Why 1-day click became the sane Meta attribution window

A 1-day click attribution window trains a better model than 7-day click 1-day view, and most accounts are still running the longer default. The reason has nothing to do with being conservative in reporting. The window is not just a reporting choice, it is the label the optimisation model learns from, and a long window feeds it slow, noisy feedback at exactly the moment the ranking system needs the opposite.

The window is a training label

This is the part that gets lost. When you set an attribution window, you are telling Meta which conversions count as a result of which impressions. That mapping is what the delivery model learns from.

With 7-day click 1-day view, a conversion arriving on day six gets attributed back to a click from last Tuesday, and a conversion from someone who saw an ad and never clicked gets attributed to the impression. The model receives that feedback late and mixed with a lot of coincidence, because a week is long enough for many people to have bought regardless.

With 1-day click, the feedback is fast and the causal link is tighter. The model learns from a cleaner signal, more often. Under Andromeda, where the system evaluates a far larger candidate pool per impression, the quality and freshness of that signal matters more than it used to. A model re-ranking more options with worse labels gets worse, not better.

What it costs you

Reported conversions will drop. For a considered purchase, they can drop a lot, because a real proportion of genuine conversions land on days two to seven and the shorter window simply stops counting them.

Your in-platform return will look worse. That is a reporting artefact, not a business outcome, but it will still show up in a board pack and somebody will ask about it.

So set the expectation before the change, not after. Tell the team the number is going down and explain why. Then hold the comparison in the Compare Attribution Settings view for a few months so both readings are visible while people recalibrate.

When not to do it

Three cases where 1-day click is the wrong call.

Long consideration cycles. If your average time from first click to purchase is 18 days, a 1-day window discards most of your evidence. Higher-ticket B2B and considered retail sit here.

Low weekly conversion volume. The model needs enough events to learn. A useful threshold is roughly 50 conversions a week per ad set. If shortening the window drops you under that, you have traded signal quality for signal quantity and lost.

Heavy retargeting programmes measured on view-through. Not because the shorter window is wrong, but because the drop will be dramatic and you should sequence it differently, which brings us to the more important point.

Drop the view-through first

Most of the inflation in Meta reporting sits in 1-day view, not 7-day click. A view-through conversion means somebody saw an ad, did not click, and bought within 24 hours. Some of those are real. Many are people who were going to buy anyway and happened to be served an impression on the way.

If you are going to change one thing, change that one. Move retargeting to click-only attribution and watch what happens to reported volume. The gap you see is a reasonable first estimate of how much of your retargeting reporting was passive credit.

Then, separately, decide on the click window.

The migration

Five steps.

1. 1.Pick one campaign, not the account. Ideally a prospecting campaign with healthy volume.
2. 2.Record the current baseline: conversions, CPA, return, and the same metrics from your own order data over the same period.
3. 3.Change the window at the ad set level and let it run through the learning phase plus a full purchase cycle.
4. 4.Keep both readings visible using the attribution comparison view, so the drop is always contextualised.
5. 5.Rebase your targets. A CPA target set under 7-day click 1-day view is meaningless under 1-day click. If you forget this step, the campaign will look like it failed and someone will turn it off.

The reconciliation nobody does

Whatever window you choose, the number that decides the budget should come from outside Meta. Your order data, your CRM, your blended CAC across all channels.

Platform attribution windows are a way of shaping what the algorithm learns. They are not a measurement of your business. Once a quarter, take total spend across every channel and total new customers, divide one by the other, and compare that trend to what the platforms are each claiming. The gap between those two numbers is the most useful diagnostic in performance marketing and almost nobody plots it.

Shorten the window to teach the model better, and keep the ledger somewhere the model cannot see.