Meta's 2026 Attribution Changes Explained: Link-Click-Only Reporting and Incremental With Cost Caps

The first thing I saw when I opened the account was a reported ROAS that had fallen off a shelf. Down something like 20% inside a fortnight, and the founder had flagged it in red. No spend change, no creative change, nothing in the account history to explain it. Just a number that used to say one thing and now said another.

It wasn't a performance drop. It was a reporting change, and there are two of them landing in 2026 that you need to be able to tell apart from an actual decline. Because if you can't, you'll do what that founder nearly did, which is start cutting things that were working fine.

Let me decode both, because they're easy to confuse and they pull in different directions.

Change one: click attribution now means link clicks only

Here's the old behaviour, which most people never realised was happening. When Meta reported a click-attributed conversion, "click" meant any click on the ad. Not just the link. Someone tapping to expand the caption, hitting pause, opening the comments, tapping the image to enlarge it, any of that counted as a click. And once they'd clicked, any purchase in the next day or seven days could be credited to that ad.

You can see why that was generous. Meta could serve a deliberately clicky ad, the one with hundreds of comments you can't help opening, to someone already about to buy, and then take the credit when they bought. The click didn't change their mind. It just put Meta's hand up at the right moment.

From late March 2026, that stops. Click attribution now counts link clicks only. Someone has to actually click through to your site to be eligible as a click-driven conversion. Tapping the comments no longer does it.

That's a meaningful tightening, and on balance I think it's a good change. It should sharpen how Meta optimises, because it's no longer chasing cheap incidental clicks that signal nothing about buying intent. The knock-on effect is that ads which were quietly hoovering up spend by farming cheap clicks on cheap placements will likely get less budget going forward. If you've got a top spender that always struck you as oddly clicky, watch it across the March line. It may settle down.

The catch is what it does to your reported numbers. Because you're no longer crediting all those soft, any-click conversions, your in-platform ROAS and conversion counts come down. Not because you sold less. Because Meta stopped counting sales it had a weaker claim on in the first place.

Why this actually closes a gap you've been complaining about

If you run a third-party tool, Triple Whale, Northbeam, whatever you trust, you'll know the frustration of Meta's number and your tool's number never agreeing. Meta always read high. You'd spend half a reporting call explaining the gap to whoever signs the cheques.

A big reason for that gap was the any-click thing. Your external tool can't see someone who tapped the comments and bought later. It only sees a real click through to the site. So Meta was counting a category of conversion your tool structurally couldn't, and the two numbers drifted apart.

Tightening Meta to link clicks pulls the two closer together. Meta's stated reason for the change is exactly this, getting in-platform metrics to line up better with what brands measure outside the platform. So the reporting drop you're seeing isn't a loss. It's the platform number walking back toward the truth you were already tracking elsewhere. I'd frame it to your team as the dashboard getting more honest, not the ads getting worse.

Change two: incremental attribution can now use cost caps

This is the quieter one, and almost nobody announced it, so it's worth spelling out properly.

Meta rolled out an incremental attribution setting around the middle of 2025. The idea is genuinely useful: instead of reporting every conversion it can reach for, it reports the conversions it believes happened because of your ads. Incremental, meaning the sale wouldn't have happened otherwise. It's a different lens from 7-day-click, and a more honest one if what you care about is real, additional revenue rather than credit-claiming.

The problem at launch was that you couldn't pair it with cost caps or bid caps. So if you ran a cost-controlled buying strategy, and a lot of serious accounts do, you couldn't really use it. It was off the table for the people most likely to want it.

That changed quietly in December 2025. Incremental attribution now works with cost caps. There was no fanfare, no email from your rep, nothing much on the timelines. I only went looking because a few of the usual sharp voices in the space mentioned it in passing. If you asked your Meta contact for an update and got nothing, you're not alone.

Why it matters: it opens incremental attribution up to cost-capped accounts, which is most of the ones spending real money. For a brand that's already well known, running on other channels too, that's potentially a better optimisation target. You're telling Meta to chase sales it's genuinely causing, at a controlled cost, rather than sales it can merely associate itself with.

I'll be honest about my confidence here. I haven't run it across enough accounts and enough time to call it the default yet. What I'd say is that for bigger spenders it's now worth a proper test, where six months ago it wasn't even available to test. That's the shift.

How to read these two changes together

Stack them up and the through-line is the same. Meta is moving its own reporting toward incrementality, toward only taking credit for sales it had a real hand in.

The link-click change does it bluntly, by refusing to count clicks that meant nothing. The cost-cap change does it by opening up the setting that measures genuine lift to the accounts that buy on caps. Different mechanisms, same direction of travel.

And both produce the same surface symptom: in-platform numbers that look lower than they used to. That's the bit your founder, your board, or your own gut will misread as a decline. It isn't one. It's the same business with a more conservative scorekeeper.

The mistake I'm watching people make is treating a reporting reset as a performance event. They see the ROAS dip, they yank budget off the top campaigns, and then real performance actually does fall because they pulled spend off things that were working. The reporting change becomes a self-fulfilling decline, caused entirely by reacting to it wrong.

What we're changing in accounts because of this

So here's what this turns into in practice, rather than just theory.

• Reset the expectation before the number moves, not after. If you wait for the dip and then explain it, it reads as an excuse. Tell whoever watches the dashboard, in advance, that in-platform ROAS is going to step down by some margin and here's exactly why. Get ahead of the reaction.
• Re-anchor on a number the change doesn't touch. Blended MER, total revenue over total spend, doesn't care how Meta attributes anything. Through a period where in-platform reporting is shifting, your blended number is the steady hand. Lean on it.
• Re-examine your top spenders across the March line. Specifically the ones that always looked a bit clicky. Some will lose budget now that incidental clicks don't count, and that's the system working, not breaking. Don't reflexively rescue them.
• Test incremental-with-cost-caps deliberately, on bigger accounts. Not a wholesale switch. A controlled test on the accounts with enough spend and external measurement to read the result honestly, then decide.
• Stop comparing this month's reported ROAS to last month's like-for-like. Across the change, you're comparing two different definitions. Note where the line falls in your records so nobody benchmarks against a number that no longer means the same thing.

None of this is dramatic. It's mostly about not letting a definitional change masquerade as a business problem, and quietly restructuring expectations so the reporting drop reads as what it is.

The deeper question these changes raise is the one worth sitting with. If the platforms are slowly walking their own reporting toward only counting sales they truly caused, then the comfortable in-platform ROAS a lot of brands have steered by for years was always softer than it looked. So what's the real number you'd run your account on, the one that wouldn't flinch whether Meta tightened its definitions or not? That's the thing I'd be working out now, before the next change lands without an announcement.