Is Your Agency Andromeda-Ready? A 2026 Meta Self-Audit for CPG Brands

Meta Andromeda is the AI retrieval engine that decides which of your ads actually get shown, and it quietly rewrote the rules of paid social in 2026. The short version: if your agency is still cloning one winning ad into ten near-identical variants, Meta now treats all ten as a single ad, so you pay to produce a full slate and get the reach and fatigue curve of one creative. This is a self-audit to tell whether the team running your Meta account is on a 2026 playbook or a 2019 one. By the end you can grade an Andromeda-ready agency on six signals in about ten minutes, no ad-account login required.

We run this same check on our own CPG accounts every month, so this is the actual rubric, not a theory.

What Meta Andromeda actually changed

Meta shipped Andromeda in late 2024 as a new retrieval engine, the layer that sifts through every ad in your account and picks which ones enter each auction. Meta's own engineering team reported it evaluates tens of millions of ad candidates per query and delivered a 6% recall improvement and an 8% lift in ads quality (Engineering at Meta, 2024).

Here is the part that hits your budget. Andromeda reads the transcript, imagery, message, and sentiment of every ad, then clusters ads that say the same thing into one auction entry. Run 100 look-alike variants and Meta treats them as one. You get the delivery ceiling and fatigue of a single ad while paying to make dozens.

Underneath it is a bigger shift. With far less demographic data to target on, Meta stopped betting on who you are and started betting on what you react to. Creative is the targeting now. We broke down the mechanics in our Meta Andromeda creative strategy guide. This piece answers a different question: has the team running your account actually adjusted?

Your account can look fine and still be behind

Most brands do not leave their agency because performance fell off a cliff. They stay because the numbers look okay. ROAS is holding, spend is steady, the monthly report is green. That is the trap.

Andromeda punishes a stale playbook slowly. Meta's own research shows conversion likelihood drops about 45% by the fourth time someone sees the same creative (Analytics at Meta). Costs keep climbing at the same time: median Meta CPM rose about 20% year over year in 2025 across roughly 35,000 brands (Triple Whale). A setup that held flat in 2023 does not hold flat now. It leaks.

So the question is not whether your ROAS is okay this month. It is whether your agency is doing the things that keep it okay six months from now. Those are different questions, and the second one is what this audit answers.

The Andromeda-readiness self-audit

Score your agency on these six signals. You can answer most of them by asking your team directly and watching whether the answer is specific or vague.

Count the rows that land in the right column. Six of six, your team is current. Three or below, you are paying 2026 prices for a 2019 engine.

One nuance worth stating, because the fix is not just make more ads. Volume without variation backfires. Meta's systems reward distinct concepts, not more copies, and stuffing an ad set with 20 near-identical uploads often hurts. The goal is diversity Meta can tell apart, not raw count.

What current actually looks like

The most common thing we find in an inherited account is the old variant strategy still running. Find a winner, clone it into ten close variants, ship them for more auction entries. That was the right move in 2018. Today Andromeda collapses those ten into one, and you get more fatigue, worse delivery, and higher costs out of what looks like a full creative slate.

We audited a specialty-food CPG brand earlier this year with a quieter version of the same problem. Server-side tracking was switched off entirely, so Andromeda was making delivery decisions on a degraded view of who actually bought. Every ad used the same Shop Now button, and roughly 8 to 12 real creative concepts were hiding behind 40 to 80 near-identical ads. They were paying for dozens of ads and getting the auction power of a handful.

The fixes were not exotic. Turn on the Conversions API, which typically recovers 20 to 40% of the attributed conversions browser-only tracking loses. Split the account into genuinely distinct concepts so Andromeda registers them separately. Match the landing page to the ad concept, which has lifted cold-traffic conversion rates 15 to 30% in our accounts. Add founder-led creative, which has run 2 to 3x ROAS against baseline across our CPG accounts.

None of that is a secret tactic. It is just the current playbook, run properly.

Why this hits CPG brands hardest

CPG brands feel Andromeda more than most, for a few reasons.

• Broad appeal: a snack, a supplement, or a pet treat is not a niche B2B tool. Broad Advantage+ targeting suits it, so clinging to narrow lookalikes leaves performance on the table.

• Creative is the lever: founder-led and story-driven creative is where CPG wins, and it has run 2 to 3x ROAS for our CPG accounts. An agency treating creative as an afterthought is optimizing the wrong thing.

• Volume math is real: with creative win rates near 30% and winning ads decaying in as little as six days, some accounts need 70 to 120 net-new creatives a month. A team shipping four is not in the game.

• Meta is most of the budget: brands now put about 68% of their ad spend on Meta, so an out-of-date Meta setup is not one channel underperforming, it is most of your paid budget underperforming.

This is also where Meta is pushing everyone. Advantage+ Shopping passed a $20B annualized run-rate, up about 70% year over year, with more than 4 million advertisers now using at least one of Meta's generative-AI ad tools (Meta Q4 2024 earnings, via Marketing Dive). Meta's streamlined Advantage+ setups have shown 9% lower cost per action, and its gen-AI creative tools 11% higher click-through and 7.6% higher conversion rates (Meta for Business). The automation is not a trend to wait out.

Where this leaves you

None of this means your current agency is bad. Andromeda moved fast, and plenty of good teams have not fully caught up. The point of the audit is to see the gap clearly, because a stale playbook does not announce itself with a crash. It shows up as a slow tax on every dollar you spend on Meta, and it compounds month over month.

We keep running this readiness check on our own CPG accounts, because the algorithm keeps moving and we would rather catch the drift early than explain it later.