and D2C brands. The words are always the same: “Leads are coming. Sales are not.”

Sometimes the ads are the problem. More often, the ads did their job. Someone clicked. Someone asked. The sale was lost later, between that message and the payment. It was lost in the part of the business nobody is watching.

I have worked in performance marketing for 8+ years, with 25+ brands since 2017. I have run ₹1Cr+ in ad spend for 9+ clients. The pattern repeats: more leads poured into a leaking funnel just make a bigger puddle.

So let us walk with Neha through the audit I run. Seven questions, in the order a buyer moves. Each one comes with a 10 minute test, one fix and a line you can copy today.

Why does a lead go cold at all?

Start at the root. A lead is a person who showed interest. A DM, a form, a missed call. Inside each one is a question and a little hope.

Think of a chai stall at a busy railway station. You walk up. The chaiwala does not look at you. One minute. Two minutes. You walk to the next stall. You are not angry. You just wanted chai, and the next stall gave it to you sooner.

Your buyer works the same way. Hope fades with two things: waiting and effort. Every hour of silence is waiting. Every extra tap, question or confusing page is effort. When the waiting and the effort feel bigger than the hope, the person leaves. No complaint. No goodbye.

A funnel is the path from “who are you?” to “here is the money”. A leak is any spot on that path where people drop out. The seven steps below are the seven spots where I see it happen most.

What does each cold lead cost you?

Most founders count the leads they get. Very few count the leads they lose. So try this small sum.

Say 1 in 10 of your leads buys. Then every cold lead is a tenth of a sale walking away. Now say 20 leads a week go cold because nobody replied in time. That is two lost sales every week. Over a year, it is about 100 lost sales. Now multiply that by your price.

These are example numbers. Put in your own. Neha did this sum with her own numbers. It hurt more than any ad report.

How to run this audit in one hour

Neha opens a blank sheet. Do the same. Pull your last 20 leads from every place they came in: Instagram DMs, WhatsApp, lead forms, calls and email.

Then go through the seven questions below. Mark each one yes or no for your business. A small tick box sits at the end of every step to help you keep score.

Be strict, like the third umpire. “Mostly” is a no. You are not grading yourself. You are looking for the hole.

1. Can a stranger tell what you sell in 5 seconds?

Neha shows her Instagram page to her cousin for five seconds. “So you sell soap?” Neha sells face serums. Her heart sinks a little.

Why it leaks: a stranger does not read. A stranger glances. If that glance cannot tell what you sell, who it is for and what to do next, they leave. Or they send “details?”, and now you have a lead you must win twice.

The 10 minute test: show your Instagram profile or home page to a friend outside your industry for 5 seconds. Then ask three things. What do I sell? Who is it for? What should you do next? If they get two answers out of three, you have a leak.

Hidden prices make it worse. Baymard Institute studies online checkouts. In its research, 12% of US online shoppers said they had given up on an order because they could not see the total cost up front. Indian buyers who read “DM for price” face the same wall, plus a wait.

The fix: one line at the top of your profile or page. It says what you sell, who it is for and the next step. If your price changes, give a starting price or a range.

Copy this:

“Skin care for oily skin in humid cities. Ask us on WhatsApp and get your routine in one message.”

“12 week strength program for working women. Plans from ₹[price]. Tap below to check your fit.”

Neha changes her bio that night. Her cousin gets it in two seconds.

2. Does every post and ad give one next step?

Neha’s favourite reel is full of hearts and saves. She waits for orders. None come from it.

Why it leaks: people do what is clear and easy. A reel with no ask is a shop with no counter. People like it, then scroll on. Often the caption says “DM me” with no reason. Or the link in bio opens a page with nine buttons.

The 10 minute test: open your last 9 posts. Count how many end with one clear action a viewer can take in under 10 seconds.

The fix: one keyword ask per post. Add one link that opens the chat with a message already typed. That is called a prefilled link: you tap it, and WhatsApp opens with the words already written.

Copy this:

“Comment PLAN and I will send you the 7 day meal plan on WhatsApp.”

“Not sure of your size? Send SIZE on WhatsApp and we will reply with your fit.”

3. Does every lead land in one list?

Neha tries to count yesterday’s leads. Some are in Instagram. Some are in WhatsApp. There are three missed calls and one email from a shop in Goa. Just finding them all takes her twenty minutes.

Why it leaks: you cannot follow up with a person you forgot. Leads spread across five apps are like a cricket score kept on five scraps of paper. Nobody knows the real total, so nobody sees when a wicket falls.

The 10 minute test: count yesterday’s leads across every channel. If you need four apps to get the number, you found the leak.

The fix: one sheet or CRM with six columns: date, name, number, source, last message and next follow up date. A CRM is a contact list that remembers what you said and when. Every lead goes in on the day it arrives, from every channel.

Copy this (for your team):

“If a lead is not in the sheet, it did not happen.”

4. Does every lead hear back within an hour?

Neha opens her new sheet and finds her Sunday night buyer. The message sat for a day and a half. Nobody answered it.

Why it leaks: hope cools fast. While you are busy, asleep or on a call, the price question waits. By the time you reply, the buyer is cold or has bought somewhere else.

This is the leak with the strongest research behind it. A Harvard Business Review study checked 2,241 US companies with a test lead. 37% replied within an hour. 23% never replied at all. The average reply took 42 hours.

A separate study in the same article looked at 1.25 million leads. Some firms tried to reach a lead within an hour. They were nearly seven times as likely to qualify it as firms that tried even an hour later. To qualify a lead means to find out if the person is a real buyer. Against firms that waited 24 hours or longer, the gap was more than 60 times.

That data is from the US in 2011. Think of your family WhatsApp group. When your mother asks something, she expects a reply soon. Your buyers in 2026 are no more patient than she is.

The 10 minute test: ask a friend to message your business at 9 PM with “what is the price?”. Note when they hear back.

The fix: fixed reply times, at least three a day. Save one warm opening reply, and have a person send it. It answers the question and asks one easy question back. To see what slow replies may cost you in rupees, try the Reply Leak Calculator.

Copy this:

“Thanks for asking, [name]! [Program] is ₹[price] for [duration]. One quick question so I guide you right: are you hoping to start this month?”

Neha picks three reply times: 10 AM, 2 PM and 8 PM. She saves her opening reply and sends it herself. For the full method, read my guide to answering “what’s the price?” in DMs.

5. Does every maybe get a planned follow up?

Neha scrolls her chats. A whole screen of people said “let me think about it”. She never wrote back to most of them. She did not want to sound pushy.

Why it leaks: a maybe is not a no. People get busy and forget. A kind reminder feels like the kirana uncle who says, “the atta you asked for is back”. You do not feel sold to. You feel remembered. Most businesses send one follow up if someone remembers. Then the lead dies in the chat list.

The 10 minute test: in your sheet, find every lead from the last 14 days who went quiet. How many got a second message? How many got a third?

The fix: a planned rhythm with an end. Day 1, Day 3 and Day 7. Each message adds something useful and asks one easy question. The Day 7 message closes the loop politely.

Copy this (Day 3):

“Hi [name], one quick one: is [their goal] still on your list for this month? A yes or a no is completely fine.”

The Day 1 and Day 7 messages are in my guide to follow up messages that do not feel pushy.

6. Can a yes pay through one link?

A buyer finally says yes to Neha. Then the back and forth starts. “Send your UPI ID.” “Which address?” “Did my payment go through?” Halfway, the buyer goes quiet.

Why it leaks: every extra step is one more chance to change their mind. For D2C, cash on delivery (the buyer pays when the parcel arrives) has its own trap. An order nobody confirmed can come back as a return.

