Meta Andromeda update: what Indian D2C brands should change (as of October 2026)

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.

Key takeawayTen copies of one ad are about one idea to Meta. Ten different ideas are ten chances.

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.

Key takeawayChange who speaks, the problem, the format or the language. Not just the headline.

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.

Meta now does more of the ad work. The chats after the click still need a plan: see which D2C jobs an AI agent can take, and which stay human.

Andromeda changes how ads find buyers. It does not change who replies to them. If you are deciding who should run that for you, read AI growth partner vs agency vs freelancer.