Originally written in French. Translated by AI — the meaning has been preserved, not the prose.
Main idea
When a structuring decision is written down nowhere, it gets taken again. The obvious complaint is the wasted time: the same people redo the same discussion, sometimes with others, often under commercial pressure, always with a partial memory of what had already been settled.
That cost is visible and secondary. The real damage is that each retaking produces a slightly different answer, and that these answers coexist. On a marketplace dispute, support leans toward preserving the customer relationship, the partnerships team toward protecting suppliers, product toward not complicating the rules, sales toward the option that serves the account of the moment. None of these readings is unreasonable; they are simply incompatible, and they apply in parallel across dispute policies, support messages, product priorities and customer communication.
The company therefore ends up not with a delayed decision, but with several simultaneous decisions that nothing forces to meet. A written record does not save time first: it makes a single answer apply everywhere.
Why it matters
This changes the argument with which you defend writing a decision down. Costing out the meeting hours avoided is weak and contestable; showing that four functions are today applying four different rules to the same conflict is verifiable and hard to brush aside.
It also indicates where to look for the evidence of the problem: not in the calendars, but in the gaps between what support answers, what sales promise and what product builds.
Nuances and limits
Inconsistency is not always costly: on subjects with little visibility or rarely compared with one another, letting each function decide as it sees fit costs less than establishing and maintaining a common rule.
And a written rule does not guarantee coherence: if it is neither known nor applied, the divergent readings continue, with a document claiming the opposite on top.
Open questions
- How can the gap in treatment between functions on the same type of case be measured, without a heavy survey of each of them?