If your Meta conversions fell off a cliff in 2026 while your spend, creative and audiences stayed identical, this is almost certainly why: Meta narrowed the definition of a click in March 2026, so likes, saves, shares and comments no longer count as click-through conversions. They moved into a new bucket called engage-through, which has a 1-day window instead of the 7-day one they used to enjoy. Nothing about your ads got worse. A rule changed underneath your reporting.
The part almost nobody says out loud is what happens to the conversions that fall between those two windows. They don't move to another column. They are simply not attributed to your ad at all.
What Meta Actually Changed
Two things shipped together in March 2026, for campaigns optimizing toward website or in-store conversions.
Click-through attribution got narrower. Meta used to count all sorts of ad interactions as "clicks" — a like, a save, a share, expanding a caption. Now only a genuine link click that lands someone on your site counts toward click-through. Meta's stated reason is alignment: third-party tools like GA4 only ever counted link clicks, which is a large part of why Ads Manager and Analytics never agreed on anything.
Engaged-view became engage-through, and got broader. The old engaged-view bucket only credited video views of 10 seconds or more. Engage-through now covers all those non-link social interactions plus video views at a 5-second threshold. Meta's justification for shortening the video threshold is Reels behavior — the company has cited that 46% of Reels purchase conversions happen within the first two seconds of attention. Watch time stopped being a useful proxy for intent on a surface people scroll that fast.
You can read the mechanics on Meta's own engage-through setup page, and Search Engine Land's write-up covers the announcement.
Billing didn't change. Delivery didn't change. Only the arithmetic behind the numbers you make decisions with.
The Conversions That Just Disappeared
Here is the mechanic that costs agencies the most and gets the least coverage.
Picture a real journey. Someone sees your ad on Monday, saves it to look at later, doesn't click. On Thursday they remember, search your brand name, and buy.
Before March 2026: that save counted as a click. The Thursday purchase landed inside the 7-day click window. Meta credited the conversion.
After March 2026: the save is an engage-through interaction, and engage-through only reaches back 1 day. Thursday is day four. The conversion is not credited to engage-through, it is not credited to click-through, and it does not get reassigned anywhere else. It falls out of the report.
That's a real sale, genuinely caused by your ad, that Meta's own reporting now disowns. Your CRM still has it. Ads Manager doesn't.
The businesses hit hardest by this are the ones with the longest consideration gap and the highest social engagement — coaching, high-ticket services, considered-purchase DTC, anything where people save the ad because they're not buying at 11pm on a phone. Those are exactly the accounts where Meta's reported CPA suddenly looks 30 to 50 percent worse than the finance spreadsheet says it is.
If you've had this argument with a client since spring, this is the answer. And it's the same structural problem we wrote about in TikTok's Events API attribution window: a platform ships a window that's shorter than the customer's actual decision cycle, and everyone quietly under-reports.
Your Historical Comparisons Are Now Invalid
This is the second-order damage and it's worse than the first.
Every trend line you have that crosses March 2026 compares two different definitions of a conversion. Year-over-year decks. Quarter-over-quarter performance reviews. The "we used to hit a 3.2 ROAS" benchmark somebody wrote into a scope of work in 2025.
None of those comparisons are like-for-like anymore. A campaign that looks 25 percent worse this September than last September may be performing identically, or better. You genuinely cannot tell from the platform number alone.
What we do on client accounts:
- Rebaseline every target on post-March data. Pick a clean 60-day window after the change and treat that as the new zero. Anything older is a different measurement system.
- Annotate the change in every recurring report. Put a line on the chart at March 2026 so nobody, including future you, reads the step-down as a performance collapse.
- Rewrite the targets in the contract if they were set pre-change. A ROAS target agreed under the old counting rules is not the same target under the new ones, and pretending otherwise sets up a fight you'll lose.
Skipping that last one is how good agencies get fired for a reporting change they didn't cause.
Should You Leave Engage-Through On?
The setting lives at the ad set level, under Optimization & Delivery, behind Show More Options. Defaults for website conversion campaigns are now 7-day click-through, 1-day engage-through, 1-day view-through.
