Originally written in French. Translated by AI — the meaning has been preserved, not the prose.
Angle
A backlog is not defined by what it contains but by the threshold it materializes: a subject enters it when it stops being an individual thought and requires a developer, a designer, a data analyst or support. Placing that threshold at the moment of capture — "I'm noting it down so I don't forget" — is the single cause of the disorders later blamed on the size of the queue: illegibility taken for transparency, customer requests turned into tasks, interviews flattened into cards, tickets going stale with nobody daring to close them. A tool for action holds its items for a short time and in small numbers; everything that has to last — signals, decisions, reasons for refusal, product knowledge — belongs upstream, and it is there, not in the tracking tool, that the product is thought through.
Synthesis
The starting point is an entry rule, and it speaks neither of maturity nor of value. It only asks: whom do I need? As long as a Product Manager is digging alone, their workspace can be Obsidian, Apple Notes, a notebook — it makes no difference, since nobody else comes in. As soon as a feasibility has to be studied, an interface designed, a dataset explored, the collective tool becomes useful.
Everything else follows from where that threshold sits. When it is placed at the moment the idea appears, the backlog becomes the default receptacle for everything crossing the organization. The argument invoked is transparency — and it turns against itself: a queue of five hundred lines does not expose decisions, it exposes the absence of decisions. What an executive or a salesperson is looking for is not the list, it is the answer to "why not this subject".
A badly placed threshold also explains the two symmetrical excesses in what gets written. Too little: three lines dropped in transfer a vague thought to someone who will have to reconstitute it without having seen the original signals. Too much: templates everywhere, fields, mini-dossiers written to cover oneself. No quantity settles between the two, because the right quantity depends on the context already shared — and the artifact was in any case never made to replace the conversation it triggers.
Once the subject is in, two forces push to leave it there. The team's capacity is finite, so adding ought to be traded against a removal; and a tightened plan backed onto an unbounded queue has restricted nothing at all. But nothing is removed, because a written card looks like capital. The entry cost already paid passes itself off as future value — even though the world around the card has moved on and it describes a vanished context.
What remains is what you think you are protecting by keeping everything. A customer request is valuable as learning material, and what is worth keeping in it is the situation described, not the solution proposed; a refusal is valuable through its reason, which can be recorded without keeping any work to be done. In both cases the information is saved upstream, and nothing justifies housing it in the tool for action.
Upstream, moreover, has stopped being a vague intention. Sources, transcribed interviews, notes, glossaries, decisions, product data, signals linked to existing and absent features: this layer can be questioned, cross-referenced, recomposed. A card tool can do none of that — not for lack of features, but because a list knows only items and their order, never their relations. Hence the tracking tool's final place: an output chain into which you push an already structured thought, and whose items, serving up to delivery, are written for that lifespan and not for the archive.
Tensions / contradictions
Holding the memory upstream presupposes that somebody holds it. The border shifts a load as much as it removes one: the collective space gets lighter, a knowledge space appears, and nothing says it will be better maintained than the queue it replaces.
A second tension, about the personal space. Its freedom of tooling is presented as an advantage, and at the same time it makes everything under investigation invisible to the team — which is exactly the reproach levelled at keeping immature ideas out of the backlog.
Questions
- Can an outside observer establish where an organization has placed this border, without questioning those who drop the subjects in?
- What arrangement keeps the upstream space from becoming an inventory in its turn, since no entry rule protects it?