Baymard Institute found that 17% of US online shoppers had given up on an order because the checkout was too long or complicated. 9% gave up because there were not enough payment methods.

The 10 minute test: say yes to your own business. Count the taps and messages between “I am in” and money received.

The fix: one payment or booking link with the amount filled in. Send it in the same message as the yes. For cash on delivery, confirm the order on WhatsApp before you ship.

Copy this:

“Great, here is your link for ₹[amount]: [link]. Once it goes through, I will send your start date right here.”

“Before we ship your order, can you reply YES to confirm your address and ₹[amount] on delivery?”

7. Can you name the ad behind your last 5 sales?

Neha runs two ads. Ad A brings cheap leads. Ad B costs more per lead, so she is about to switch it off. Then she checks who actually paid. Most of her buyers came from Ad B.

Why it leaks: cost per lead (CPL) is what you pay for one enquiry. It tells you how cheap a lead is, not whether they buy. Judging ads on CPL is like judging a batter by the balls faced, not the runs scored. You switch off the ad that brings buyers and keep the one that brings cheap leads who never pay. I explain why in my guide to cost per lead on Meta ads in India.

The 10 minute test: write down your last 5 paying customers. Next to each, write the ad, post or referral that brought them. Every blank is a blind spot.

The fix: use the source column from step 3. Ask every new lead where they found you. Give each ad its own prefilled WhatsApp message, so the opening words tell you which ad sent them.

Copy this (prefilled message for one ad):

“Hi, I saw your reel about [topic] and want to know more.”

We judge ads on buyers, not leads, in our work with a D2C fragrance brand. That account held 3.5x+ ROAS on Meta for 6+ months, with 35% lower CPL. ROAS means rupees of sales for every rupee spent on ads.

Where should you start?

Three Sundays later, at 11:40 at night, a new message reaches Neha. “What is the price?”

This time the message lands in her sheet. At 10 AM on Monday, Neha replies herself, with the price and one easy question. By lunch, the buyer has paid through one link. Neha does not get rich that day. But she feels something she has not felt in months: in control. The studio on FC Road feels a little closer.

Neha did not fix all seven leaks in one go. Neither should you. Pick the weakest step that sits closest to the money. Steps 4, 5 and 6 sit right next to the payment. A leak there costs you buyers who already said they were interested.

Fix one leak. Run it for a week. Count your sales from the same number of leads. Then move to the next one. Spend more on ads after the funnel holds water, not before.

Want a score without doing this by hand? The free Growth Leak Check asks these same seven questions in 2 minutes. It shows your top 3 leaks, each with one fix.

Your funnel leak audit checklist

Tick what is true for your business today. Any step you ticked above is already here.

Sources (checked 21 October 2026)

• Harvard Business Review: The Short Life of Online Sales Leads, March 2011 (2,241 companies audited, 37% replied within an hour, 23% never replied, average 42 hours, and the 1.25 million lead study)
• Baymard Institute: Cart abandonment rate statistics (reasons US online shoppers gave for abandoning an order: total cost not shown up front 12%, checkout too long or complicated 17%, not enough payment methods 9%)

This article was published on madhurtyagi.sites.tab.bot, written by Madhur Tyagi, The Human Algorithm.

If you sell coaching or courses through WhatsApp, your messaging bill changed on 1 October 2026. Replies that were free are now charged once you pass a monthly free tier.

The change is small per message, but it adds up when you run launches, reminders and broadcasts from one number. Here is what changed, what each message costs in India, and what a typical month might look like.

Everything below comes from Meta's own developer documentation and its rupee rate card, checked on 20 October 2026. These rates apply to the WhatsApp Business Platform, the API that business tools connect to. Meta's rate card does not cover the free WhatsApp Business app on your phone.

What changed on 1 October 2026

Meta's pricing page, updated 30 September 2026, lists two messages that became charged again:

1. Service messages. These are normal replies (not templates) sent inside the 24 hour window after a customer writes to you. They had been free since 1 November 2024. Now each business phone number gets 1,000 free service messages a month, and Meta charges from the 1,001st.
2. Utility messages sent in reply to a customer. These are templates like a booking confirmation sent while the 24 hour window is open. They had been free since 1 July 2025. They are now charged.

The free tier is per phone number, resets every month, and does not roll over.

Meta also lists a fifth message category, Meta Business Agent, for replies written by Meta's own AI agent. It is charged differently (see below).

WhatsApp rates per message in India (October 2026)

From Meta's INR rate card, effective 1 October 2026, for messages delivered to Indian numbers (+91):

A few rules from the same page:

• Meta charges per delivered message, not per sent message.
• Messages your customers send you are not charged.
• Volume discounts apply to utility and authentication messages, not to marketing or service.

When Meta does not charge

Meta's page names two cases:

1. The free entry point window. When someone messages you from a Click to WhatsApp ad on the WhatsApp phone app and you reply within the 24 hour window, that reply is not charged and opens a free entry point window. Meta says it can stay open for up to 7 days, and while it is open Meta does not charge for marketing, utility, authentication or service messages. Meta Business Agent replies are still charged.
2. The monthly free tier of 1,000 service messages per business phone number.

What a coach might pay in a month (a worked example)

This is an illustration, not a quote. Say your business number sends these in one month to Indian numbers, none from an ad window:

In this example the one broadcast costs far more than all the replies put together. For most coaches, marketing templates are where the bill grows.

AI replies cost more: Meta Business Agent

Meta charges Meta Business Agent replies at one global rate of $2.00 per 1 million tokens. Meta says one message typically uses 20,000 to 25,000 tokens, which it puts at about 4 to 5 US cents per message. These replies get no free tier and no free entry point window.

In the example above, if 300 of the 1,400 replies were written by Meta Business Agent instead of a person, Meta's estimate puts those 300 at about $12 to $15. Your service count would drop to 1,100, so your typed replies would cost ₹11.50.

A person typing a reply costs ₹0.115 after the free tier. An AI agent reply costs roughly 4 to 5 US cents. That gap is worth knowing before you switch AI replies on for every chat.

How to keep your WhatsApp bill low

1. Reply while the 24 hour window is open. Inside it, normal replies are service messages, with 1,000 free a month.
2. Send fewer, better broadcasts. One marketing template to a large list costs more than hundreds of replies.
3. Use the right template category. A real booking reminder is utility, not marketing. Meta has its own rules for how templates are categorised.
4. Use Click to WhatsApp ads wisely. Chats that start from these ads open a free entry point window.
5. Keep a payment method on file. Meta says that without one, it stops delivering service messages once the free tier is used up.

For what to actually say in those replies, see my guides on answering "what's the price?" in DMs and follow up messages that don't feel pushy. Ready made scripts are in the free Script Vault.

Want to see where your own chats leak before you spend more on messages? The free Creator Leak Check takes 2 minutes.

Billing in rupees: a date to note

Meta's page says businesses whose billing country is India must move their messaging accounts to INR billing by 31 December 2026. From 1 January 2027, Meta says it will no longer deliver messages from non INR accounts of those businesses. If a partner runs your WhatsApp number, ask them whether your account is already billed in rupees.

A note on older pages

When we checked on 20 October 2026, Meta's WhatsApp Business marketing page still said service messages are not charged. The developer documentation, updated 30 September 2026, describes the new charges. When the two differ, go by the dated developer documentation and the rate card.

For the full DM to WhatsApp to payment flow, read my guide on getting coaching clients on Instagram.

Check your WhatsApp setup

Sources (checked 20 October 2026)

• Meta for Developers: Pricing on the WhatsApp Business Platform, updated 30 September 2026 (categories, the 1,000 free service messages, the free entry point window, Meta Business Agent rate, INR billing dates)
• Meta's INR list rates file, "Cost per message in INR on the WhatsApp Business Platform, effective October 1, 2026", linked from the page above (India row)
• WhatsApp Business: Platform pricing (the marketing page mentioned above)

This article first appeared on madhurtyagi.sites.tab.bot, written by Madhur Tyagi, The Human Algorithm.

