How a lookback window works
When a conversion is recorded, the system looks backwards in time for the touches that led to it. The lookback window is the boundary on that search. Anything inside it is eligible for credit; anything older is treated as though it never happened, however genuinely it contributed.
The window is set before the model runs, so it decides what the model is allowed to see. A short window hands the account to whatever happens near the decision. A long window lets earlier discovery touches into the calculation, which usually spreads credit towards awareness activity.
Two details cause most of the confusion. First, click and view windows are set separately, and the view window is normally much shorter, because seeing an ad is weaker evidence than clicking one. Second, a change to the window is applied going forward in some tools and retrospectively in others, so the same past month can be reported two different ways depending on where you read it. This is closely related to the attribution window used inside ad platforms.
Why the lookback window matters
It changes reported results without changing a single sale. Widen it and older touches start collecting credit, so discovery channels improve and the closing channels appear to weaken. Narrow it and the reverse happens. Nothing about the business has moved; only the boundary has.
It also decides which conversions the bidding systems learn from. A window shorter than your real buying cycle means the platform never connects the click to the sale it produced, so it optimises towards whatever converts quickly rather than towards what is worth most.
Common mistakes with lookback windows
The most damaging is leaving the default in place for a long, considered purchase. An education consultancy, a property developer or a business-to-business service can have months between first contact and enquiry, and a default window built for quicker decisions writes off most of the journey.
The second is changing the window and then comparing periods across the change. That comparison is meaningless, and I have seen budget moved because of it. The third is comparing tools without checking their windows at all; two reports built on different boundaries were never going to agree.
How to act on it
Measure your actual time from first touch to conversion for deals you win, then set the window a comfortable margin beyond it. If the data to do that is missing, the honest first step is fixing measurement rather than guessing a number, which is what a tracking audit exists to sort out.
Record the window alongside the model in whatever document your reports are built from, and note the date whenever either changes. When you do change it, keep the old setting running in a parallel view for a while so you can see how much of any movement is real. And accept the limit built into the idea: a window cannot recover a touch the browser never allowed you to store, so a very long window on badly retained data is false comfort rather than better measurement.
