A media buyer opens the account on Monday and ROAS is down by more than 10% week over week. Nobody shipped new creative, nobody touched budgets, the offer is the same offer it was in January. By Tuesday afternoon there is a panic brief sitting in the shared doc and three ads that were performing perfectly well on Friday have been paused. Two weeks later, Shopify's numbers for that same window come in flat. The sales never went anywhere. The reporting did.

This is happening constantly in 2026, and it is happening for a reason that has nothing to do with the work. Meta changed how it counts conversions in March, and separately made incremental attribution available as an alternative model that reports fewer conversions on purpose. Read either of those as a performance drop and you will kill creative that was working, brief replacements you did not need, and burn a production cycle chasing a problem that only ever existed in a column heading.

What follows is the plain-English version of both changes, and then the part that actually matters day to day: how to tell a measurement change apart from a real creative problem, and how to judge creative in a way that does not fall apart every time Meta moves the goalposts.

‍

What Actually Changed in March 2026

Meta shipped its measurement update on March 2026. Three things moved. None of them changed how many people bought from you. Only which bucket the purchase landed in.

‍

A "click" now means a link click

Before March, a like, a comment, a share or a save all counted as clicks. Someone tapped the heart on Tuesday, bought on Friday, and it showed up as a 7-day click-through conversion.

That's over. Click-through now requires an actual link click, a tap that sends someone to your site, form, app or Messenger.

So click-through conversions fell in March across accounts where nothing else changed. And this is the sentence the whole article hangs on: those conversions didn't disappear, they were reclassified.

‍

Engage-through arrived with a one-day ceiling

Everything that left the click bucket landed in a new category called engage-through, likes, comments, shares, saves, carousel swipes, plus qualifying video views.

The catch is the window. Engage-through is fixed at one day and can't be extended.

Old rules: share on Monday, purchase on Friday, full 7-day click-through credit. New rules: that share buys a 24-hour window, and the Friday purchase is attributed to nothing at all.

So some conversions didn't move buckets in March. They fell out of reporting entirely. The share still influenced the sale, Meta just no longer has anywhere to record that it did.

Video engagement now qualifies at five seconds

The old threshold was ten seconds, calibrated for long Feed video. It made no sense for Reels, where Meta's data shows much of the purchase impact traces to the first couple of seconds. Five is the new bar.

This one runs the opposite direction to the other two, so know about it mainly to avoid crediting yourself for a lift you didn't cause.

Your 2026 default, if you haven't touched ad set settings since last year: 7-day click, 1-day engage-through, 1-day view-through, standard model.

The Diagnostic: Reporting Change or Real Creative Problem?

Four checks. Ten minutes. Run them before you brief anything.

‍

1. Did it drop overnight, or slowly?

Pull daily conversion for the last 30 days, and break it out by day. Not by week, a weekly view hides the exact thing you're looking for.

Now read the purchase column down the page.

One normal day, then a 30% lower day, then it just stays there? That's a counting change. Nothing about your creatives changed on a single night.

Losing a bit more every day while frequency climbs? That's fatigue. Your audience has seen the ad too many times.

If you can name the exact date it broke, it wasn't the creative.

‍

2. Did the change affect everything or only to a few things?

Strongest single signal. Rank ads by spend, compare conversion counts either side of the break.

Everything down 25–35%, including ads you launched last week with no fatigue history → reclassification. It applies account-wide and doesn't care which creative it lands on.

Two ads down 60%, three flat, one up → real. Fatigue is specific by definition. It attaches to a creative and the audience that's already seen it, so it shows up unevenly.

‍

3. Do the Engagement metrics agree?

The check most teams skip, and the cleanest one available.

Attribution governs what happens after a click. It has no influence over what happens before one. Which makes hook rate, hold rate, thumbstop and CTR an effective control group.

ROAS down, hook rate steady? The audience is responding exactly as it was. The change is downstream in the counting.

Hook rate sliding, hold rate collapsing? People have stopped watching. That's creative fatigue, and no attribution setting will fix it.