"What's the price?" is the most common message a coach or creator gets in their DMs. It is also the one most often answered badly.

Some people paste a number and wait. Some dodge the question with "DM me for details" inside the DM itself. Both lose good leads every week.

A price question is a buying signal, not a nuisance. The person has seen enough to wonder if they can afford you. Your reply decides whether that thought turns into a conversation or a silent scroll away.

I have written and reviewed reply flows for 25+ brands since 2017. The pattern below is the one I keep coming back to, because it respects the buyer's time and yours.

Why "what's the price?" deserves a real answer

A person who asks your price has moved from "is this interesting?" to "is this for me?". That is a big step.

Two common replies throw it away:

1. The bare number. "₹[amount]." With no context, the number has nothing to stand next to. The lead compares it with a free video and goes quiet.
2. The dodge. "Let's get on a call and I'll explain." To someone who asked a simple question, this reads like a sales trap.

People do not object to your price. They object to a price they cannot place. Your job is to give the number a frame: who it is for, what it includes, and what happens next.

The 4 step price reply (copy these)

Change the words in [brackets] and send. Each step is one short message.

Step 1: Thank them and promise the price

Thanks for asking, [name]! Happy to share the price. Two quick questions so I point you to the right option.

This tells them you are not hiding anything. The price is coming.

Step 2: Ask two questions

1. What would you like to change in the next 90 days?
2. Are you looking to start this month, or later?

Two questions, not five. You learn their goal and their timing, and they feel heard.

Step 3: Give one price for the one option that fits

Thanks! Based on that, [program name] is the right fit for you. It is [duration], includes [3 key things], and the fee is ₹[price].

One option, one price. A menu of three plans in a DM makes people stall. If you have several offers, pick the one that fits their answers.

Step 4: Offer one next step

Would you like the payment link, or a quick 15 minute 1:1 call on Zoom to ask questions before you decide?

Give a choice between two yeses. Both move the conversation forward.

More ready messages for this moment are in the free Script Vault: how to explain your offer in one message and the payment message to send when they say yes.

Want to see where your own DMs lose people? The free Creator Leak Check takes 2 minutes and gives you 3 fixes.

What the full exchange looks like

Notice what the coach never does: argue, pressure, or send a wall of text. Every message is short and ends with an easy reply.

What to say when they reply "too expensive"

"Too expensive" usually means "I am not yet sure it is worth it", not "I cannot pay". Ask before you discount:

Totally fair. Can I ask what feels like too much: the total amount, or paying it all at once?

Then answer the real worry:

• The total. Point to a smaller option if you have one, or what they can start with on their own today.
• Paying at once. If you offer instalments, mention them here.
• Not sure it works. Share one honest client story with their permission, or offer the 1:1 call.

Never cut your price in the chat to save a sale. Buyers who got a discount for asking will ask again next time, and so will their friends.

When they go quiet after the price

Silence after a price is normal. People check with a partner, look at their budget, or simply get busy.

Follow up three times over a week, then stop. My guide to DM follow up messages that don't feel pushy has a Day 1, 3 and 7 message for exactly this.

The India version: "Price kya hai?"

In India the same question often arrives as "Price kya hai?" or "Kitne ka hai?". The framework does not change. Match the person's language in your reply and keep it short:

Bilkul, price share karta hoon! Bas 2 quick sawaal, taaki sahi option bataun.

If the chat moves to WhatsApp, remember that a business can send free form replies there for 24 hours after the person's last message. After that, it needs an approved template, which Meta charges for (Meta's WhatsApp documentation; my guide to WhatsApp pricing in October 2026 has the India rates). So reply while the window is open.

For the full path from an Instagram DM to WhatsApp to a UPI payment, read my guide on getting coaching clients on Instagram.

Check your own price reply

Sources (checked 20 October 2026)

• Meta for Developers: Pricing on the WhatsApp Business Platform (customer service window), updated 30 September 2026

This article first appeared on madhurtyagi.sites.tab.bot, written by Madhur Tyagi, The Human Algorithm.

A lead asked about your program, you replied, and then nothing. No "no", no "yes". Just two grey ticks.

Most coaches do one of two things next. They never message again, or they send "Hi, any update?" every other day until the lead blocks them.

There is a calmer way. Three short messages over one week, each one more helpful than the last, then a polite stop.

I have built follow up systems for 25+ brands since 2017. The single habit that matters is not clever wording. It is following up at all, on a fixed rhythm, and stopping on time.

Why leads go quiet (it is rarely a no)

People stop replying for ordinary reasons:

• They got busy and meant to reply later.
• They need to check with a partner or look at their budget.
• They are not sure the offer fits them, and do not want to say so.
• Your last message was hard to answer.

Silence is usually "not now", not "no". A good follow up makes "now" easier.

4 rules that keep a follow up from feeling pushy

1. Every message adds something. A tip, an answer, a short story. Never a bare "any update?".
2. One easy question per message. A question that takes five seconds to answer gets answered.
3. No guilt, no fake urgency. No "last chance", no countdowns that are not real.
4. A clear end. After the Day 7 message, you stop. That is what keeps you trusted.

Day 1: add value, ask one easy question

Send this about a day after they went quiet.

Hi [name], I was thinking about what you said about [their goal]. One thing that helps most people here is [one useful tip]. Did you get a chance to look at the details I sent?

You are not chasing. You are continuing the conversation with something useful.

Day 3: make replying easy

Two days later, offer a choice they can answer in one word.

Hi [name], quick one so I don't keep messaging you: are you still thinking about [program], or is now not the right time? Either answer is completely fine.

"Either answer is fine" is the line that gets replies. It gives people permission to say no, and many of them say yes instead.

Day 7: close the loop

Hi [name], I'll stop messaging about [program] so I don't clutter your chats. If you want to start later, just reply "start" here and I'll pick it up. Rooting for you with [their goal]!

This message often gets more replies than the other two, because it is clearly the last one.

More follow up lines are in the free Script Vault: the WhatsApp follow up message template and what to reply to DMs you missed.

Want to see where your own DMs leak? The free Creator Leak Check takes 2 minutes and gives you 3 fixes.

Following up on WhatsApp: mind the 24 hour window

On WhatsApp, a business can send normal (free form) messages for 24 hours after the person's last message. Each new message from them resets that clock. Outside the window, a business on the WhatsApp Business Platform needs an approved template message (Meta's WhatsApp documentation, checked 20 October 2026).

What this means in practice:

• Day 1 often lands inside the window, so a normal message works.
• Day 3 and Day 7 usually fall outside it on the platform, so they need approved templates. Templates are charged per message. My guide to WhatsApp pricing in October 2026 shows what each one costs in India.

What never to send

• "Hello??" or "Any update?" with nothing else.
• A discount you did not plan, sent to get a reply.
• A long paragraph repeating your whole offer.
• Messages after the person said no or asked you to stop.

Respecting a no is part of the follow up, not a failure of it.

When the follow up is about price

If the lead went quiet right after you shared your price, the Day 1 message can answer the question they probably have:

Hi [name], one thing people often ask after seeing the fee: [the common worry, answered in one line]. Happy to explain anything else.

My guide on how to reply when a lead asks "what's the price?" covers the price message itself.

Check your follow up system

Keep a simple tracker

You do not need software to start. A sheet with five columns is enough: name, where they came from, last message date, next follow up date, and status (open, won, closed). Check it at the same time every day.

For the full path from DM to payment, start with my guide on getting coaching clients on Instagram.

