Definition
Every attribution system has to decide how much time may pass between an interaction and an outcome before the two stop being connected. That limit is the lookback window. A conversion that happens inside it is eligible for credit; one that happens a day after it closes is not, however clearly the ad started the journey. The window does not decide which touchpoint gets the credit, which is the job of the AttributionAssigning conversion credit to touchpoints. model; it decides which touchpoints are allowed into the contest at all.
Most systems keep at least two windows:
- The click-through window covers conversions that follow a click on the ad.
- The view-through window covers conversions by people who saw the ad without clicking it. It is usually much shorter, because the link between an unclicked impression and a sale is weaker.
Some platforms add a third window for engaged video views. The vocabulary also varies: Google Ads calls it a conversion window and measures it forwards from the ad interaction, while GA4Google's event-based analytics platform. calls it a lookback window and measures it backwards from the key event. The two descriptions arrive at a similar idea from opposite ends, and a conversion that falls inside one window and outside the other is counted by one system only.
In practice
Each place a window lives answers a slightly different question.
- Google Ads sets windows per conversion action, separately for clicks, engaged views and views.
- GA4 sets property-level lookback windows, one for acquisition events and one for other key events. The GA4 attribution settings guide explains how those windows sit beside the reporting model and channel groups.
- Meta applies an attribution setting at the ad set level, combining a click window and an optional view window, and Meta's help documentation describes how the setting shapes both reporting and the people the system optimises towards.
- Affiliate programmes use a cookie window, a commercial term that decides whether a partner is paid. The guide to how affiliate tracking works treats it as a contract setting rather than an analytics one.
- App measurement partners apply one set of click and view windows to every network so that installs are judged by the same rules, which is why the app attribution guide on MMPs and SKAdNetwork treats inconsistent network defaults as a setup error.
How window length changes the numbers
A longer window credits more conversions to ads, because it lets in buyers who took their time and, with views, people who might have bought anyway. Reported conversions and ROASRevenue generated per unit of ad spend. therefore rise as the window lengthens, without a single extra sale. A shorter window does the reverse: it concentrates credit on interactions close to the purchase and undercounts products with long decision cycles. Neither is correct in the abstract. The right length is the one that matches how long real customers take to decide, which the business can read from its own time-to-conversion data.
The window also explains why recent results look weak and then recover. A report for the last few days keeps growing until the window has closed, because conversions are still arriving for clicks that already happened.
Aligning windows before comparing
When two systems both claim the same sale, a DeduplicationRemoving double counts when one conversion is reported by more than one source. rule decides which counts once. But before deduplication matters, the systems have to be measuring on comparable terms. The practical sequence is to put both on click-only windows of the same length, wait until that window has closed for the period being compared, and only then look at what gap remains. The guide to why Google Ads and GA4 numbers differ works through the other causes, from view-through credit to date logic, that account for what is left. Changing a window mid-period breaks trend lines, so any change belongs in a change log.
A furniture seller lines up its windows
Lindqvist Office Furniture sees its social platform report 140 sales in a month while GA4 credits that platform with 52. The platform is counting conversions up to 7 days after a click and 1 day after a view; GA4 uses a last-click model with a 30-day lookback. Re-run with the platform's columns set to 7-day click only, the platform figure falls to 81, because 59 of its sales came from people who had only seen an ad. The remaining gap of 29 sales is mostly journeys where a later search or email click took the credit in GA4. Neither number was wrong; they were answering different questions. The team now reports both on click-only windows and keeps view-through results in a separate column.
Related terms
AttributionAssigning conversion credit to touchpoints. is the rule that shares credit among the touchpoints a window admits, and the attribution models framework argues that the window should be chosen to match the sales cycle rather than left at a default. GA4Google's event-based analytics platform. keeps its own lookback windows that rarely match an ad platform's. DeduplicationRemoving double counts when one conversion is reported by more than one source. settles which system counts a sale that two of them claim, and ROASRevenue generated per unit of ad spend. is the figure that moves most visibly when a window is lengthened or shortened.
Common confusions
- The window is not the review period. Teams also say "lookback" for the span of data they examine before pausing an ad; the SOP for pausing underperforming ads uses it in that sense. That is a decision rule, not an attribution setting.
- A longer window is not a better one. It produces bigger numbers, not truer ones.
- View-through credit is not proof of influence. A person who saw an ad and bought may have been buying anyway; only a holdout test separates the two.
- Defaults are decisions. Leaving each platform at its own default means every source is judged under different rules.
- Window changes are rarely retroactive. Comparing periods either side of a change compares two settings, not two levels of performance.