Our default recommendation splits by business model:
- Lead gen and local services: turn engage-through off. It's the softest credit in the account, it flatters top-of-funnel creative, and if you're optimizing toward booked calls you want the model chasing people who actually clicked. Report click-through only and prove lift separately.
- DTC with heavy Reels and video: leave it on, but report it separately. On visual-discovery buying, social engagement genuinely does precede purchase and stripping that signal out understates the channel. Just never let it sit blended into one conversion number on a client dashboard.
- Anything with a sales cycle longer than a week: turn it off and stop caring about it. A 1-day window is structurally useless to you. Your measurement has to live in the CRM regardless, so don't spend attention tuning a column that can't see your customers.
The one thing that's always wrong is flipping the setting mid-quarter. Attribution setting is part of what the optimization model learns from, so a change on a scaled ad set buys you a re-learning period and a discontinuity in your own trend line, at the same time. Make the change at a campaign boundary, once, and hold it.
What Incremental Attribution Does and Doesn't Solve
Meta's incremental attribution option, now widely available for Sales and Leads objectives, optimizes toward conversions the model predicts were caused by the ad, rather than every conversion that touched it. It's a genuinely better objective than last-touch, and it's worth testing.
It is not a fix for what we've described here. Incremental attribution changes what Meta optimizes toward. It doesn't recover the conversions that fell outside the windows. Your reported volume goes down further, not up, because the model gets more conservative about claiming credit. That's honest, and it's also a hard conversation to have with a client in the same quarter that engage-through already cut their numbers.
Sequence it. Absorb the engage-through rebaseline first, get everyone comfortable with the new normal, then test incremental as a separate deliberate change with its own before-and-after.
The Setup That Doesn't Break When Meta Changes the Rules
Every window narrowing over the last five years has punished the same thing: reporting that depends entirely on what the ad platform is willing to tell you.
The accounts we manage that barely felt March 2026 all had the same three things in place beforehand.
Server-side conversions with persisted click IDs. Capture fbclid on first landing, write it to a first-party cookie, and persist it against the lead record in your database — then attach it from storage at conversion time, never from the current URL. Send hashed email and phone alongside it. Our Meta Conversions API setup guide covers the build. Deterministic matching doesn't care how Meta redefines a click.
An attribution layer that isn't owned by the ad platform. This is the whole argument for running Hyros or an equivalent independent tracker: one system that sees every touch across Meta, Google and TikTok and reconciles to closed revenue, so a unilateral change inside one ad platform can't rewrite your history. We laid out how we deploy that in our Hyros attribution guide for agencies.
Weekly reconciliation against closed-won revenue. Not leads. Not platform conversions. Money that arrived. Feed those outcomes back into Meta as offline or CRM conversions so the algorithm optimizes toward customers instead of form fills. This is also the only reporting number that survives the next change, whatever it turns out to be — and there will be a next one. Meta's aggressive automation shift, which we covered in recovering ROAS under Andromeda, moves in exactly one direction: less manual control, less granular reporting, more trust required.
Build for that, and a March-2026-shaped event becomes a footnote in a report instead of an emergency.
The Honest Summary
Meta made its reporting more accurate and less flattering at the same time. Narrowing click-through to real link clicks is defensible — arguably overdue, since it finally puts Ads Manager and GA4 in the same conversation. Shortening the video threshold to 5 seconds matches how people actually consume Reels.
The cost is that a slice of genuinely ad-driven revenue now has nowhere to be counted, historical comparisons broke silently, and nobody sent you an email about it.
That's the job. Platform numbers are a rented signal. The ones you own are in your CRM, and everything we build on the attribution side exists to keep those two as close together as they can honestly get.
Find Out What Your Reporting Is Missing
We audit Meta, Google, TikTok and Hyros setups for exactly this: attribution windows, click-ID capture, server-side match quality, and how far platform-reported conversions have drifted from closed revenue in your CRM.
Get a Free Automation Audit and we'll show you what the engage-through change cost you, and what recovering that visibility is actually worth.