Sources (checked 20 October 2026)

• Meta for Developers: Pricing on the WhatsApp Business Platform (customer service window and template messages), updated 30 September 2026

This article first appeared on madhurtyagi.sites.tab.bot, written by Madhur Tyagi, The Human Algorithm.

Imagine you hire a new assistant who never gets tired, never forgets, and reads and writes very fast. There is one catch: they do exactly what your checklist says, even when the checklist is wrong.

That assistant is AI automation.

So the real question is not "which AI tool should I buy?" It is "which of my tasks has a clear enough checklist to hand over?"

This guide answers that in plain words. You get the 5 tasks I would hand over before anything else, a 4 question test to pick your own, and the order that stops you from automating a mess.

I have worked on growth and follow up systems for 25+ brands since 2017. The founders who get real time back from AI are rarely the ones with the fanciest tools. They are the ones who picked small, boring, repeated tasks close to money, and kept a human yes at the end.

What does AI automation mean for a founder?

AI automation means a repeated task in your business runs on its own steps, with AI doing the reading and writing part, and simple rules doing the moving part.

Every automation, however fancy it sounds, has the same four parts. Think of a recipe:

1. The trigger. What starts it. A new lead fills a form. A payment lands. It is Monday 9 AM.
2. The steps. What happens, in order. Read the message, look up the customer, pick the right reply.
3. The output. What comes out. A drafted reply, a row in a sheet, a one page report.
4. The check. Who looks at it before it reaches a customer. For anything a customer reads, that check is you or your team.

If you cannot write down all four parts for a task on one page, it is not ready to automate yet. That single test saves most founders weeks of wasted setup.

Before you automate: question, delete, simplify, speed up, automate

Automating a messy process is like fitting a faster engine to a car with a flat tyre. You arrive at the wrong place sooner.

Walter Isaacson's 2023 biography Elon Musk describes a five step order Musk uses on any process. It works just as well for a ten person business:

1. Question the step. Who asked for it? Does a customer care?
2. Delete it if you can. The cheapest automation is a step that no longer exists.
3. Simplify what is left. Fewer fields, fewer approvals, one template.
4. Speed it up. Make the manual version fast and smooth.
5. Automate it last. Now the checklist is short and clear, so AI can follow it.

The free THA Playbook walks through this 5 step audit with examples.

Most founders jump straight to step 5. That is why so many automations get switched off a month later.

How to pick which task to automate: the 4 question test

Ask four questions about any task:

• Does it happen often? Daily or weekly beats once a quarter.
• Does it follow the same steps every time? If you can write the checklist, AI can follow it.
• Does it sit close to money? Leads, follow ups and payments beat internal tidying.
• Is a mistake easy to catch and undo? A drafted reply that you approve is low risk. A refund sent by a bot is not.

Four yeses means automate it soon. Two or fewer means leave it alone for now. Try it on one of your own tasks:

The 5 tasks to automate before anything else

These five pass the test for most founders, coaches and creators I talk to in India. They all sit between "someone showed interest" and "money landed", which is where most businesses quietly lose sales.

1. Draft the reply to every new lead

A lead comes in from an ad form, a website form, an Instagram DM or WhatsApp. AI reads what they asked, drafts a short reply in your voice with one helpful answer and one question, and puts it in front of you. You read it and tap yes.

Why it leads the list: interest fades fast. Research published in Harvard Business Review, The Short Life of Online Sales Leads (March 2011), found that most companies respond to online enquiries far too slowly. A ready draft removes the slowest part, the blank page.

What you still do: approve every reply, and answer anything about price, health, money or a complaint yourself. The message says it was written with AI help when a bot or AI is involved.

Hi Ankit, thanks for asking about the 8 week program. Quick question so I send the right details: are you starting from zero, or have you trained before?

2. Follow ups that are planned in advance

Most sales happen after the second or third message, and most founders never send it. "Seen" is not "no".

Plan the follow ups once: day 1, day 3, then stop. AI drafts each one from the conversation so far, and it goes out after your yes, or on the schedule you approved. Every follow up ends with an easy way to say no.

Why it goes early: it costs nothing to build and it rescues leads you already paid for. My guide on cost per lead shows why a lead that never hears back is the most expensive lead you have.

3. One list for every lead

Leads arrive in five places: ad forms, DMs, WhatsApp, email, calls. Most founders cannot answer "how many leads came in last week, and from where?"

Automate the boring part: every new lead lands in one sheet or CRM with name, contact, source, what they asked, and status. AI can summarise the conversation into one line so you can scan 30 leads in a minute.

Why it matters: you cannot fix a leak you cannot see. Once every lead sits in one list, you know your real ROAS and cost per customer, not just what the ad dashboard says.

4. Payment check and welcome message

Someone pays by UPI or a payment link. Today you check the app, find the chat, type a thank you, and send next steps. Sometimes a day later.

Automate it in two halves. The payment is matched to the customer in your list. A welcome message with next steps is drafted and waiting for a quick check. You confirm the money landed before anything goes out.

Why it matters: the hour after payment sets the tone for the whole relationship. A clear welcome cuts "did you get my payment?" messages and refund requests.

5. A weekly one page numbers report

Every Monday, one page lands in your inbox or WhatsApp: leads by source, replies sent, follow ups sent, calls booked, payments, and the step where most people dropped off.

AI does the counting and writes three plain lines on what changed. You spend five minutes reading and pick one thing to fix that week.

Why it matters: this is the task that makes the other four better over time. Without it, automation runs in the dark.

What not to automate yet

Keep a person on these, even if a tool promises it can do them:

• Price negotiation and discounts. Your margin is not a template.
• Complaints and refunds. An upset customer needs a human who can decide.
• Health, legal, money or tax advice. Rules exist for a reason, and a confident wrong answer hurts people.
• The sales call itself. AI can prepare your notes. The trust is yours to earn.
• Anything you have not done by hand at least ten times. If you do not know the steps yet, neither will the AI.

How to keep AI automation honest

Three rules I follow for every system I build:

1. Say it is AI. If a bot or AI drafted a message, the customer can know. People forgive an honest assistant. They do not forgive one pretending to be you.
2. Keep a human yes. Anything a customer reads gets approved by a person, at least until the drafts are consistently right.
3. Log everything. Every message, every change, every yes. When something goes wrong you can see why, and fix the checklist.

I wrote more about this on how we keep AI human.

A 7 day starter plan

• Day 1: List every task you did twice this week. Run each through the 4 question test.
• Day 2: Pick one task. Question it, delete what you can, simplify the rest.
• Day 3: Write its recipe on one page: trigger, steps, output, check.
• Day 4: Do it by hand with the new checklist. Note where you hesitated.
• Day 5: Set up the automation for that one task, with your approval before anything goes out.
• Day 6: Run it on real leads. Read every draft.
• Day 7: Fix the checklist. Then, and not before, pick task two.

One working automation beats five half built ones. If leads are where you want to start, my guide for coaches and creators has message scripts you can use as the checklist, and WhatsApp follow up for D2C brands shows the same idea for an online store.

This article first appeared on madhurtyagi.sites.tab.bot, written by Madhur Tyagi, The Human Algorithm.

Written on 19 October 2026. Every price below is a third party figure with its source and the date I checked it. Prices move, so read them as "true as of October 2026". Rupee amounts use ₹94.96 to the US dollar, the reference rate for 16 October 2026 (ECB rate via Frankfurter).

Think of your ads like a car trip. The ad spend is the fuel. It goes to Meta, every rupee of it. The agency is the driver. You pay the driver to pick the route, watch the road and not waste fuel.

