How to Measure a Journey You Cannot Fully See

Customer journey measurement has a first step that most journey reports skip: work out what fraction of your real orders the data can resolve at all. That fraction is your coverage rate, it is one query against two systems you already have, and until you know it, every journey percentage you read is a share of an unknown share of your business.

The one query

Take a fixed date range. Count the conversions your analytics attributed to any source at all. Count the orders your store actually shipped over the same dates. Divide the first by the second.

That is coverage. If it comes back at 0.6, then 40% of your orders left no resolvable trace in the journey data, and every cost-per-order figure you quote is the real figure divided by 0.6, which is to say 67% above reality before a single media price changes.

The Price of Being Found reports its publisher's own census at 48.6% of sessions with no resolvable source over 17 October 2025 to 2 September 2026. That is one company's data and the book says so rather than presenting it as an industry figure. Your number is yours, and it is an hour of work. The number nobody checks walks the query.

Why journeys go missing

Consent declines remove the tracking that would have resolved the session. In-app browsers, from social and messaging apps, break the handoff. Cross-device journeys start on a phone and finish on a laptop, and nothing joins them without a login. Assistants and answer engines increasingly send buyers with no resolvable referrer at all, which lands in Direct alongside everything else that could not be identified. When AI sends the shopper, who gets the credit covers that last one.

None of these is a tagging bug you can fix your way out of. Better tagging raises coverage. It does not close the gap.

The three systems, and which one is money

Your ad platforms report conversions they can associate with their own inventory, inside their own attribution windows, under their own rules. Your analytics reports sessions it could resolve. Your store reports orders. The three disagree by construction, and the sum of platform claims routinely exceeds the order count, which is arithmetically impossible as a description of distinct sales.

Only one of the three is money that arrived. Anchor every journey metric to it. The claim ratio audit is the hour that makes the disagreement visible.

Journey metrics that survive the coverage question

Once you know the fraction, some metrics still mean something and some do not.

Survives: anything computed as a ratio within the visible set and compared against itself over time, such as conversion rate by resolved source, week over week. The coverage bias is roughly constant, so the trend is readable even when the level is not.

Does not survive: any absolute count or revenue share presented as a share of the business, and any comparison between two periods where consent rates or traffic mix moved, because the denominator moved underneath the comparison.

Needs a different design entirely: any claim that a touchpoint caused revenue. That is a counterfactual and no amount of journey data supplies it. Causal inference for marketing teams covers what does.

What to do this week

• If you own the budget: compute coverage for the last 40 days, write it down with the date, and put it on the first slide of the next review. It reframes every number after it.
• If you have to defend the number: quote journey figures as shares of the visible set, in those words. It costs nothing and it makes the report honest.

A causal read states its coverage rather than presenting a total that may be half visible. The interactive demo shows the output on a sample store, no signup.

The coverage census (5,858 of 12,054 sessions) is from The Price of Being Found (Edition 2.10), Chapter 8, rated Supported rather than Established by the book's own appendix because the underlying data is not independently reviewable.