Meta's March 2026 Attribution Overhaul: Why Your Reported Conversions Dropped 10-30%

On March 3, 2026, Meta changed the ruler, not your results. Click-through attribution was restricted to link clicks only, and view-through attribution was renamed engage-through with a one-day attribution window. Reported conversions fell 10 to 30 percent overnight across advertisers, while actual sales did not move. Your ads did not suddenly get worse. Meta simply stopped counting conversions that followed non-link clicks like reactions, comments, and shares, and the drop exposed how much of the old number was never causal.

What exactly did Meta change on March 3, 2026?

Two changes, same day, no transition period.

First, click-through attribution was restricted to link clicks only. Before March 3, a click-through conversion could be credited after almost any click on an ad: a reaction, a comment, a share, a tap to expand the text, a click on your page name. Under the new rules, only clicks on the link itself earn click-through credit. Every conversion that used to ride on those other interactions left the column overnight.

Second, view-through attribution was renamed engage-through and given a one-day window. The new name is more honest about what it measures, exposure and light engagement rather than a deliberate click, but the tighter window cut the volume of conversions eligible for this credit at the same time.

For clarity, a link click is exactly what it sounds like: a click on the outbound link in your ad, the action that sends someone to your site. Everything else, the likes, the comments, the shares, the taps on your page name, is engagement with the ad unit itself. Meta's change notes live in its Ads Help Center updates from early March 2026, and the new definitions applied across advertisers at once.

There was no grandfathering for historical comparisons. Reports that span March 3 mix two counting systems, so annotate the date in every dashboard you own. We logged the change alongside the year's other platform moves in the 2026 attribution changelog.

Why did reported conversions drop 10 to 30 percent overnight?

The mechanics are simple subtraction. Remove non-link clicks from click-through credit, and a chunk of previously counted conversions disappears. Tighten view-based credit to a one-day engage-through window, and another chunk goes with it. Nothing happened to your ads, your audience, or your revenue. The counting rules moved.

Where you land inside the 10 to 30 percent range depends on your account's mix. Accounts heavy on retargeting and engagement-style creative collected more credit from non-link clicks and view-based windows, so they lost the most on paper. Accounts running tight link-click prospecting saw smaller drops. Same policy, different exposure.

One housekeeping warning: any report that spans March 3 blends two counting systems in a single row, which makes week-over-week comparisons across the seam meaningless. Split your analysis there and treat February and April as different eras.

Here is what the shift looks like on one illustrative prospecting campaign with €60K of monthly Meta spend and a €55 average order value. Same campaign, same budget, same reality; only the definitions change.

A 24.5 percent reported decline with zero change in revenue. That is the whole story of March 2026 in one table.

Why platform-reported numbers were never causal

Sit with that table for a moment. If a definitions change can erase a quarter of your reported conversions without touching a single euro of revenue, then the reported number was never measuring what you thought it measured. It was measuring contact: conversions that happened after someone saw or touched your ad. The causal question, did this ad change what anyone did, was never on the table.

Retargeting is where this gets uncomfortable. Serve an ad to someone already heading to checkout, let them convert, collect view or engagement credit: the old system called that a win. The test for whether that credit was deserved is the counterfactual, would this person have bought anyway, and it is the same test behind our retargeting cannibalization analysis.

None of the familiar models escape this: last click attribution describes where credit landed, not what caused the sale, and multi-touch attribution spreads the same untested assumption across more touchpoints. An attribution model is a credit-assignment convention; incrementality is a claim about causation. March 3 simply made the gap between the two visible to everyone at once.

None of this means Meta ads stopped working. For many DTC brands they remain the primary growth engine. The point is narrower: the platform's counter was never the instrument to prove it, and March 3 proved that in public. A number that can drop by a quarter on a definitions change was carrying a large share of non-causal credit. The honest response is not to mourn the old number but to stop making budget calls on it.

And be honest in the other direction: the new, narrower numbers are not truth either. A one-day engage-through read is a stricter ruler, but it is still the platform grading its own homework.

How should you measure Meta performance now?

See also: How to Measure TikTok Ad Performance for Beauty Brands

1. Keep Meta's numbers for trends, not truth. Week-over-week direction inside the platform is still useful for creative and audience decisions. Absolute conversion counts, and any ROAS figure Meta computes for you, should no longer anchor budget decisions.

2. Triangulate against GA4 with eyes open. GA4 sees a different slice of reality, but its data driven attribution is its own opacity problem, as we covered in our review of GA4's attribution black box. Keep your UTM parameters disciplined so at least the traffic side of the comparison is clean.

3. Anchor budgets on incrementality. Geo holdouts and conversion lift tests answer the causal question directly when you can run them. When you cannot, a causal read on your GA4 export estimates what your Meta spend actually caused, controlling for seasonality and everything else running at the same time.

4. Re-baseline before you re-forecast. Your reported KPI targets from February are invalid under the new definitions. Rebuild them from post-March data: annotate the break, recompute acceptable cost per acquisition and reported ROAS thresholds, and tell stakeholders plainly that the drop is definitional. The worst response to March 3 was panic-cutting spend that was actually working. The second worst was pretending the old targets still applied.

What changes in your week-by-week reporting after March 3?

Rebuild the weekly rhythm around the seam. In week one of each month, compare like with like: post-March weeks against other post-March weeks, never against February. When a stakeholder asks why conversions look lower than last quarter, show the definitions note first and the performance read second, in that order, or you will spend the meeting defending a counting change. Track two trend lines from now on: Meta's reported conversions for direction, and your own store revenue per euro of Meta spend for truth. When the two lines diverge for more than two weeks, that is your cue to pull a fresh GA4 export and re-run the causal read rather than tweaking creative against a broken ruler. Finally, log every Meta policy update in the same annotation sheet you started on March 3. The next ruler change is easier to explain when the last one is already documented.

One practical note for Q3 planning: pull any platform exports with the new definitions only. Mixing pre-March and post-March rows in the same trend line is the most common reporting error we still see in July, four months later, and it makes healthy accounts look broken and broken accounts look fine. If a year-over-year comparison matters to your board, present it alongside revenue from your own store data, not alongside last year's Meta dashboard.

The causal step is what Causality Engine does. GA4 export in, causal read out, 5 to 10 minutes, with stated uncertainty. €99 per read, €299 per month for Pro, no pixel, no annual lock-in. The next time a platform rewrites its ruler, run a causal read on your own GA4 export before you touch the budget.

Key takeaways

• On March 3, 2026, Meta restricted click-through attribution to link clicks and replaced view-through with a one-day engage-through window; reported conversions fell 10 to 30 percent with no change in reality.
• If a definitions change can erase a quarter of reported conversions, the number was measuring contact, not causation.
• Accounts leaning on retargeting and engagement credit lost the most on paper; tight link-click prospecting accounts lost the least.
• The new numbers are a stricter ruler, not the truth: platform-reported conversions remain the platform grading its own homework.
• Anchor budget decisions on incrementality and causal reads of your own data; treat Meta and GA4 as context, not verdicts.

Further reading

• The retargeting cannibalization test: is your branded search actually incremental
• The running changelog of every 2026 attribution change
• A deeper look inside GA4's attribution black box
• Run a causal read on your own GA4 export