Originally written in French. Translated by AI — the meaning has been preserved, not the prose.
Angle
What a company knows about its customers is worth nothing until it has been through five passages — captured, structured, connected, shared, fed back — and any of them can be missing without anything in the organization flagging it: a team that has interview transcripts, a clean filing plan and a synthesis circulated to everyone has cleared none of the last three. It is that chain, and not the material, that constitutes the advantage: private conversations, objections and adoption frictions exist at the competitor's too, who lets them scatter. The price of a broken loop, moreover, is paid not in delay but in the quality of the next arbitration, which makes it invisible — a decision taken on an impoverished version of the context looks exactly like a good decision. And the loop does not stop at the company: knowledge accumulated about a trade can go back out to the customer in the form of a capability, which closes the circuit instead of merely restarting it.
Synthesis
Taken one by one, these notes look like recommendations for keeping good documentation. Placed end to end, they describe a chain in which each link has its own mode of failure, and it is that catalogue of failures that constitutes the level of reading.
The first link breaks through insufficient transformation: filing a verbatim quote in a "customer feedback" folder gives exactly the same impression of order as requalifying it into a signal, then a situation, then a business stake, whereas the first gesture added nothing to what the organization understands. The second breaks through impoverishment at the moment of sharing: circulating a synthesis while withholding access to the raw material is hard to distinguish, when it is sent, from circulating it with access left open — the two diverge only at the moment someone wants to reread it differently, and by then it is too late. The third breaks through silence: an impeccable base nobody opens at the prioritization committee does not announce itself as unused, since the decision gets taken anyway.
What links these three failures is their shared form, and it explains why the problem persists in organizations that document a great deal: none of them produces a symptom. An organization that starts over from scratch every six months notices it nowhere.
The rest of the network says what the closed loop makes possible, and that is what justifies maintaining it. At the input, bringing the prisms together — the problem the Product Manager sees, the perception the Product Marketing sees, the objection the sales rep sees, the friction the CSM sees — brings out the blind spots no single position can see alone. At the output, that diversity becomes the condition of an arbitration backed by independent paths rather than by one person's memory. And in between, the same unit of knowledge produces five different deliverables depending on which function picks it up, which settles the question of cost: the loop is not amortized on the decision that triggered it.
There remains what sets it apart from a mere document system. A warehouse keeps the first level — the calls, the tickets, the verbatim quotes. The loop creates the passage between the three levels, all the way to the business uses: PRD, roadmap, positioning, sales narrative, onboarding. That passage is not a property of the tool, it is work somebody does, and that is why it cannot be copied.
Tensions / contradictions
Two stages pull in opposite directions. Structuring requires requalifying the material — the request becomes a signal, the signal a situation, the situation a business stake — and each passage discards something; sharing, on the contrary, requires keeping the raw material reachable so it can be reread differently. The loop therefore demands both that you transform and that you not lose what the transformation leaves behind, and nothing here says who pays for maintaining both states.
Second tension, more awkward because it sets a stage against its own purpose: the strength of a body of evidence rests on the independence of its paths, yet a well-shared common memory is precisely what circulates interpretations between functions. Sharing, which is the fourth passage, erodes what the fifth draws on — the more the readings converge, the less their convergence proves anything.
Finally, the claimed advantage rests on a practice being uncopiable. A practice can be observed, told at a conference and documented; what resists is not its principle, it is sustaining it over time — a more fragile answer than one based on non-public material.
Questions
- What, in an organization, could make it visible that one of the five passages has been skipped, given that no break produces a symptom at the moment it occurs?
- How much of this loop rests on the Product Manager's role, and how much must be carried by a mechanism their departure from the company would not break?
- Up to what size of organization does a single customer memory hold, before the number of functions and segments forces it to be split?