Your ad platform stops watching on day 7. Your customers don't stop shopping.
This one started as a routine question on the same account I wrote about in the cross-domain piece: a Shopify client with a $300+ average order value. The question was simple. How long does it take a buyer to get from ad interaction to purchase? GA4's answer: ad-involved purchase journeys averaged around 7 days and around 3 touchpoints.
Then I checked what the ad platforms were set to watch. Google Ads and Meta were both running 7-day click attribution windows. Both platforms stop counting at almost exactly the moment the average buyer in this account decides.
Nobody planned that collision, and that's the point. Nobody plans it. The attribution window gets set once, during account setup, usually by whoever built the first conversion action. Your customers' consideration time is a behavior that belongs to them, not to your settings. The two numbers live in different places, get owned by different decisions, and in most accounts have never once been compared.
Where the 7 days comes from
On Meta, 7 days is not just the default, it's the ceiling. The click-through options are 7 days or 1 day, and the old 28-day window is gone. If your buyers take longer than a week, Meta cannot be configured to watch them.
Google is more generous on paper. The default window is 30 days, adjustable from 1 to 90, and Google's own documentation tells you to check where your conversions actually land before choosing. In practice, accounts frequently get set shorter: someone tightens the window years ago to "clean up the data," the decision outlives its author, and the setting never gets revisited. A window is a setting. Lag is a behavior. Settings persist while behaviors drift, and nothing in either platform warns you when they diverge.
The cliff in the lag data
You can see the collision directly if you know where to look. On this account, I pulled conversion counts by lag bucket through the Google Ads API and looked at the shape of the curve as it approached the window edge. Conversion density was still healthy in the days approaching the cutoff. Then, at day 7, the count dropped to exactly zero.
A natural decay curve tapers. A truncated one falls off a cliff.
A parallel conversion action on the same account with a 90-day window showed what the cliff was hiding: roughly 5 to 10 percent of conversions land beyond day 7. Those sales were never missing. The store rang them up, the revenue was real. They were only missing from ad reporting, which means they were missing from ROAS, from cost-per-acquisition, and from every decision built on those numbers.
The part nobody talks about: your bidding trains on what it can see
Underreporting is the visible cost. The expensive cost is quieter. Smart bidding learns from the conversions the platform records, which means a truncated window isn't just a reporting gap, it's a training-data gap. The algorithm gets a complete education on your fast closers and never hears about your deliberate ones.
Here's what made that concrete on this account: the beyond-window conversions carried roughly 35% higher average order value than the account-wide average. That pattern makes sense for a considered purchase. The buyer who takes nine days is comparing, researching, coming back. Deliberation and order value travel together. And those are precisely the buyers the window edits out of the algorithm's world.
You are not just undercounting sales. You are teaching the algorithm to ignore your best customers. Every optimization cycle, the bidding leans a little further toward the impulse-shaped buyer it can see and away from the high-value buyer it can't.
How to know if this is you
Two checks take about ten minutes and require nothing but access to your own accounts. In GA4, open Advertising and look at the attribution path reports: average days to conversion and touchpoints to conversion are sitting right there. In Google Ads, open your conversion action's settings and read the actual window on it. Don't assume the default; assume someone changed it, and verify.
If your measured lag is approaching your window length, you're in collision territory. The risk profile is exactly what it sounds like: considered purchases, higher order values, multi-touch journeys. If you sell something people buy in one sitting, this probably isn't your problem. If your buyers take a week, it probably is.
The deeper read, the shape of the lag curve at the window edge, takes API access and is the difference between suspecting truncation and seeing it. A curve still carrying weight at the cutoff is being truncated, full stop.
Sizing the window to the buyer
The fix is not "make every window 90 days." Longer windows trade recency of signal for completeness, and the right answer depends on the shape of your actual lag curve, which is why the diagnostic comes before the change. A full pass covers the lag-curve shape at the window edge, window sizing against measured lag rather than platform defaults, a parallel long-window action to expose the true tail, and a reporting cadence that stops anyone from judging spend before its conversions have had time to arrive.
This is the same category of problem as the cross-domain leak I wrote about: configuration quietly misaligned with reality, invisible on every dashboard, compounding monthly. Checking it costs half an hour. Running misconfigured costs you every slow, deliberate, high-value buyer your account attracts. Measurement is where I start every analytics engagement, because spend decisions are only as good as the data underneath them. If you don't know what your window is set to right now, that's your answer: get in touch.