GA4 Custom Conversion Windows: How to Set the Lookback That Fits Your Sales Cycle

Every conversion action in GA4 carries a lookback window: the rule that decides how far back a marketing touchpoint can sit and still claim credit when the conversion happens. For most properties that rule has been one size fits all. Clicks were measured on a 30-day window, engaged views on a fixed 3 days, whether the client sells £30 repeat purchases or £30,000 software on a six-month sales cycle.

On 11 August 2026 Google changed that. Click-through conversion (CTC) windows can now be any whole number from 1 to 90 days, and engaged-view conversion (EVC) windows any whole number from 1 to 30 days. The window is set per conversion action, so a newsletter signup and a booked demo no longer have to share a rule. The GA4 changelog entry is short. The effect is not, because most setups have never touched this setting, and the default has been quietly mis-measuring slow sales for years.

What changed on 11 August

Before the update, the click-through window on a GA4 conversion action was a picklist: 1, 7, 14, 30, 60, or 90 days. You could not enter 45 or 75. The engaged-view window was not a picklist at all. It was fixed at 3 days and could not be changed anywhere in the interface.

Since 11 August both numbers take any integer:

• Click-through conversion windows: 1 to 90 days.
• Engaged-view conversion windows: 1 to 30 days.

The engaged-view part is the one most teams will miss. An engaged view is a view that met the engagement threshold without producing a click. If you have switched on engaged-view conversions for a key event, the old 3-day limit decided whether that view earned any credit at all. Three days suits an impulse product. It does not suit a service people research for three weeks before acting.

Where the setting lives

The window is configured per conversion action, not per property. That matters more than it sounds: two actions in the same property can now follow different sales cycles, which is the whole point of the release.

In GA4 the path is Advertising, then Conversion management, then the three-dot menu on the action (labelled more options), then Settings. The same conversion action appears in the linked Google Ads conversion management interface, and both consoles read from that shared action, so the window you set in GA4 is the window Google Ads reports against. Set it in one place and verify the value in the other view before you close the tab.

Which actions carry the setting? Every action under Conversion management, including Google-marked events such as purchase and custom key events you have marked for conversion. The two acquisition events are the exception: first_visit and first_open keep a 30-day default with a 7-day option, because they mark when someone became a user rather than when a campaign paid off. Expect to leave those alone.

Why the 30-day default was wrong for most clients

A window is a filter on the credit pool. Touchpoints older than the window are invisible to the conversion, no matter how well the attribution model likes them. The model can only split credit between touchpoints that are still in the pool.

Now apply that to real client shapes:

• E-commerce with a short path. Most purchases happen within days of the first click. A 30-day window keeps old clicks in the pool for weeks after they stopped influencing anything, so credit spreads across channels that did the early work while the channel that closed the sale gets diluted.
• B2B and high-ticket services. A lead reads three posts in June, returns in July, books a call in August. With a 30-day window the June touchpoints drop out entirely. The August session looks like the whole story, content looks useless, and paid search looks as if it did everything.
• Local and professional services. Enquiries usually land within a week or two of the trigger. Windows of 60 or 90 days would let a random click from last month steal credit from the search that actually converted.

Run those three client types on one default and you are mis-measuring at least two of them. That is the quiet cost this release removes.

Pick the window from data, not from the dropdown

The wrong reaction is to set everything to 90 days because longer sounds safer. Long windows put more conversions in the pool, but they also spread credit across more touchpoints and blur the story. Pick each window from the time-to-conversion distribution for that specific action.

1. If the property exports to BigQuery, measure the gap between the first touch in a conversion path and the conversion itself, per converted user. Look at the distribution, not the average.
2. No BigQuery export? Use the Google Ads time lag report, or the CRM's own lead-to-close data as a proxy. You need the shape of the curve, not a precise figure.
3. Take roughly the 90th percentile of the distribution. The average is dragged down by the mass of fast conversions and will still cut off the slow tail you are trying to keep.
4. Round to a sensible whole number, add a small buffer, and set it for that action. Then move to the next one.

A few worked shapes: an e-commerce client whose conversions cluster inside 9 days gets 14. A SaaS client whose demo bookings arrive between 20 and 55 days gets 60. A trades business that converts within 5 days gets 7. The numbers come from the data; the dropdown just makes them possible.

If you are querying BigQuery, keep the shape simple. Take the users who converted on the action, find their first session_start in the weeks before that conversion, and measure the gap in days. Plot the gaps and read the 90th percentile off the curve. Cross-device journeys and merged sessions will blur the edges, which is exactly why you use a percentile with a buffer rather than an exact day. For a client without BigQuery, the Google Ads time lag report gives the same shape for paid touchpoints and the CRM gives the full picture; either is good enough to choose between 30, 60, and 90.

What still does not change

Three things stay the same, and teams that miss them will be confused later.

• The change is not retroactive. Historical conversions are not reprocessed with the new window. Set the window and let it accumulate forward; do not expect last quarter's numbers to shift.
• The attribution model is separate. The window decides which touchpoints are eligible for credit; the model decides how the credit is split between them. Changing one does not change the other. The earlier guide to GA4 attribution models covers the split side.
• Acquisition key events keep their own rules. first_visit and first_open still run on the 30-day default with a 7-day option, because they measure when someone became a user, not campaign credit.

One consequence deserves its own warning: changing a window changes the conversion count from the moment you save it. A month-over-month comparison will show a step change that has nothing to do with marketing performance. When you widen a window for a client, note the change in the report that covers the switchover, or the client reads the jump as a win and you explain it twice.

Consent mode makes the choice bigger

Most UK properties now run consent mode v2 with analytics_storage denied until a visitor accepts, which means a share of conversions is modelled: reconstructed from observed patterns rather than recorded directly. The modelling works inside the window you set. Widen the window and the modelled span widens with it.

Two habits keep this honest:

• When you change a window, check the modelled share in the conversion reports before and after. A jump in modelled conversions is not proof the new window works; it can be the model filling a wider span with estimates.
• Do not let micro-conversions inherit the revenue window. A newsletter signup that converts within a week does not need 90 days of credit eligibility. Give it 7 or 14 and keep the long window for the event that pays.

Longer windows generally help lower-traffic properties more than they hurt: the extra conversions add signal, and the wider credit spread matters less at small volumes. High-volume properties feel the blur sooner. Set the window per client, not per fashion.

What to check on client accounts this month

If you manage several properties, run this pass once, then revisit it quarterly:

• List every conversion action and the window currently set on it. Most will still show the old default; that is the point of the audit.
• For each action, decide whether the client's sales cycle is shorter or longer than 30 days, and set the window from the data rather than from the list.
• Keep micro-conversions on short windows even when the revenue event moves to 60 or 90.
• Confirm the same numbers in Google Ads, where the imported GA4 conversions mirror the shared action.

The release does not force anyone to change anything. It removes the excuse for leaving every client on a window that was never theirs.

Record it in the measurement plan

A window decision without an owner and a date is drift waiting to happen. When the measurement plan exists, add one line per conversion action:

• the window you chose,
• the data behind it (the 90th percentile of time to conversion, measured on a named date),
• the date it was set and who set it,
• the trigger to revisit: a change in offer, pricing, funnel, or traffic mix.

Revisit the windows when the client changes what they sell or how they sell it. A 90-day window on a product relaunched as a flash sale is a new mistake. Google finally built the dial; the work is using it deliberately, per client and per event, and writing the decision down. Do that and the conversion numbers you report start matching the way your clients actually buy.