The One-Hour Black Friday Claim Ratio Audit

The claim ratio audit takes one hour: pull orders for 24 to 30 November 2025 from the store, pull each platform's claimed conversions for the same seven days under the windows they actually used, sum the claims without deduplication, divide by orders. Then pull deduplicated attributed conversions from analytics and divide those by orders for the coverage rate. Two numbers, last year's data, no vendor, and both reframe the Q4 plan before a euro is allocated.

Octalysis's Rolling Rewards are small payoffs that arrive on a short cycle and keep the player going. This is one: an hour's work with a result nobody in the company has seen, and the result is usually a surprise in a specific direction.

The six steps

1. Orders. From the store or finance system, orders placed 24 to 30 November 2025. This is the fact about money. Call it O.
2. Meta. Purchases reported for the same dates under the attribution setting the account used at the time, which was probably 7-day click and 1-day view. Call it M.
3. Google. Conversions for the same dates for the purchase action, under Google Ads' window. Call it G.
4. Everything else. TikTok, Pinterest, affiliates, and the email and SMS platform's attributed orders under its own window. Call the sum E.
5. Claim ratio. (M + G + E) divided by O. Do not deduplicate; the point is to see the overlap.
6. Coverage. Purchase conversions your analytics attributes to any named source for the same dates, deduplicated (one per order), divided by O.

Then repeat both for a quiet week, say 9 to 15 February 2026. The difference between the peak and the quiet week is how much more generous the platforms get when every channel is on and every window is open.

An illustrative example

Say the store shipped 4,000 orders in Cyber Week. Meta claims 2,600 purchases, Google 1,500, TikTok 400, and the email platform 2,900 attributed orders. The claims sum to 7,400. The claim ratio is 7,400 divided by 4,000, which is 1.85: the suppliers collectively believe they did 85% more work than exists. Analytics, deduplicated, attributes 2,500 orders to a named source, so coverage is 0.625. Every cost-per-order figure computed from analytics is therefore the real figure divided by 0.625, which is 60% too high, before any platform overlap is considered.

Those numbers are made up to show the arithmetic. Yours are not, and they are an hour away.

Reading the two numbers

The Price of Being Found's guidance: the claim ratio's level matters once and its movement matters always, because movement is usually a definition change rather than a performance change. Meta narrowed its click definition in March 2026, per the book's account; if your ratio falls this November, that is a candidate explanation before any improvement is. Coverage is the first of the four properties of a defensible number: revenue of X, of which the systems can attribute Y, reported rather than hidden.

The two together tell you what the platform ROAS figures in the plan are made of. Your Black Friday ad budget is finite. The platforms' claims are not. is the allocation consequence; the number nobody checks is the coverage half in depth.

What the audit cannot tell you

It cannot say which channel holds the double counts, and it cannot say what any channel caused. A claim ratio of 1.85 means the credit is inflated; it does not say whose. That question needs a design that could have found nothing: a holdout that starts by 2 October, or a causal read on the GA4 export that estimates each channel's incremental contribution with an interval. The audit is what makes the case for either.

What to do this week

• If you own the budget: run the six steps this week. Put both numbers on the first page of the Q4 plan, above any ROAS.
• If you have to defend the number: run them for three periods, Cyber Week 2025, a quiet week, and last month, so the review sees a series rather than a snapshot.

The calendar has the dates that follow.

As of 9 September 2026. The claim ratio, the coverage identity and the four properties are from The Price of Being Found (Edition 2.10), Chapters 8, 9 and 19. The worked example is illustrative.