So the real question is never "how much does the driver cost?" It is "how much fuel does this driver save, or turn into sales?" A cheap driver who burns fuel on wrong turns costs more than a good one.

This guide gives you the real numbers I could verify, explains how agencies price their work, and shows you how to compare two quotes without getting fooled.

How much does a Meta ads agency charge in India in 2026?

There is no official price list. What exists are marketplaces and directories that publish rates. Here is what they showed when I checked on 19 October 2026. Every row is a third party figure, not my price.

• Indian freelancers for Facebook and Meta Ads: typical rates of $10 to $15 an hour, about ₹950 to ₹1,425. Source: Upwork, India page (third party).
• The same role in the US, for comparison: $54 to $94 an hour, about ₹5,128 to ₹8,926, on the same Upwork page (third party).
• Indian social media marketing companies: a median hourly rate of $25, about ₹2,374, across 8,578 companies reviewed. Source: GoodFirms, India (third party).
• Indian PPC agencies on Clutch, page one: of the 56 that list a rate, 34 list under $50 an hour (under about ₹4,748), and 43 list a minimum project of $1,000+ (about ₹94,960). Source: Clutch, India PPC (third party; my count).

Three things to know before you use these numbers:

• Upwork's range is for freelancers, from its own historical contracts. Upwork says what clients pay is subject to negotiation.
• Directory rates are what agencies list, not what they charge you. A listed rate is a starting point for a talk.
• Clutch page one is a sample, not a survey. It includes sponsored listings, so I counted it as a rough picture of how agencies there list themselves. I did the counting myself on 19 October 2026.

What that looks like each month (my arithmetic, not a quote)

Agencies rarely bill Indian small businesses by the hour. But hours help you sense check a monthly fee. Say a small account needs about 20 hours of work a month for setup checks, new ads, tracking and reporting. That number is my assumption; your account may need more or less.

• 20 hours at Upwork's $10 to $15 is about ₹19,000 to ₹28,500 a month.
• 20 hours at GoodFirms' median $25 is about ₹47,500 a month.

If a monthly quote is far below these, ask what is left out. If it is far above, ask what extra work you get. Neither is wrong by itself.

What large brands pay big media agencies

Very large advertisers usually pay big media agencies a commission on spend. According to a Reuters report published by The Economic Times (19 June 2025), an industry guideline circulated in August 2023 set a minimum commission of 3% on digital ads for clients spending more than $29 million a year, and up to 8% for smaller clients. India's competition regulator is examining that guideline, so treat it as a reported data point, not a market rate. It also has little to do with what a growing D2C brand or a coach pays.

The 3 ways agencies price Meta ads work

1. A fixed monthly fee (retainer)

You pay the same amount each month for an agreed scope. Simple to plan. The risk: the fee says nothing about how much work you get, so the scope has to be written down.

2. A percent of ad spend

The fee is a share of what you spend on Meta. It grows when your budget grows. That can be fair when more spend truly means more work, like more creative and more testing. It can also reward spending more, whether or not it brings sales.

3. A mix, sometimes with a results bonus

A base fee plus a percent above a certain spend, or a base fee plus a bonus tied to an agreed result. If you agree a bonus, tie it to a number you can check yourself, like delivered orders or qualified leads, never just clicks or reach.

On top of any of these you may see a one time setup fee, and creative (new videos and images) may be priced separately. Ask.

Here is the simple way to see it. A fixed fee is a flat line: the same at ₹1 lakh or ₹5 lakh of spend. A percent fee is a rising line. Where the two cross depends on the numbers in your quotes. Put yours into the calculator below.

Work out your real monthly cost

What the quote usually does not include

• The ad spend itself. That goes to Meta. An agency fee is never the whole cost of advertising.
• GST on the ads. Meta says that for ad accounts with India as the business country, ads are billed by Meta India in rupees and "GST and TDS are added whenever you're charged for your ads" (Meta Business Help Center).
• GST on the fee. Advertising services are taxed at a standard GST rate of 18% according to Razorpay's GST guide (third party). Ask whether a quote includes GST. If you are GST registered, ask your accountant about input tax credit.
• Creative. Many quotes cover running ads, not making them. With Meta now rewarding ads that are truly different, creative is a big part of results. My Andromeda guide explains why.

Why two quotes are so hard to compare

Two agencies can quote ₹15,000 and ₹35,000 for what sounds like the same job. Often it is not the same job. One sets up campaigns and sends a report. The other also makes new ads every month, fixes your tracking and reviews results with you.

The fix is simple. Write one scope list and send the same list to everyone: number of campaigns, new creatives each month, tracking setup (Pixel and Conversions API), report frequency, review calls, and who makes the ads. Now the prices mean something.

Freelancer, agency or in house: which fits your stage?

• A freelancer suits a small, steady account where one skilled person can do the work. Ask who covers when they are away.
• An agency suits you when you need several skills at once: media buying, creative, tracking and reporting.
• In house suits you when ads are central to the business and spend is large enough to keep a person fully busy.

Whatever you pick, judge results on money: cost per delivered order for D2C, or cost per qualified lead for services. My guides on good ROAS with COD and RTO and on cost per lead show how to work out your own break even numbers.

Before you sign: 8 checks

Warning signs in a pitch

• A fixed ROAS promise before they have seen your numbers. Nobody can know your result before seeing your margins, prices and data.
• The ad account or Pixel sits in their business, not yours. If you part ways, you lose your history.
• Reports full of reach and clicks, with no sales or qualified leads.
• No access for you to Ads Manager.
• A long lock in with no exit before you have seen any work.

None of these makes someone dishonest. They just leave you with less control over your own money.

What I charge, and a small way to start

I have run Meta ads for 8+ years, for 25+ brands, since 2017 (my own record). For one D2C client (unnamed), we held a 3.5x+ ROAS for 6+ months and brought cost per lead down by 35% (my client result, not a promise for your account).

Before you hire anyone, me included, find where your money is leaking. That is what my Growth Leak Audit is for: a 45 minute 1:1 call on Zoom, recorded, and a 1 page plan within 24 hours, for ₹9,999 (my price). I am not registered for GST at present, so that price is final. Bigger projects are priced after a scoping call.

Sources (checked 19 October 2026)

• Upwork: Facebook and Meta Ads specialists in India
• GoodFirms: social media marketing companies in India
• Clutch: PPC agencies in India
• The Economic Times (Reuters), 19 June 2025: agency commission guideline report
• Meta Business Help Center: India's goods and services tax
• Razorpay: GST on advertising services
• Exchange rate: ECB reference rate for 16 October 2026, via Frankfurter

This article first appeared on madhurtyagi.sites.tab.bot, written by Madhur Tyagi, The Human Algorithm.

Written on 19 October 2026. Meta changes its ad system often, so every fact below carries its source and date. Read it as "true as of October 2026".

Picture a huge library with tens of millions of books. A reader walks in. A librarian cannot read every book to them, so she does two things. She sweeps the shelves and pulls a small pile that might suit this reader. Then she picks the one to hand over.

Meta's ad system works the same way. Andromeda is the sweep. If your ad does not make it into the pile, it never gets a chance in the second step, no matter how good your bid or your targeting is.

This guide explains what Andromeda is, what Meta itself has said about it, and the 7 changes I would make for an Indian D2C brand today.

What is Meta Andromeda, in plain words?

Andromeda is the name Meta gave to its ad retrieval system: the stage that narrows every ad that could be shown down to a shortlist, before the ranking models choose the final ad.

In Meta's own words, from the Engineering at Meta blog (2 December 2024): retrieval "is tasked with selecting ads from tens of millions of ad candidates into a few thousand relevant ad candidates". After that, larger ranking models decide what each person sees.

