The Meta ads analysis system that worked in 2025 is now actively destroying ROAS (Return on Ad Spend) for ecommerce brands in 2026. After Meta released the Andromeda update and GEM (the generative model powering ad distribution), the way you read your Ads Manager, decide which ads to turn off, and reallocate budget completely changed. Brands still optimizing for the single “winning ad” are killing the creatives that feed their entire funnel.
In this guide, we break down the exact 5-step Meta ads analysis system we use at Skaleit to manage millions in monthly ad spend across ecommerce clients, post-Andromeda. This is the same framework keeping our accounts profitable while competitors watch ROAS collapse.
TL;DR
– Fix tracking first: Event Match Quality (EMQ) must be 9+ for the purchase event before any analysis is valid. – Restructure your columns: Build a funnel-logic column layout in Ads Manager covering impressions, frequency, CTR, hold rate, ATC, and incremental attribution. – Stop optimizing single ads: GEM optimizes the delivery sequence, not the individual creative, so kill an upper-funnel ad and the whole sequence collapses. – Judge by incremental attribution: This is the only metric that tells you which purchases would not have happened otherwise. – Reallocate budget every 2-3 days: Move 20-30% from low-incremental ads into the highest-ROAS campaigns.
Step 1: Fix Your Tracking Foundation (EMQ 9+)
Event Match Quality (EMQ) above 9 is non-negotiable for a reliable Meta ads analysis system in 2026. If your tracking is broken, every optimization decision after it is wrong.
Here is how to check it:
- Open Meta Events Manager.
- Go to Data Sets and select the pixel tied to your brand.
- Scroll to the Purchase event.
- Confirm Event Match Quality is at least 9/10.
Why this matters: the more identifying information Meta receives through your pixel and Conversions API (CAPI), the more accurate the targeting signal sent to the Andromeda algorithm. For ecommerce, the Purchase event is the primary optimization event, so EMQ here must be as high as possible. Other events (ViewContent, AddToCart, InitiateCheckout) should also score high to give GEM a full behavioral picture of who actually converts.
If you skip this step, the rest of this analysis system is built on noise.
Step 2: Structure Ads Manager Columns Like a Funnel
Your column layout in Ads Manager dictates the speed and accuracy of every decision you make. The default columns hide the metrics you actually need.
Here is the exact column order we use across client accounts:
- Campaign / Delivery
- Purchases
- ROAS
- Cost Per Purchase
- Budget
- Amount Spent
- Impressions
- Reach
- Frequency
- CPM
- Link Clicks
- CTR (Click-Through Rate)
- Hold Rate
- Scroll Stopper (3-second video plays / impressions)
- CPC (Cost Per Click)
- Landing Page Views
- Add to Cart
- Cost per ATC
- Initiate Checkout
- Cost per IC
- Partial Conversion Value
- Average Order Value
- ROAS (incremental view)
The logic: top metrics give you headline performance, then the funnel metrics (impressions → reach → frequency → CTR → hold rate → ATC → IC → purchase) tell you exactly where in the journey each ad is breaking. This funnel-logic column structure is what allows the rest of the analysis to happen in minutes instead of hours.
Step 3: Read GEM Sequence Logic, Not Single-Ad Logic
GEM distributes content across a sequence of ads, not as isolated creatives, and this is the single biggest mindset shift required post-Andromeda. If you analyze ads the old way, you will turn off the creative that is feeding every other ad in your account.
Here is a real example from a client account:
– 1 campaign, 1 ad set, 30 ads inside. – Total monthly spend: $5,800. – Top-spending ad: $1,200 (lower ROAS than the #2 ad at $900). – The next ads spend $800, $700, etc.
The instinct from 2025 thinking: turn off the top-spending ad because its ROAS is lower. The reality post-Andromeda: that ad is the top-funnel reach driver. It is the creative bringing new audiences into the sequence, who are then retargeted by the other ads in the set. Kill it, and every other ad’s ROAS collapses within days.
This is why we build creatives as a journey, not as independent tests. For the deeper logic behind this, our breakdown of sequencing creative strategy on Meta ads shows how to build ads designed to be read by GEM as a connected funnel. And if you want the full algorithmic context for why this works, our complete guide to Meta’s Andromeda algorithm for ecommerce covers how GEM evaluates creative distribution end to end.
