Main idea

Two verbatims can talk about the same object and have nothing to do with each other. "The export button is hard to find" and "I want to steer all my sites from one place" both mention reporting; one is a friction on a screen, the other an organizational objective. Filing them under the same "reporting" label makes them incomparable, because they are not handled at the same moment, nor by the same people.

The classification criterion is therefore not what the customer is talking about, but the height at which they are talking about it. Four questions are enough to settle it, and they bear on the customer's phrasing, not on the interpretation you put on it: are they talking about a business objective or an organizational constraint, about what they are trying to accomplish in their work, about what the product should make possible, or about a specific screen and behavior?

Classifying by apparent topic is the natural reflex because the topic is what you read first. The level requires looking at the shape of the sentence: a verb of business intent, a work action, an expected capability, a situated irritation.

Why it matters

This makes the decision fast and repeatable. A taxonomy by topic forces you to ask, for every piece of feedback, which theme it goes in — a question with no right answer, which plays out differently depending on who is filing. A taxonomy by level is settled in seconds and gives the same result across two people.

Above all it prevents a flattening: the strategic piece of feedback drowned in a feature label, which never comes back up at the moment product strategy is being discussed.

Nuances and limits

The same verbatim sometimes contains several levels — a customer explains their stake, then illustrates it with an irritation. The level retained is that of what is being asked for, not that of the context given to explain it; genuinely double cases are split into two entries rather than forcing the choice.

The rule doesn't do away with a topic axis: it says which plane to place the feedback on, not what it talks about. The two combine, they don't replace each other.

Open questions

• What do you do with a piece of feedback whose level changes between the interview and rereading it six months later?