What made Andromeda different, by Meta's account:

• Bigger brain at the sweep. On its Q4 2024 earnings call (29 January 2025), Meta said Andromeda "enabled a 10,000x increase in the complexity of models we use for ads retrieval" and "has driven an 8% increase in the quality of ads that people see on objectives we've tested". (Meta's own figures.)
• Built for far more creative. The engineering post says Andromeda added an index designed to scale to a large volume of ad creatives, because AI tools mean advertisers upload many more ads than before.
• Personal, not rule based. Before Andromeda, Meta says retrieval leaned on many separate model stages and fixed rules. Now a much larger model learns which ad suits which person.

What changed, and when (a dated timeline)

There was no single "switch day" that Meta announced. Andromeda arrived in the second half of 2024 and Meta kept improving it. Some agency blogs date a global rollout to October 2025. I could not find Meta confirming one date, so I list what Meta itself published:

Every percentage in this table is Meta's own reported figure for its whole system, not a promise for your account.

The big idea: lookalike ads now count as one

This is the change most Indian brands have not caught up with.

In December 2025, Meta wrote that its system "analyzes creatives and groups together ads that share key visual and thematic attributes. When multiple ads look or feel alike, these are seen as variations of the same creative."

Think about what that means. Ten ads with the same model, the same background and five different headlines are not ten chances. To the system they are close to one idea with small changes. Ten ads built on ten different ideas (a founder story, a customer review, a demo, a comparison, a festive offer) are ten real chances to find ten different groups of buyers.

Meta draws the line itself: creative iteration is testing small changes to one ad, like a new headline. Creative diversification is making ads that are, in Meta's words, "truly different in look, feel, storyline, and message". Iteration still has a place. It just does not replace new ideas.

7 changes Indian D2C brands should make

1. Fewer, simpler campaigns

Ten ad sets slicing the same audience by age and interest made sense when you had to tell Meta who to find. Now the system does most of that finding itself, and splitting the budget across many small ad sets starves each one of data.

What to do: one prospecting campaign per goal, broad targeting, and let the budget pool. Meta's own April 2025 post reports that advertisers saw "a 9% lower cost per action" on average with the streamlined setup for Advantage+ sales campaigns (Meta's figure, averaged across advertisers).

2. Brief ideas, not variations

Your creative brief should list different reasons to buy, not different headlines. A simple way to build them for an Indian D2C brand:

• Who is speaking: founder, customer, creator, expert.
• Which problem: the one each buyer type cares about most.
• Which format: short video, static image, carousel, a creator's partnership ad.
• Which language: English, Hindi, Hinglish or a regional language, where your buyers are.

Change at least two of these between ads, and they stop being lookalikes.

3. Make what you can sustain every week

Meta's April 2025 post quotes one brand that moved from 3 or 4 new creatives a week to almost 50. That is one brand's story, not a target. A steady flow of new ideas you can keep up beats one big batch followed by silence. Most small brands can start with a few truly different concepts a week, recorded on a phone.

4. Give Meta clean purchase signals, with COD counted honestly

The system learns from the conversions you send it. If those signals are messy, it learns the wrong buyer.

• Set up the Meta Pixel and the Conversions API, and optimise for purchases, not add to cart.
• Remember that a COD order counts as a purchase when it is placed, not when it is delivered. If many COD orders come back as RTO, the system keeps finding people who order and refuse.
• Judge your results on delivered orders. My guide on ROAS with COD and RTO has a calculator for your real break even ROAS.

5. Read results by idea, not by ad

Group your ads by concept in a simple sheet: founder story, review, demo, offer. Look at cost per delivered order per concept. Kill ideas, not single ads, and make more of the ideas that win.

6. Watch fatigue and similarity

Meta separates two problems in its December 2025 post: creative fatigue (people have seen the same ad too often) and creative similarity (your ads look too alike). Meta says it is testing creative insights in Ads Manager to flag both. If your account shows them, check them weekly. If not, a falling click rate on a concept that used to work is your cue.

7. Fix what happens after the click

Andromeda can find you better buyers. It cannot reply to them. When a buyer asks "is COD available?" on WhatsApp and waits a day, the sale is lost after Meta did its job.

Confirm COD orders before you ship, reply to product questions quickly, and follow up on abandoned carts with a person's yes. That is where I usually find the biggest leak in a D2C brand's numbers, and why I built WhatsApp follow up for D2C brands.

Check your own setup

What not to believe

• "Targeting is dead." Broad targeting does more of the work now, but your offer, price and creative still decide who responds. Exclusions and location still matter.
• "You need 50 ads a week." You need ideas that are actually different. Five strong, different concepts beat fifty copies.
• "Andromeda will fix bad ads." It decides which ads get a chance. It does not make a weak offer strong.
• "Andromeda is a setting you turn on." It is part of how delivery works for every advertiser. There is nothing to switch on.

A 3 week test plan

• Week 1: Merge overlapping ad sets into one broad prospecting campaign. Check that Pixel and Conversions API both send purchases.
• Week 2: Launch 4 to 6 concepts that differ in who speaks, which problem, format or language. Do not touch them for a few days.
• Week 3: Compare concepts on cost per delivered order. Pause the losers, make two new versions of the winner's idea (not its headline), and add one new idea.

This is a test, not a promise. Your results depend on your product, price, margins and what happens after the click.

This article first appeared on madhurtyagi.sites.tab.bot, written by Madhur Tyagi, The Human Algorithm.

If you searched for the average cost per lead on Facebook in India, here is the honest answer: there is no India wide number worth planning a budget around.

Your CPL moves with your city, your offer, your price and the form you use.

I have not found a public CPL benchmark for India, from Meta or from a large ad dataset, that I would trust with real money.

What you can trust is your own break even CPL. It takes two numbers to work out, and you already have both.

In 8+ years of running ads for 25+ brands since 2017, the costly mistake I keep seeing is not a high CPL.

It is a low CPL that nobody checks against sales.

What is a good cost per lead on Meta ads in India?

A good CPL is one that stays below your break even CPL.

Break even CPL is the most you can pay for one lead and still make money, once the leads who buy have paid for the leads who don't.

That is why the same CPL can be great for one business and a loss for another.

A ₹600 lead can be cheap for a clinic selling a ₹40,000 treatment. A ₹60 lead can be expensive for a course that almost nobody buys.

(Both are illustrations, not benchmarks.)

So stop asking "is my CPL good?" and ask "is my CPL below what one lead is worth to me?"

What do published CPL benchmarks say?

The CPL numbers you will find online come mostly from US advertisers. The most recent one I could check:

• LocaliQ's Facebook advertising benchmarks for 2026 (last updated 23 September 2026) put the average cost per lead for the leads objective across industries at $27.39, a little below the $27.66 of the year before.
• In the same report, Career and Employment leads averaged $12.30, Education and Instruction $26.31, and Beauty and Personal Care $50.91.

Read these for the spread, not as a target.

In that data a beauty lead costs about four times a career lead, because the buyer, the price and the competition are different.

Auction prices, ticket sizes and buying habits in India are different again, so a US average tells you very little about your own account.

The number that tells you something is the one you can work out yourself.

How to work out your maximum CPL

Here is the formula I use with clients. It fits on one line:

Break even CPL = profit per customer × lead to customer rate

• Profit per customer is what you keep from one sale after product, delivery and payment costs, before ad spend. If customers reliably buy again, you can include the repeat profit you are confident about. If you are not sure, leave it out.
• Lead to customer rate is how many leads out of 100 actually paid. Take it from your last 30 to 60 days of real leads, not from a hope.

Break even means the ads earn you nothing, so you also want a target below it:

Target CPL = break even CPL × the share of your profit you are happy to give to ads

Worked example (fill in your own column)

The example is an illustration for a coaching program, not a benchmark. Write your numbers in the right column.

