Your Meta dashboard looks worse this month. That is the point.
If reported conversions on your Meta account dropped 15 to 25% in May without anything else changing, the explanation is on the platform side, not on yours. Meta rolled out its largest attribution overhaul in three years between March and early May 2026 1. The change tightened what counts as a click, dropped the engaged-view threshold from 10 seconds to 5, and moved likes, comments, shares, and saves into a separate attribution bucket called engage-through.
Reported conversions look different. The account did not get worse. The work this month is reading the new numbers correctly, not panicking and turning off campaigns that are still working.
What Meta changed and why
Three changes, three reasons.
The first change tightened click-through attribution. Meta now counts a click only if it was an actual link click that loaded a landing page. The old definition counted incidental taps, post-tap dwell, and certain swipe actions that did not necessarily indicate intent 2. The result is a stricter click count that more closely matches what third-party measurement tools have reported all along. Click-through conversions drop. That drop is reporting accuracy, not lost demand.
The second change cut the engaged-view threshold from 10 seconds to 5. Meta cited internal data showing that 46% of online purchase conversions tied to Reels happen within the first two seconds of viewing 3. The implication: a 10-second threshold was too long; meaningful intent forms earlier. The shorter window credits more conversions to video and reduces credit to clicks-after-views that happened beyond Meta’s old window.
The third change is the new engage-through attribution bucket. Likes, reactions, comments, shares, saves, and engaged video views that lead to a later conversion now sit in their own column. Previously these mixed into the click-through column and inflated reported credit for upper-funnel activity 4.
Underneath all three: Meta is trying to publish numbers that look closer to what Northbeam, Triple Whale, and post-purchase surveys have been showing. The platform’s reported ROAS gets less optimistic. The honest ROAS does not change. The gap was always there; now it is in the dashboard.
What this looks like in a real account
A direct-to-consumer skincare account the firm reviewed last week shows the pattern clearly. Reported purchases in Meta Ads Manager dropped from 412 in April to 318 in early May, a 23% drop. The actual purchases reported in Shopify for the same period rose 4%. Spend was flat. Nothing in the account changed.
The 23% drop was the new click-through definition removing the incidental-tap conversions that Meta used to count. Those conversions were always real; they just were not strictly attributable to a click. They were happening through some combination of brand recall, retargeting, organic, and direct traffic. Now they sit outside the click-through column, with some moving to engage-through and some falling out of Meta’s reported attribution entirely.
This is the pattern across most accounts the firm has audited this month: Meta-reported conversions down 15 to 25%, real revenue flat or up, no underlying account change. Operators who do not understand the attribution change are pausing campaigns that are still performing.
How to read the new Conversion Count Breakdown
Meta also rolled out a new view called Conversion Count Breakdown, which shows first-click, last-click, and data-driven attribution side by side 1. The view is buried in the Columns menu under “Performance and Clicks.” Turn it on. Compare the three numbers for any campaign older than two weeks.
When first-click is meaningfully higher than last-click, your upper-funnel creative is doing the work and your bottom-funnel creative is taking credit. When last-click is higher, the reverse. When data-driven sits between the two and tracks the same direction over time, the platform’s attribution model is internally consistent and you can trust the relative comparisons across campaigns.
The absolute numbers will all be lower than they used to be. That is fine. What matters is the relative performance across campaigns, and that comparison is now more honest than it was three months ago.
What to do with reported CPA looking worse
The instinct when CPA goes up is to fix something. Resist for two weeks. Most of the May CPA inflation in Meta accounts is the new attribution model, not a real performance shift. Wait for the next full reporting cycle, look at the trailing 28-day window under both old and new attribution definitions if your historical data is still available, and compare to your downstream revenue.
If reported CPA is up 20% but Shopify, your CRM, or your post-purchase survey shows revenue holding or growing, the account is fine. Continue spending. The dashboard is just reading more accurately than it did in March.
If reported CPA is up 20% and downstream revenue is also down, you have a real problem and the attribution change is masking it. The fastest way to separate the two is to look at engage-through conversions for the same campaigns. If engage-through is meaningfully higher than it was, the demand is still there; just spread differently across credit columns. If engage-through is also down, the audience interest has shifted.
Engage-through is directional, not measurement
This is the part most operators are getting wrong. Engage-through is a credit category for actions that signal interest but do not constitute a click 2. It is useful as a creative diagnostic: ads with high engage-through volumes are landing emotionally even when they do not generate clicks. It is not a substitute for click attribution and it should not be summed with click-through to defend a campaign’s ROAS.
Treat engage-through the way you would treat video completion rate: directional, useful for picking which creative to scale, useless as a primary KPI. The campaigns that survive the next quarter are the ones that look reasonable on click-through alone. Engage-through is a tiebreaker, not a defense.
Where to spend the diagnostic hour this week
Open the campaign that produced your best result in February. Pull its current data. Add the Conversion Count Breakdown view and the engage-through column. Note the relative shape of attribution across click-through, engage-through, first-click, and data-driven. That shape is your new baseline.
Run the same exercise on the campaign that produced your worst February result. The comparison will tell you which campaigns the new attribution model is treating more harshly and which are roughly the same. The first group needs more patience before you cut. The second group can be evaluated normally.
A separate reminder, since creative fatigue has compressed in parallel with these attribution changes: an ad showing CPA drift in May 2026 may simply be on its fatigue cycle, which the firm covered in the creative fatigue piece. Look at the creative’s age first. Fatigue and attribution can compound in a way that makes a healthy campaign look broken when the only real fix is a new creative variant.
The dashboard got more honest. The work did not change.
The single sentence to carry into the next reporting meeting: Meta’s reported numbers are closer to your real numbers than they used to be. That is good. It makes the conversation with the CFO easier next quarter, when the dashboard and the bank account finally agree.
What did not change is the work that moves the number. Creative quality. Offer. Landing page. Bid pacing. Audience clarity. Spend less, on fewer things, with creative that earns the five seconds. The accounts that win the new attribution era are the same accounts that won the old one. The only difference is that the scoreboard now reads them honestly.