The Journey Map KPIs That Survive an Audit

A customer journey map usually carries one KPI per stage, and the map rarely says which of those numbers describes the customer and which describes how well your tracking resolved that stage. The two behave differently. Behaviour metrics move when customers change. Resolution metrics move when consent rates, traffic mix or tagging change, which is to say they move for reasons that have nothing to do with the journey being mapped.

Sort the KPIs into three piles first

Pile one: it happened, and the store knows. Orders, revenue, average order value, refund rate, repeat purchase rate, time between first and second order. These come from the commerce platform, not from tracking resolution, so they are as complete as your business is. Put these on the map with confidence.

Pile two: a ratio inside the visible set. Conversion rate by resolved source, add-to-cart rate, session-to-order rate. These are computed over the fraction of journeys your analytics could resolve. They are useful compared against themselves over time, because the resolution bias is roughly constant week to week, and they are not shares of your business.

Pile three: credit under a rule. Attributed revenue per stage or per channel. This is an allocation, not a measurement, and it changes if someone changes the model. Keep it if you like, and name the rule beside it.

The failure mode on most maps is that all three piles are printed in the same typeface, so a number from pile three gets discussed as if it came from pile one.

The metric that decides how to read the rest

Coverage: attributed conversions divided by the orders the store shipped over the same dates. If it is 0.6, every pile-two number describes 60% of your business and every cost figure is inflated by the reciprocal. Compute it once, write it down with the date, and put it in the map's header rather than in a footnote. How to measure a journey you cannot fully see has the query.

Two KPIs worth adding that maps usually lack

The claim ratio. Sum what each platform says it caused, divide by orders. Above one, your stages are being credited more times than sales occurred, and the map is displaying an arithmetic impossibility. The one-hour audit is the procedure.

The not-measurable list. Every map has stages whose effect is too small to detect at your spend. Naming them is more useful than assigning them a number, because it stops the map implying precision that the data cannot support. Whether your channels are measurable at all has the arithmetic.

What a stage KPI cannot do

It cannot tell you that the stage caused the outcome. A stage appears in the journeys of customers who converted because those are the journeys you kept. The customers who saw the same stage and did not convert are thinner in the data, and the customers who would have converted without the stage are not distinguishable at all. That is a counterfactual question, and it needs a design built for it rather than a better dashboard.

For the biggest stage, a holdout answers it. For the rest, a causal read on the export estimates it with an interval and says which stages fall below the floor. Incrementality testing for ecommerce is the playbook.

What to do this week

• If you have to defend the number: relabel your existing map into the three piles. Nothing new to collect, and the conversation changes immediately.
• If you own the budget: add coverage and the claim ratio to the header. Two numbers, one hour, and every other number on the map becomes readable.

The interactive demo shows a causal read on a sample store, no signup.

Product facts as stated on causalityengine.ai on 9 September 2026. Coverage, claim-ratio and measurability material is from The Price of Being Found (Edition 2.10), Chapters 8, 9 and 19, with the book's caveats.