How to read it: in this example, a CPL under ₹240 makes money with room to spare. Between ₹240 and ₹480 it still makes some money.

Above ₹480, every new customer costs more than they bring in.

One more number settles most arguments about lead quality:

Cost per customer = CPL ÷ lead to customer rate

Try it with your numbers

Why cheap leads still don't buy

Here is the trap in numbers. Two campaigns for the same ₹12,000 profit offer (illustration):

Campaign A looks five times cheaper in Ads Manager. It loses money on every sale.

Where do the buyers go? Most of the leak happens after the form is sent.

These are the reasons I see most often.

1. The form asks for nothing

A Meta lead form fills in the name and number for the person, so submitting can take two taps, sometimes by accident.

Meta offers a form type called Higher intent, which adds a review step before the form is sent (Meta Business Help: form types).

Add one or two short questions a casual person will not bother with: city, budget, or when they want to start.

You get fewer leads and more buyers.

2. The reply comes too late

A lead is warm for minutes, not days.

Research published in Harvard Business Review, The Short Life of Online Sales Leads (March 2011), found that most companies were not responding to online enquiries nearly fast enough.

If a person calls or messages within minutes during working hours, the same lead is worth more.

If an AI assistant sends that reply for you, it should say it is an AI assistant in its opening line.

3. Nobody follows up

Many leads do not answer the opening message. They are busy, not uninterested.

A simple plan works better than none: a call and a WhatsApp message on day 0, one useful message on day 1, a short check in on day 3, then stop.

Ready words for each step are in the free Script Vault.

4. Meta is learning from the wrong signal

If the campaign is set to get leads, Meta finds people who fill forms. That is exactly what it was asked to do.

When you send back which leads became customers, from your CRM through Meta's Conversions API, you can use the conversion leads goal so Meta looks for people likely to buy (Meta Business Help: performance goals for lead ads).

Even a simple sheet with a "bought: yes or no" column is the place to start.

5. The ad promises something the offer can't keep

"Free", "lowest price" and "get price list" ads pull in people hunting for a deal.

If the ad says one thing and the call says another, the lead goes cold on the call.

Name the problem you solve and who it is for, so the wrong people skip the ad.

How do I lower my cost per lead without getting junk leads?

Lower CPL helps when cost per customer falls with it. These changes usually help both:

• Make the hook specific. Name one problem and one kind of buyer in the opening line of the ad. The right people stop for an ad that describes them, and the wrong people scroll past.
• Add one qualifying question. It raises CPL a little and usually lowers cost per customer more.
• Serve the area you can actually serve. Leads from cities you cannot deliver to are pure waste.
• Test a click to WhatsApp ad against the form. A conversation tells you more about intent than a form does, and it starts in the app people already use.
• Judge ads by cost per customer, not CPL. Pause the ad that brings cheap leads who never buy, even when it looks like the winner.

How to know what a lead is really worth

Track five numbers every week. A Google Sheet is enough:

1. Spend
2. Leads
3. Qualified leads (people who answered and fit)
4. Customers
5. Revenue from those customers

From these you get CPL, cost per qualified lead and cost per customer.

Tag every lead with the ad or form it came from, so you can see which ad brings buyers and which brings form fills.

This is the same discipline behind our D2C fragrance case study: a D2C solid perfume brand from India got a 35% lower CPL and held 3.5x+ ROAS on Meta for 6+ months.

A lower CPL counts when the sales hold too.

Where the leak usually is

If your CPL looks fine and sales do not, the problem is usually after the lead: the reply, the follow up or the offer.

The free Growth Leak Check takes 2 minutes.

It asks 7 questions on reply speed, follow up, lead capture, ads, offer, content and payments, and shows your top 3 leaks with one fix each.

If you are a coach or creator, the Creator Leak Check asks about your DMs and how you earn instead.

This article first appeared on madhurtyagi.sites.tab.bot, written by Madhur Tyagi, The Human Algorithm.

A good ROAS for Meta ads is not 2x, 3x or 4x.

It is any number above your break even ROAS, and in India that number is often higher than it looks, because Ads Manager counts cash on delivery orders that never get delivered.

If you sell with COD, the ROAS in your dashboard and the money in your bank are two different numbers.

This guide gives you the formula that joins them, with a worked example you can fill in with your own numbers.

I have run Meta ads for 25+ brands since 2017, many of them D2C brands shipping across India.

The question I ask on every account is the same: what ROAS do we need just to stand still?

What is a good ROAS for Meta ads?

A good ROAS is one that leaves profit after every cost, including the orders that come back.

You will hear rules of thumb like "aim for 3x".

None of them knows your margin, your COD share or your return rate, so none of them can tell you if you are making money.

Here is the one line answer:

Good ROAS = any ROAS above your break even ROAS, measured the way Ads Manager measures it, but with your real costs behind it.

What do published ROAS benchmarks say?

Public ROAS data comes mostly from brands outside India. The most recent dataset I could check:

• Triple Whale's Meta ads benchmarks report a median ROAS of 1.88 on Meta for August 2025 to July 2026, across 17 ecommerce industries, almost flat from 1.86 the year before.

That median tells you something useful: half of the brands in that data tracked less than 1.88 in sales for every 1 spent on Meta ads.

It does not tell you what a good ROAS is for you, because it says nothing about your margin, and it is not Indian data with Indian COD habits.

Why your Meta ROAS is not your real ROAS in India

Three gaps sit between the ROAS in Ads Manager and the profit you keep.

1. COD orders that come back (RTO)

When a COD order is refused or the buyer cannot be reached, the parcel returns to you. That is RTO, return to origin.

Meta still counted that purchase. You paid shipping both ways and earned nothing.

How big is it?

The Shipway ShipNotes FY25 analysis, as cited by Amazon Shipping India, reports about 26% RTO on COD orders overall and under 2% on prepaid orders.

Your own number can be higher or lower by category, price and city. Use yours if you have it.

2. The value Meta sees is not the money you keep

The purchase value your website sends to Meta often includes GST and the shipping charge.

Your margin has to be counted on that same value, or your break even ROAS will look lower than it really is.

3. Meta takes credit for some sales that would have happened anyway

The default attribution setting in Ads Manager is 7 day click and 1 day view.

A person who saw your ad, ignored it and later bought from a Google search can still show up as a Meta sale.

That is not wrong, but it means Meta ROAS is an estimate, not a bank statement.

How to calculate break even ROAS with COD and RTO

Here is the formula I use. It needs three inputs:

• Order value (A): the average purchase value Meta reports per order.
• Margin (m): the share of A you keep on a delivered order after product cost, packing, shipping and payment or COD fees, before ads.
• Delivered share (d): 1 minus (COD share × RTO rate on COD).

And one cost: RTO loss (L), what one returned order costs you in shipping both ways and packing.

Then:

Profit per order before ads = (d × m × A) minus (COD share × RTO rate × L)

Break even ROAS = A ÷ profit per order before ads

If every order is prepaid, this becomes the familiar break even ROAS = 1 ÷ margin.

Worked example (fill in your own column)

This is an illustration for a D2C brand, not a benchmark. Write your numbers in the right column.

What this shows: the brand above thinks it breaks even at 2.2x. It really breaks even at 2.8x.

At a 2.5x ROAS in Ads Manager, it spends ₹480 in ads per order (₹1,200 ÷ 2.5) and keeps ₹429.

It loses ₹51 on every order while the dashboard looks fine.

Try it with your numbers

What ROAS should you aim for?

Break even means the ads pay for themselves and nothing more. To keep a profit, decide how much you want left per order, then:

Target ROAS = A ÷ (profit per order before ads minus the profit you want to keep)

In the example, to keep ₹150 per order: ₹1,200 ÷ (₹429 minus ₹150) = ₹1,200 ÷ ₹279 = about 4.3x. (Illustration.)

