Andromeda Killed Your ROAS? +93% Fix (Case Study)

The Andromeda Facebook Ads ROAS fix is a three-part recovery system (unified CBO structure, incremental attribution winner selection, and net new creative concepts) that we use to rebuild ecommerce ad accounts wrecked by Meta’s Andromeda algorithm update. In this case study, we walk through how we took an ecommerce brand from a 2.01x ROAS (Return on Ad Spend) in July to a 3.89x ROAS in September, a +93% lift, by scrapping the old testing-and-scaling playbook and rebuilding around post-Andromeda mechanics.

If you are still running Meta ads the way you did in early 2024, this is why your ROAS collapsed, and this is exactly what we changed.

TL;DR

– Brand baseline (July): $23K spend, 2.01x ROAS on old ABO testing structure. – After fix (August): $17K spend, 4.29x ROAS. – Sustained (September): $22K spend, 3.89x ROAS. – Three fixes applied: unified CBO full-funnel structure, winner selection via incremental attribution (not raw purchases), and creative diversification with net new concepts. – Key mindset shift as of 2026: headline swaps and background-color tweaks are no longer valid iterations. Andromeda re-targets the same audience if the visual concept does not change.

What Broke After Meta Andromeda (And Why Old Strategies Stopped Working)

Meta Andromeda is the algorithm architecture Meta rolled out that consolidated audience targeting and shifted the way the delivery system reads creative signals. In practical terms, three legacy tactics that used to work now actively burn budget:

1. ABO testing campaigns (Ad Set Budget Optimization) running in parallel with scaling campaigns. Andromeda favors consolidated learning, so fragmented structures starve the algorithm.
2. Picking winners by amount spent and total purchases. These metrics are polluted by 7-day-click / 1-day-view attribution and existing-customer traffic.
3. Iterating winning ads by changing only the headline. Andromeda reads the visual as the primary signal, so headline-only variations get delivered to the same audience segment and burn out fast.

When we audited this brand’s account, all three problems were present. According to Meta’s own documentation on Advantage+ and consolidated delivery, the system is designed to reward broader signals and diverse creative inputs, which confirms why the old fragmented approach fails.

Fix #1: Unified CBO Full-Funnel Structure

The unified CBO (Campaign Budget Optimization) full-funnel structure replaces the old prospecting-plus-retargeting split with one consolidated campaign that lets Meta distribute budget across the entire funnel automatically.

Here is what changed in the account:

– Before: separate testing campaigns, separate scaling campaigns, separate retargeting. Six-plus campaigns fighting for the same auction. – After: one unified CBO with cold prospecting, warm retargeting, and existing customers all inside. Meta decides distribution.

This structure aligns with how Andromeda actually processes signals. Instead of forcing Meta to relearn each campaign, we give it one budget and one signal pool. Our consolidated CBO strategy breakdown covers the full setup if you want the campaign-level walkthrough.

Fix #2: Pick Winners Using Incremental Attribution, Not Total Spend

Incremental attribution winner selection is the practice of judging ad performance by ROAS on NEW audiences only, filtering out purchases coming from existing customers, retargeting, and 1-day-view attribution.

This is where most brands get it catastrophically wrong. Here is the example from the account:

| Metric | Creative A | Creative B | |—|—|—| | Total ROAS (blended) | 4.20x | 2.92x | | Incremental ROAS (new audience) | 1.93x | 2.42x | | Decision | Looks like winner, actually stale | True winner, worth scaling |

Creative A looks like a clear winner at a blended 4.20x, but 100% of the lift is coming from engaged audiences and existing customers. Creative B, at a lower blended 2.92x, is doing the real work of acquiring new buyers at 2.42x incremental ROAS.

The rule we now apply on every account: pick scaling winners based on incremental ROAS on new audiences, not total purchases or amount spent. This one change alone accounted for a large portion of the ROAS lift, and it is the same principle we detail in our guide on incremental attribution for winning ads.

We also introduced a hard turn-off rule: if an ad spends aggressively and its ROAS falls 30% below the campaign average, we kill it. Keeping bad ads alive hoping the funnel will “balance out” is a myth. Meta will keep feeding the highest-volume ad even if it is unprofitable.

Fix #3: Creative Diversification With Net New Concepts

Creative diversification post-Andromeda means changing the visual concept, the avatar, and the messaging together, not swapping headlines on the same image.

What Does Not Work Anymore

– Same visual, new headline. – Same layout, new background color. – Same product shot, new button color.

Under Andromeda, these variations get served to the same audience segment. That means creative fatigue accelerates and ads die 2 to 3 times faster than they did pre-Andromeda.

What We Implemented Instead

Our creative system now enforces three rules for every iteration on a winning ad:

1. Match the angle to the visual. If the angle is “lose weight with this meal plan,” the visual is a light, low-calorie salad. If the angle is “gym bros are stacking mass with this,” the visual is a chicken-and-rice high-protein dish with 50g protein callouts.
2. Match the avatar to the persona. Targeting men 40+? The person in the ad is a 40+ man. This sounds obvious. Most brands do not do it.
3. Match the copy tone to the target. Gym-audience copy is direct and aggressive. Wellness-audience copy is calmer and benefit-led.

70% of the winning creatives in this account were static ads. Only 30% were video. That distribution matches what we cover in our static ads Meta strategy breakdown.

Net New Concepts, Not Recycled Templates

The brand’s biggest creative sin was copying the “classic” ad formats you see 200 times a day: product shot with arrows, us vs them tables, generic point-out visuals. These get scrolled past because users have seen them for years.

We replaced every recycled format with net new concepts, meaning original visual executions that deliver the same angle in a way the audience has not seen. This is aligned with the broader shift we cover in our complete guide to Meta Andromeda for ecommerce, which is the hub post for everything post-Andromeda on this blog.

The Actual Results: 2.01x to 3.89x ROAS

Here is the month-over-month breakdown from the ad account:

| Month | Spend | ROAS | |—|—|—| | July (pre-fix) | $23,000 | 2.01x | | August (fixes applied) | $17,000 | 4.29x | | September (stabilized) | $22,000 | 3.89x |

Spend held roughly steady. ROAS nearly doubled. That is what a properly rebuilt post-Andromeda system looks like, and it holds up as the account scales further.

Why This Works: The Post-Andromeda Logic

Andromeda rewards three things: consolidated signals, incremental new-customer acquisition, and creative diversity. Everything we changed serves one of those three levers.

– Unified CBO = consolidated signals. – Incremental attribution winner picks = incremental new-customer focus. – Net new creative concepts + matched avatar/angle = creative diversity.

If any one of these three is missing, ROAS will decay. All three together is what unlocks the +93% lift. Independent analysis from industry benchmark reports confirms that consolidated structures with diverse creative outperform fragmented setups by wide margins across ecommerce verticals.

FAQ

About the Author

Antonio Ventre is the founder of Skaleit, a Meta ads agency specialized in scaling ecommerce brands post-Andromeda. Skaleit has managed over $10M in ad spend across fashion, supplements, skincare, and DTC verticals.

Want Skaleit to Build This System for Your Brand?

If your ROAS dropped after Andromeda and you want us to rebuild your account using the exact three-part fix in this case study, book a call with our team here.