The short answer
- Did ROAS drop abruptly, on one specific day? Then something most likely changed in your measurement, not in your ads.
- Did ROAS slide gradually over a few weeks? That points more towards your creative or your market: tired ads, more competition, an offer that lands less well.
- Always compare with your real revenue. If it stays flat while ROAS in Ads Manager falls, you're measuring less, not selling less.
Step 1: find the date
Put your ROAS, reported purchases and real revenue (from your store or accounting) side by side, per day. Two patterns already tell you a lot:
- A cliff: fewer conversions from one day to the next, while clicks and real revenue stay roughly the same. That smells like a measurement problem. Look for what changed that day: a new cookie banner, a theme or checkout update, a new tracking app, a change in your reporting.
- A slope: a gradual decline over weeks, with rising frequency and a falling click-through rate. That points more towards your creative or market. More on that in step 3.
Step 2: did your measurement change?
These are the measurement causes we see most often, and the ones to rule out first.
Meta reports fewer attribution windows
Meta announced that from 12 January 2026, its reporting API no longer returns the "7-day view" and "28-day view" attribution windows (Meta for Developers, 16 Oct 2025). If your dashboard or reporting tool relied on those windows, you've seen fewer attributed conversions since then, and therefore a lower ROAS, without selling any less. Check which windows your reporting uses, and only compare periods with the same setting.
Pixel and Conversions API don't add up
Many stores send a purchase both through the browser pixel and through the Conversions API from the server. Meta merges them into one purchase when the event ID and event name match, within 48 hours (Meta documentation). When that breaks, you get one of two problems: double counting (your ROAS looks too good, until someone fixes it and it "crashes"), or purchases that aren't attributed properly. An app or theme update can break that link without anyone noticing.
Your cookie banner doesn't pass on consent properly
In the EU and UK, tracking cookies need the visitor's consent. After a new or updated banner, the technical side often goes wrong: the visitor clicks "accept", but the signal never reaches your tags. Conversions then disappear from your reporting. The fix is never to work around consent, but to make sure that a "yes" is actually passed on as a "yes". Google, for example, uses four consent signals in its Consent Mode (Google documentation).
What else can change
- A new checkout, payment provider or thank-you page where the purchase event no longer fires.
- A tracking or consent app that was updated or replaced.
- A different optimisation event or campaign setting, so you're comparing apples with oranges.
A typical scenario (for illustration)
This is a composite example of how such situations usually play out, not a specific client case.
On Monday, an online store sees ROAS in Ads Manager fall from just over 3 to below 2. Clicks are unchanged, and so is revenue in the store. The team's first reflex: pause the "bad" ads and rush out new creatives.
Looking at the date first tells a different story. The checkout was updated on Sunday evening, and since then the purchase event on the new thank-you page no longer fires in the browser. Meta stops seeing part of the purchases, and the ads look worse than they are. Once the event is fixed, reported ROAS returns to its old level, without a single ad being changed.
The lesson: had they paused the ads, they would have switched off their winners for a problem that was never in the ads.
Step 3: only then, your creative and your market
If your measurement is right and ROAS still falls, look at what the customer sees.
- Tired ads: frequency climbs, click-through rate drops, the same people see the same message too often.
- Too little real variety: Meta's systems treat near-identical variants (different background, different headline) largely as the same ad. New angles, not just new variants, give the system something new to test.
- More competition: a competitor who just found a winner can push up your costs.
- Your offer or landing page: clicks without purchases often point to a mismatch between what the ad promises and what the page delivers.
Overview: what you see and where to start
- Conversions crash on one day, real revenue stable → measurement. Find what changed that day.
- Reported conversions higher than your real orders → double counting. Check pixel and Conversions API deduplication.
- Drop since a new cookie banner → consent isn't passed on correctly.
- Gradual decline, frequency up, click-through rate down → creative. You need new angles.
- Lots of clicks, few purchases → offer or landing page.
Why the order matters
Responding to a measurement error with new creatives, a lower budget or a different audience changes exactly the things that were working. That's why at SASDAMX we always check measurement first and creative second. They're two sides of the same problem, and they belong together.