That is why one brand can grow happily at 2.5x and another loses money at 3x.

For a real reference, our D2C fragrance case study shows how a D2C solid perfume brand from India held 3.5x+ ROAS on Meta for 6+ months with a 35% lower CPL.

How to raise your real ROAS, not just the dashboard one

These moves lift the money you keep. Some of them barely move the Ads Manager number, and that is fine.

• Confirm COD orders before you ship. A short call or WhatsApp message asking the buyer to confirm the address and the order catches many orders that would come back. A person can send it, or an AI assistant that says it is an AI assistant in its opening line. Ready words are in the free Script Vault.
• Make prepaid the easy choice. A small prepaid discount or a small advance on COD cuts the orders people place without meaning to receive.
• Stop shipping where parcels keep coming back. Track RTO by pincode and city, then pause COD where it is clearly losing.
• Raise order value. Bundles and a sensible free shipping threshold raise A without raising your ad cost per order.
• Tell Meta which orders were delivered. Sending delivered orders back as a conversion through the Conversions API teaches the algorithm the difference between a buyer and a refuser.
• Watch blended ROAS too. Total revenue ÷ total ad spend across every channel, checked monthly, keeps attribution from fooling you.

How to find your own numbers

You need three numbers from your own store, and all of them are already in your tools:

• Margin: take one month of delivered orders. Subtract product cost, packing, shipping, payment and COD fees from the revenue Meta would have seen for them, then divide by that revenue.
• COD share: COD orders ÷ all orders, from your store dashboard, for the same month.
• RTO rate on COD: COD orders returned ÷ COD orders shipped, from your courier panel. Look at it by city too, because it is rarely even.

Work the formula once a month.

When your margin, COD share or RTO rate moves, your break even ROAS moves with it, and so should the ROAS you ask your ads to hit.

Where the leak usually is

When ROAS looks fine and cash does not, the leak is usually after the click: COD confirmation, follow up, or payment.

The free Growth Leak Check takes 2 minutes and shows your top 3 leaks across reply speed, follow up, lead capture, ads, offer, content and payments, each with one fix.

This article first appeared on madhurtyagi.sites.tab.bot, written by Madhur Tyagi, The Human Algorithm.

If your reels get views and your bank account does not notice, the problem is rarely the content. It is what happens after someone is interested.

Here is the flow I set up for coaches and creators in India: a keyword in the post, three questions in the DM, a move to WhatsApp with permission, one clear next step, and payment by UPI.

Below you get the whole flow, the exact messages, and the mistakes that quietly kill sales.

I have worked on growth for 25+ brands since 2017, and the pattern is the same for a fitness coach, a career coach or a course creator: the people who would buy are already in your DMs.

They just never get a clear next step.

Why do Instagram views not turn into sales?

Usually one of five reasons:

1. The post never asks for anything. People enjoy it and scroll on. There is no keyword, no question, no next step.
2. "Link in bio" is the next step. A cold link to a sales page asks a stranger to decide alone. A short chat answers their real doubt.
3. DMs get answered late. By the time you reply, the person has forgotten why they messaged.
4. The opening reply is a pitch or a price list. It feels like a sales desk, so people go quiet.
5. Nobody follows up. "Seen" is not "no". A nudge often gets the reply, and many creators never send one.

The fix is not more content.

It is a simple path from interest to payment, written down, so you or your team run it the same way every time.

The DM to client flow (with scripts)

Six steps. Change the words in [brackets] and keep the shape.

Step 1: The post asks for one keyword

End your post or reel with one clear ask, tied to one problem:

Struggling to [get 5 paid clients / lose weight after 40 / crack your PM interview]? Comment PLAN and I will send you the 3 steps I use with my clients.

One post, one keyword, one problem. A keyword tells you what the person wants before you say a word.

Step 2: The opening DM gives value and asks one question

Send what you promised, then ask one easy question. Do not pitch yet.

Hey [name], here are the 3 steps 👇 [short steps or a link to a free guide] Quick question so I can point you to the right thing: where are you right now with [the problem]?

Step 3: Three short questions to see if you can help

If they reply, ask three questions, one at a time, in your own words:

1. What have you already tried for [the problem]?
2. What would change for you if this was sorted in the next few months?
3. Are you looking to do this on your own, or with someone guiding you?

The answers tell you whether this is a buyer, a learner or a fan. All three are welcome. One of them needs a call.

Step 4: Move to WhatsApp, with permission

For people who fit, ask before you move the chat. WhatsApp is where people in India actually reply, and it keeps your DMs clean.

This is easier to talk through properly. Can I message you on WhatsApp? If yes, share your number, or tap here to start the chat: [your wa.me link]

Never add someone to a broadcast list or group they did not ask for.

A person who said yes to a chat said yes to a chat.

Step 5: One clear next step

Match the step to the price:

• Lower priced offers (a workshop, a short program, a template): share the price and what it includes, then the payment details.
• Higher priced coaching: offer a 1:1 call on Zoom before asking for payment.

Based on what you told me, my [program name] fits you: [2 lines on what they get]. It is ₹[price] for [duration]. Want me to send the payment details, or would you like a 15 minute 1:1 call on Zoom to ask questions before you decide?

Hinglish works just as well, and often better. Use the language your client uses:

Aapne jo bataya, uske hisaab se [program name] aapke liye sahi hai. ₹[price] mein [what they get]. Payment details bhej doon, ya pehle 15 minute ka 1:1 call on Zoom karein?

Step 6: Take payment by UPI, then confirm by hand

Send your UPI ID or a payment link with the amount and what it covers in the same message:

Here are the payment details: UPI ID [your UPI ID] or this link [payment link]. Amount: ₹[price] for [program]. Send me a message once it is done and I will confirm and share your next steps.

Check that the money has landed before you confirm. Then send a clear welcome: what happens next, when, and how to reach you.

A buyer who feels looked after in the opening hour after paying is a buyer who stays and refers.

Here is how the whole flow reads in one chat:

How to follow up without being pushy

Silence is normal. People are busy, not uninterested. Two follow ups, then stop:

Day 1: Hey [name], just checking you saw this. Happy to answer anything about [program] before you decide.

Day 3: Hi [name], last message from me on this. If now is not the right time, no problem at all. Whenever you want to pick it up, just reply here.

That second message often gets the reply, because it removes the pressure.

More ready messages for DMs and WhatsApp are in the free Script Vault.

How fast should you reply to DMs?

As fast as you reasonably can during your working hours. Interest fades quickly.

Research published in Harvard Business Review, The Short Life of Online Sales Leads (March 2011), found that most companies were not responding to online enquiries nearly fast enough, and DMs follow the same logic.

Set two fixed windows a day to clear DMs, and tell people when to expect you.

If you use an AI assistant or a chatbot to reply, it should say it is an AI assistant in its opening message, and a person should take over before any price talk or payment.

What to post so the right people message you

Content that brings buyers is different from content that brings likes:

• Problem posts: name one problem your buyer has, in their words, and give one useful step.
• Proof posts: a real client story, shared with their permission, with what changed and what it took. Never invent or stretch results.
• Process posts: show how you work, so a buyer can picture working with you.

Each of them ends with one keyword ask from Step 1. Followers are nice. Conversations are what pay.

Score your own DM flow

Tick what you already do. The score tells you which step to fix next.

Track five numbers each week

A simple sheet is enough:

The step where the number drops the most is your leak. Fix that one before you post more.

This article first appeared on madhurtyagi.sites.tab.bot, written by Madhur Tyagi, The Human Algorithm.