‍

4. Does your first-party data agree?

The tiebreaker. Meta's numbers moved. Did Shopify move with them?

Confirmed revenue flat while Ads Manager is down 30% means stop optimising against a drop that didn't happen. Both moved together means something real occurred and the brief is warranted.

If you run Northbeam or Triple Whale, this becomes a glance rather than an assembly job, because confirmed revenue sits beside Meta's attributed revenue on the same creative. Two sources disagreeing sharply on one ad is worth a look. Disagreeing uniformly across every ad is diagnosis #2, and you already know what you're looking at.

How to Judge Creative on True Results in Creative Analytics

The deeper problem: most teams are judging creative against a standard that keeps changing.

Meta redefines a click. Someone toggles view-through off in one ad set and not another. Suddenly last month's winner list isn't comparable to this month's. So the leaderboard everyone briefs from is measuring the measurement, not the work.

Two habits fix it.

‍

Fix one attribution setting and keep it

Pick one attribution setting as your creative evaluation standard and rank every creative on it, permanently.

Which one matters far less than that you stop changing it. 7-day click with view-through excluded is a reasonable default for most DTC accounts.

Consistency is what makes movement on the leaderboard mean anything. When a creative climbs or falls, you know the creative changed and the measurement didn't. Model switching still has a place, that place is budget and channel reporting, not creative reporting.

‍

Judge at the component level, not the ad level

Meta reports at ad, ad set and campaign. Creative decisions happen at hook, format, angle, talent and offer. That mismatch is exactly why knowing "which ad won" so rarely converts into a usable brief.

Creative Analytics 2.0 in adnova tags every creative automatically across 14 creative dimensions and maps each attribute back to performance. The quiet benefit in a year like this one: a reporting shift that hits every ad uniformly leaves your relative component rankings intact.

Question-led hooks still beat statement hooks. UGC still beats studio. Every absolute number underneath moved — the ordering didn't. And the ordering is what your next brief is built from.

That's the protection worth having. An account judged on absolute ROAS gets thrown into disarray every time Meta changes a definition, and 2026 has produced three of those in a single quarter. An account judged on relative component performance — one fixed standard, cross-checked against first-party revenue — mostly shrugs and keeps shipping.

Pair it with a naming layer a human can read (our ad creative naming conventions guide covers the structure) and the read stays clean while the platform moves underneath it.

‍

Conclusion

Meta narrowed the definition of a click, added a bucket with a one-day ceiling, and shipped a model that deliberately reports fewer sales than the one before it. Three changes in twelve months, every one capable of producing a chart that looks exactly like a creative collapse. None of them caused by your creative.

Run the four checks before you brief. Edge or slope. Everything or a few things. Did the front-end metrics move. Does first-party agree.

Then judge creative the way it should have been judged all along: one fixed attribution standard, confirmed revenue beside attributed revenue, rankings read at the component level.

Adnova tags every creative automatically and connects your first-party attribution source, so component-level rankings and confirmed revenue sit in the same view. Which is what keeps the read on your creative stable while the definitions keep shifting underneath it.

‍

FAQs

1. My conversions dropped in March 2026 and I changed nothing. What happened?

Meta narrowed click-through to require an actual link click. Likes, shares, saves and comments moved into engage-through, which has a fixed one-day window. Add the engage-through column and most of the missing volume reappears. A portion is genuinely gone, social interactions followed by a purchase on day two or later now fall outside every window.

2. Should I switch to incremental attribution?

It suits accounts with conversion volume to spare and budgets large enough that overstating credit gets expensive. Poor fit if you're already struggling to exit learning, since incremental reports fewer conversions and you lose the ability to edit windows once selected. If you switch, treat it as a fresh baseline, never compare incremental CPA to your historical standard CPA.

3. Which attribution setting should I use to rank creative?

Whichever one you'll commit to keeping. Consistency matters more than the choice, because a ranking is a comparison and a comparison only holds if the measurement is identical on both sides. 7-day click with view-through excluded is a sensible default. Set it once, rank everything on it, and make model changes in budget reporting instead.

‍