Attribution windows explained: why two reports disagree

Aug 22
3 min read

Updated: 2 days ago

Introduction

Your ad platform says it produced forty conversions. Your own records show twenty-two customers. Both numbers are being reported honestly.

A large part of that gap is the attribution window — the period after seeing or clicking an ad within which a sale still gets credited to it. Change the window and the same month produces a different result.

1. Attribution windows explained: a lookback period, chosen by someone

An attribution window is a rule: if a customer buys within this many days of interacting with an ad, the ad gets the credit.

Every platform has one, every platform has a default, and the defaults differ. That alone explains a good share of the disagreement between reports.

Crucially, the window is a setting rather than a measurement. Nobody discovers what caused a sale; somebody decides how long the credit lasts.

2. Click windows and view windows are very different claims

Most platforms distinguish between someone clicking an ad and someone merely seeing it.

A click window credits sales from people who actively engaged. A view-through window credits sales from people who saw the ad and did not click, then bought later.

View-through attribution inflates results substantially and is the most common reason a platform's figures exceed reality. If a channel's performance looks implausibly good, check whether view-through conversions are included before concluding anything.

3. Longer windows mean more credit and less certainty

The trade-off is straightforward. A seven-day window reports fewer conversions with a stronger causal link. A ninety-day window reports many more with a much weaker one.

Over ninety days, a customer may have seen your ad once, then searched for you, read a review, asked a friend and returned twice. Crediting the whole sale to the ad is a choice, not a finding.

Neither is wrong. What is wrong is comparing a channel measured over ninety days against one measured over seven and treating the result as a like-for-like ranking.

4. Match the window to your actual sales cycle

The window should reflect how long your customers genuinely take to decide.

For an impulse purchase, a few days. For a considered service costing a month's income, several weeks. For business software with a committee involved, longer again.

You can measure this: for a sample of customers, look at the time between first contact and purchase. Use the point by which most have bought, rather than the slowest case.

5. Set the same window everywhere before comparing channels

This is the practical fix and it costs nothing.

Go into each platform and set the same click window, and either include view-through everywhere or nowhere. Then rebuild the comparison.

Channel rankings frequently change after this, and decisions made on the previous rankings were decisions about reporting settings rather than about performance.

6. Expect double counting across platforms

Two platforms will both claim the same sale, because each saw its own interaction and neither knows about the other.

Add up the conversions reported by every channel and the total usually exceeds your actual customer count, sometimes considerably. That is not fraud; it is each platform reporting within its own window.

Which is why the total that matters comes from your own records — invoices, bookings, sales — and the platform figures are only useful for comparing channels against each other under identical settings.

7. Keep asking customers directly

The cheapest attribution data available does not come from a platform.

Asking every enquiry where they heard about you, and recording the answer, gives an independent reading. It is imperfect — people misremember and name the last thing they saw — and it is not subject to any window at all.

Where the self-reported picture and the platform picture disagree sharply, that disagreement is informative and worth investigating rather than resolving in favour of whichever you prefer.

8. Test with a holdout rather than trusting the window

The only strong evidence about a channel's contribution comes from turning it off.

Pause it for a defined period and watch total enquiries and sales, not the channel's own reported numbers. If the total does not fall, the reported conversions were largely people who would have bought anyway.

That test cuts through every attribution argument, and it is the reason to keep a record of total demand alongside channel-level reporting.

Conclusion

An attribution window is a setting that decides how long an ad keeps claiming credit, and defaults differ by platform.

Separate click from view-through, understand that longer windows buy volume at the cost of certainty, derive your window from measured decision time, apply identical settings before comparing channels, expect double counting, keep asking customers directly, and settle real disputes with a pause test against total demand.