GA4 forgets your buyer after 30 days

If your buyers take longer than a month to decide, GA4 is systematically undercrediting whatever brought them in. The acquisition conversion lookback default moved from 90 days to 30, and defaults are what most properties run on.

That single setting decides how far back GA4 will look to find the campaign that first brought a converting user to your site. Set to 30 days, a buyer who saw your ad in March and signed in May is attributed to whatever they touched last, usually a branded search or a direct visit. The channel that actually created the demand shows nothing.

Every business with a considered purchase is affected. Capital equipment, professional services, construction, software with a procurement process, anything with a committee. Which is to say, most of the businesses reading this.

Find your real lag before you change anything

Do not just set it to 90 and move on. Measure first, because the answer tells you more than the setting does.

Look at time-to-conversion in your path data: how many days elapse between first touch and conversion, distributed rather than averaged. Averages hide the shape. What you want is the 90th percentile, because that is the point where you would capture nearly all of your buyers.

Then compare that number to 30.

Three outcomes. If the 90th percentile is under 30 days, leave the setting alone and go do something more useful. If it sits between 30 and 90, change the setting to 90 and you have fixed the problem. If it sits beyond 90, you have a structural problem that no GA4 setting solves, because 90 days is the ceiling.

When 90 days is not enough

This is common in Australian B2B, where a tender cycle or a capital approval can run two or three quarters. GA4 simply cannot see that far on acquisition conversions, so the answer has to live outside GA4.

The fix is a stored first touch. When a visitor first lands, capture the campaign source, medium and campaign name into a first-party cookie with a long life, write those values into a hidden field on every form, and pass them to your CRM on the lead record. They then sit on that record forever. Six months later, when the deal closes, you join the closed-won record back to the original source, in your CRM or your warehouse, with no lookback window involved at all.

Two implementation notes that decide whether this works. Write the cookie server-side with an HTTP header rather than through JavaScript, because Safari caps JavaScript-set first-party cookies at seven days and drops that to 24 hours when the visitor arrived through a link carrying tracking parameters. And store first touch and last touch separately, in different fields. Overwriting one with the other is the single most common way this gets built wrong, and it is invisible until someone asks a question the data cannot answer.

Once that is in place, your source of truth for channel performance is closed revenue by original source in the CRM. GA4 becomes a behavioural tool for understanding what people do on the site, which is what it is genuinely good at.

Where to change the setting

Admin, then Attribution settings at property level. You will see the reporting attribution model and the acquisition conversion lookback window with options of 30 or 90 days, plus a separate lookback for other conversion events with 30, 60 and 90 day options.

Change both if your cycle warrants it. And note that the acquisition window governs the first open and first visit events specifically, which is why it matters so much for channel credit rather than for on-site behaviour.

The knock-on effect nobody flags

Lengthening the window changes your data-driven attribution eligibility. That model needs at least 400 conversions for the action and 20,000 across the property inside the lookback window. A longer window includes more conversions, which can push a property over a threshold it was previously failing.

So changing this setting can silently switch your reports from last click to data-driven, and your channel numbers will move for two reasons at once. Change the window, then wait a full cycle before drawing conclusions about channel performance, and note the change date somewhere your team will see it in three months.

The honest limitation

A longer lookback does not make the credit correct. It makes it less wrong. Attribution windows are a rationing rule for a question that cannot be answered by observation, which is whether the touchpoint caused the sale.

For a business with a six-month cycle, the campaign that mattered most might be a video the buyer watched on a phone while logged out, on a platform that reports nothing to your analytics. No window length recovers that.

Which is why the only reliable read on long-cycle marketing is a holdout: turn the channel off in one region for a full sales cycle and compare closed revenue. Long cycles make that test slow, and slow does not mean optional.

Set the window to match your buyer, then build the CRM stitch that makes the window irrelevant.