On March 3, 2026, Meta changed the definition of a click, and the reported numbers on almost every ad account moved without a single dollar of spend changing. If your Meta Ads Manager conversions and ROAS have looked worse since early spring while your CRM and Google Analytics told a calmer story, this is very likely why. The underlying campaigns did not necessarily get worse. Meta just started counting differently.
On March 3, 2026, Meta narrowed click-through attribution to actual link clicks and moved likes, shares, comments and saves into a new one-day engage-through window. Reported click-through conversions and ROAS have dropped for many advertisers since, without any real change in ad performance. Read your account by attribution type before you react to the headline number.
What actually changed on March 3, 2026?
Meta announced a structural overhaul of how it attributes conversions to ads, first reported by Search Engine Land. Click-through attribution now requires a real link click: one that sends a person to your website, app, lead form or another destination. Likes, shares, saves and comments, which used to count toward click-through conversions in some cases, no longer qualify. Instead, they moved into a new category called engage-through attribution, which runs on a one-day window. Under the old system, some of those non-link interactions could still convert a customer up to seven days later and get counted. Under the new system, that late conversion window on non-link engagement is simply gone.
Meta's stated reason is to close the gap between what Ads Manager reports and what third-party tools like Google Analytics show, which had been a long-running source of confusion for advertisers reconciling two different conversion counts. Narrower attribution should, in theory, make Meta's number look more like everyone else's number.
Why your reported ROAS dropped without your ad performance changing
Here is the mechanical reason. If a chunk of your reported click-through conversions used to come from social engagement rather than actual link clicks, and especially from engagement that converted on day three, four or five under the old seven-day window, those conversions simply stop appearing in the click-through number after March 3. Ad spend is unchanged. The site or store is converting the same visitors at the same rate. But the headline conversion count, and therefore the headline ROAS, is now smaller, because Meta is measuring a narrower thing and calling it the same name.
This matters most for accounts that lean on engagement-heavy placements, video-first campaigns, and awareness-to-conversion funnels with longer consideration windows. It matters less for direct response search-style campaigns where most conversions already came from an immediate link click.
Click-through vs engage-through vs incremental attribution
Meta now effectively reports three different things under three different names. Knowing which one you are looking at changes what decision you should make from it.
| Attribution type | Window | What it counts | What to use it for |
|---|---|---|---|
| Click-through (new definition) | 1 day | Real link clicks to your site, app or lead form only | The fastest, most conservative read of direct response |
| Engage-through | 1 day | Likes, shares, saves, comments and non-link video engagement | Middle-funnel visibility, not a sales number to bid on |
| Incremental (Conversion Lift) | Length of the test, days to weeks | Holdout-tested conversions actually caused by the ad | The number to trust for real ROAS and budget decisions |
Meta itself calls incremental measurement, built on its Conversion Lift product, the gold standard, since it compares an exposed group against a holdout group rather than counting everyone who happened to click or engage. Meta has reported roughly a 24% increase in incremental conversions identified since the latest lift model rollout in January 2026 compared to its prior standard attribution model, though that figure comes from Meta's own reporting and is worth treating as directional rather than an external audit.
Is incremental attribution actually better, or just more convenient for Meta?
Both things can be true. Holdout testing genuinely answers the question advertisers actually care about, which is whether the ad caused the sale, not whether a customer happened to see the ad at some point before buying. That is a real methodological improvement over click-window attribution, which has always overcounted conversions that would have happened anyway. At the same time, a measurement approach that reports higher performance is also the one a platform has an obvious incentive to promote. Run your own Conversion Lift tests on your highest spending campaigns rather than taking the reported incremental number on faith, and compare it against your blended ROAS from outside Meta's own reporting.
Do not reset your CAC targets or bid strategies off the March 2026 dip alone. Pull cost per result by attribution type for February and for June, side by side, and separately check your CRM or ecommerce platform's own order count over the same period. If orders and revenue outside Meta's reporting held steady, the platform's number moved, not your business.
What to check in your account this week
- Compare reported conversions and cost per result for the four weeks before March 3, 2026 against the four weeks after, split by click-through and engage-through where the reporting allows it.
- Check whether any Target CPA or Target ROAS bid strategies were set using pre-change numbers, since a bidding target calibrated on inflated old click-through data can now underdeliver against a stricter definition.
- Turn on or review Conversion Lift testing for your top two or three spending campaigns, so you have a holdout-tested number to compare against Ads Manager's headline figure.
- Reconcile Meta's reported revenue against your own order data at least monthly rather than trusting either number in isolation. This is the same discipline that matters for blended reporting across every platform, not just Meta.
Treat this as a permanent reminder to build reporting around your own conversion data, not any single platform's dashboard. Platform attribution methodology will keep changing as measurement gets harder in a cookieless, privacy-first environment. A CRM-anchored view of CAC is the only number that does not move when a platform redefines a click.
Frequently asked questions
What did Meta change about attribution in 2026?
On March 3, 2026, Meta narrowed click-through attribution to actual link clicks only, meaning a click that sends someone to your website, app or lead form. Likes, shares, comments and saves no longer qualify as click-through conversions and moved to a new engage-through attribution category with a one-day window.
Why did my Meta Ads conversions drop without changing my campaigns?
Interactions that used to count as click-through conversions under a wider window, including non-link social interactions, no longer qualify. The same ad performance now reports a lower click-through conversion count and a lower reported ROAS, purely because of the definition change, not a drop in actual results.
What is engage-through attribution on Meta Ads?
Engage-through attribution is a new category that captures social interactions like likes, shares, saves, comments and non-link video engagement that lead to a conversion. It runs on a one-day window and is reported separately from click-through, rather than folded into it.
What is incremental attribution on Meta Ads?
Incremental attribution, built on Meta's Conversion Lift product, uses holdout testing to estimate conversions actually caused by an ad versus conversions that would have happened anyway. Meta describes it as the more reliable measurement of true ad impact compared to standard attribution windows.
Should I still trust blended ROAS after this change?
Blended ROAS from your own CRM or analytics platform is now more useful relative to Meta's reported number than before, since Meta's figure moved due to a definition change rather than a real performance change. Compare trend lines across the March 2026 changeover instead of absolute numbers from before and after.
What should I check in my Meta Ads account now?
Check reported conversions and cost per result before and after March 3, 2026, whether your CAC targets and bid strategies were set using pre-change numbers, and whether Conversion Lift tests are running on your top spending campaigns to give you a measurement Meta trusts more than raw attribution.
The takeaway
Meta did not quietly make your ads worse in March 2026. It quietly made its own reporting stricter, and the gap between what it used to count and what it counts now landed on your dashboard as a drop. Read cost per result by attribution type, lean on incremental testing for real budget decisions, and keep a CRM-anchored view of CAC that does not shift every time a platform redraws the line around what counts as a click.