Does Journey Orchestration Actually Increase Revenue?

A journey orchestration platform can tell you how many conversions its orchestrated journeys touched. It cannot tell you how many of those conversions would have happened anyway, because the customers who were routed and the customers who were not are not comparable unless somebody deliberately made them so. The revenue question has a real answer. It is just not available from inside the platform that is asking to be paid for the revenue.

What the dashboard is reporting

Orchestration decides, in real time, which message or path a customer meets. The reporting then counts outcomes among customers who met those paths. Everyone who was going to buy anyway and happened to pass through an orchestrated journey is counted as an orchestrated conversion, because from the system's point of view they are indistinguishable from someone the orchestration persuaded.

This is the same structure as an ad platform reporting on its own ads, and it produces the same inflation for the same reason. The system that chooses who receives the treatment is also the system that reports the result.

The selection problem, stated plainly

Orchestration targets. That is the product. It routes the engaged customer to one path and the dormant one to another, and it does so using signals that predict purchase. So the orchestrated group is, by construction, the group more likely to buy before any orchestration happened.

The better the targeting, the larger the apparent effect and the smaller the real one, because a model that is excellent at finding people about to buy will produce spectacular reported numbers while adding very little. The experiments The Price of Being Found quotes make this concrete: in the Facebook studies it cites, observational methods overstated effects by roughly three times against randomised ground truth, and in one set of fourteen comparisons six were not statistically distinguishable from zero at all.

The design that answers the question

Randomly withhold. Take a slice of customers, chosen at random rather than by any behavioural rule, keep them out of the orchestrated experience for a fixed window decided in advance, and compare revenue per customer between the groups.

Three details decide whether the result means anything. Fix the window before you start, because stopping when the gap looks good manufactures the gap. Choose the slice at random, because any behavioural rule reintroduces exactly the selection you were trying to remove. And work out the smallest effect the design could detect at your volume before you run it, so you know in advance whether a null result would mean anything. Incrementality testing for ecommerce covers the arithmetic, and the measurability floor covers when a test cannot clear it.

Most brands can afford one such test per quarter, not one per feature.

What to do between tests

A causal read on your own export estimates what each channel contributed using the variation already in the data, reports an interval, and states its coverage. It is observational rather than experimental and it says so, which is the property that makes it usable between anchors. Incremental ROAS per channel without a geo test covers the limits honestly.

The question to put to the vendor

What would your system report if the orchestration had no effect at all, and have you tested that? An honest answer describes a placebo design and its result. A vendor who has never asked the question of their own product is not withholding the answer; they do not have one. Ask your vendor for a placebo test covers the exchange, and it applies to this site's publisher as much as to anyone else.

What to do this week

• If you own the budget: ask your orchestration vendor for the holdout design behind the revenue number in their QBR. The answer tells you what you bought.
• If you have to defend the number: label orchestration revenue as touched rather than caused, until a holdout says otherwise.

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

As of 9 September 2026. The Facebook experiment comparisons and the measurability arithmetic are from The Price of Being Found (Edition 2.10), Chapters 12 and 19, with the book's stated caveats.