Originally written in French. Translated by AI — the meaning has been preserved, not the prose.
Angle
Business stake, use case, product capability, local friction point are not four categories on one and the same plane: they are four altitudes, and the altitude of a verbatim is set by the one at which the customer spoke, not by the subject they were speaking about. This scale does not serve to file more finely — it serves to route: each rung is the only entry point of a distinct product decision, strategy at the top, discovery and roadmap in the middle, design and backlog at the bottom. A repository that flattens the four rungs onto a single axis has therefore not merely lost detail, it has cut the supply to three arbitrations out of four — which will go on being made, without the signals that were meant to feed them.
Synthesis
The problem presents itself as a filing problem: you centralize feedback in Productboard or Harvestr, and a taxonomy is needed. Posed that way, it has no solution — each candidate axis, the feature, the stake, the theme, works for part of the material and breaks on the rest.
The shift that unblocks it is to stop looking for the right axis. What separates "the export button is hard to find" from "absorbing growth without hiring" is not a different subject, it is a different height; both may speak of the same functional domain. Once height is recognized as the real dimension of the problem, classification stops being an interpretation and becomes a reading of the customer's wording: four questions, a few seconds.
But it is what follows that makes the scale valuable, and it is visible in no rung taken in isolation. Each altitude has a consumer and a moment: the business stake is what you need in order to arbitrate a segment, the work action is what feeds discovery, the capability is what a roadmap is discussed on, the situated irritant is what a designer can fix the following week. The four do not go in the same drawer because they do not travel to the same people.
Hence the real effect of a single axis, which is not the one you would expect. You imagine a deficit of granularity — you could have filed better. The deficit is elsewhere: three levels of decision find themselves without material. Nothing complains, because the decisions get made anyway, with whatever is available. A roadmap built on a repository flattened to feature level looks perfectly documented.
What remains is the constraint that keeps the reasoning from being generalized indefinitely. A scale is paid for at every entry, by the person classifying a verbatim after an interview. Adding rungs because they are conceptually justified produces a sound model that nobody fills in — and an empty rung feeds no decision, exactly like an absent one. Four is a compromise, not a truth; the criterion that set it is that beyond it, the scale stops being applied.
Tensions / contradictions
Two requirements clash without being settled here. Routing towards decisions pushes to add rungs — every badly fed decision calls for its own. Adoption pushes to remove them, because the cost is paid at every entry, by people whose job is not classifying. The figure of four is one context's point of balance, not a result.
A second tension, inside the scale itself: the use case rung is valuable for its independence from any solution, and it is flanked by two rungs — capability and friction — whose vocabulary is that of the product. A verbatim dropped one rung too fast loses that independence with nothing to signal it.
Questions
- Do the rungs this scale distinguishes hold up when the repository is shared between several product teams that do not arbitrate in the same place?
- Can feedback classified at one altitude be raised a rung later without going back to the customer?