Step 4: Judge Ads by Incremental Attribution (Not Last-Click ROAS)
Incremental attribution is the only metric that tells you which purchases would not have happened without that specific ad. Standard 7-day click or 7-day click + 1-day view attribution counts conversions that often would have occurred anyway. Incremental strips that out.
In the same client account, the top-spending ad had the lowest standard ROAS but the highest incremental attribution. The #2 ad had a higher reported ROAS but lower incremental attribution. Translation: the #1 ad was generating actual new revenue, the #2 ad was largely capturing demand already created by ad #1.
Our decision rules for turning ads off:
– Keep: high incremental attribution, even if standard ROAS looks weak. – Kill: low incremental attribution AND poor performance over the last 7 days AND zero purchases in the last 2 days. – Watch: ads with mid-tier incremental attribution and declining 4-day ROAS (~1.27x or lower).
Workflow tip: filter your view by “Ad Delivery is Active” to remove the noise of inactive ads. Then run a 7-day window to identify candidates for cuts, and confirm with a 2-day window to catch recent collapses.
This incremental-first approach is consistent with Meta’s official guidance on using incrementality testing to evaluate true ad-driven lift, especially as iOS and signal-loss challenges continue into 2026.
Step 5: Reallocate Budget Every 2-3 Days by 20-30%
Budget reallocation under normal conditions should move 20-30% every 2 to 3 days, from low-incremental ads or campaigns into your highest-ROAS performers. This pace gives GEM enough learning time to recalibrate without destabilizing delivery.
The rules:
– Decrease budget on campaigns with low incremental attribution (e.g., 0.20x incremental ROAS). – Increase budget on campaigns with strong ROAS AND strong incremental attribution. – Black Friday / peak periods: you can 2x or 3x budgets in a single move. We have gone from $100/day to $500/day in one push. Day 1 ROAS may drop (e.g., 4x to 2x), but Meta’s learning recovers within 2-4 days, faster than Google.
Real example from today’s account adjustments:
– Pulled $20/day from a campaign with 0.20 incremental ROAS. – Added $20/day to a campaign with strong incremental + strong ROAS. – The highest-budget campaign in the account ($500/day) is the one with one of the highest ROAS values, which is the correct distribution.
The mistake we see constantly: brands spending more on testing than on scaling their best performer. If your best campaign is doing 2x ROAS at $500/day and your second is doing 2.90x at $120/day, increasing the second too fast usually drops its ROAS. The 20-30% / 2-3 day cadence is how you find that ceiling without crashing through it. For brands wrestling with this exact ceiling problem, our deeper read on scaling Meta ads post-Andromeda covers the four structural fixes we apply most often.
FAQ
What is the most important metric in Meta ads analysis post-Andromeda?
Incremental attribution. Standard 7-day click ROAS over-credits ads that capture existing demand. Incremental attribution shows which conversions would not have happened without that specific ad, which is the only honest measure of value in 2026.
Should I turn off ads with low ROAS but high spend?
Not automatically. Post-Andromeda, GEM uses high-spend ads as top-of-funnel sequence drivers that feed warmer ads later in the journey. Check incremental attribution first. If incremental is high, keep the ad. If both standard ROAS and incremental are low for 7+ days with zero recent purchases, then turn it off.
What Event Match Quality should I aim for in 2026?
A minimum of 9 out of 10 on your Purchase event for ecommerce brands. Lower scores mean Meta receives weaker identity signals, which degrades targeting precision and inflates CPMs.
How often should I reallocate budget on Meta ads?
Every 2 to 3 days under normal conditions, shifting 20-30% from low-incremental campaigns to high-ROAS performers. During peak periods like Black Friday, 2x to 3x daily-budget jumps are acceptable because Meta’s learning phase recovers within 2-4 days.
Why does the Ads Manager column layout matter so much?
A funnel-logic column structure (purchases → ROAS → impressions → frequency → CTR → hold rate → ATC → IC) lets you diagnose exactly where an ad breaks in the customer journey. Default columns hide the diagnostic metrics, which forces guesswork instead of data-driven decisions.
About the Author
Antonio Ventre is the founder of Skaleit, a Meta ads agency for ecommerce brands managing millions in monthly ad spend. Skaleit specializes in post-Andromeda creative strategy, incremental attribution analysis, and profitable scaling frameworks for DTC brands.
Want Skaleit to Build This System for Your Brand?
If your ROAS has slipped since the Andromeda update and you want our team to install this analysis and reallocation system inside your ad account, book a free strategy call with Skaleit.