The roadmap and the real workflow have drifted apart. This finds exactly where — so the backlog reflects what users actually do.
Every product implies a journey: the sequence it was designed around, and the order its screens assume. Customers work out their own way of using it, which drifts from that over time. Backlogs are usually built from the implied journey, so as the drift grows, more of the roadmap addresses steps that matter less than they used to.
When nobody has looked at the sequence people actually follow, everyone's version of the customer is partly invented, and each version is internally consistent. The argument can't resolve, because there is nothing outside the room to check it against.
A journey has stages, and the break is not always at the stage people are looking at. A team can spend a long time making a problem easier to fix while the real break is that customers cannot tell the problem is happening. When the break sits earlier than assumed, every improvement downstream of it has less effect than expected.
Two journeys, mapped side by side, and a backlog that reflects the gap between them:
Two maps of the same thing. One is drawn from the product — what it assumes people do, and in what order. The other is drawn from asking people to walk through what they actually did most recently, in their own words. The value is in the comparison, so both have to be drawn properly.
Prerequisitewhatever documents the intended journey — flows, specs, onboarding material. The implied journey has to be read off something.
Not every difference matters. Scoring the gaps by how often the step happens, and by what it costs when it goes wrong, separates the ones worth fixing from the ones worth knowing about.
Prerequisiteusage data showing how often each step occurs, so the gaps can be weighted rather than only listed.
An ordered list of work with the evidence for each item attached. Attaching the evidence is what keeps the order from being renegotiated on the next strong opinion.
Prerequisiteaccess to the current backlog, so the output merges into it rather than sitting beside it.
