There is a precise moment, near the end of an engagement, when we mistake the document for the result. The file is heavy, the structure is clean, the last proofread is done. We think we have finished. We have only finished writing.
A deliverable is a trace. It attests that a piece of reasoning took place, it allows it to be replayed, it protects an organisation's memory against its own staff turnover. That is useful. It is not the work. The work is the movement obtained: a decision taken earlier than it would have been, a trade-off made explicit, a team that stops waiting for a permission that will never come.
The confusion is comfortable for everyone. For the consultant, because a document is finished, dated, billable. For the client, because a document can be filed away. A decision, on the other hand, is paid for in internal conflict, in abandoned priorities, in people who have to be convinced twice.
Nobody has ever changed their mind reading page 47.
01 — What is left six months later
Run the exercise on your last three engagements. Six months on, what remains? Almost never the report. What remains falls into three categories: a shared vocabulary, a ritual, and a person. The vocabulary, because an organisation cannot deal with a problem it has not named. The ritual, because a decision that is not in the calendar does not exist. The person, because every change needs an owner with no way out but to succeed.
If your engagement produces none of those three, it produces a memory. That is an honourable trade, but it is not the one you were bought for.
02 — Against exhaustiveness
Exhaustiveness is a polite form of cowardice. To say everything is to refuse to prioritise; and refusing to prioritise means handing the client the hardest part of the work — precisely the part they are paying not to do alone. An eighty-page document rarely contains eighty pages of ideas; it contains two ideas and seventy-eight pages of insurance.
The useful constraint is brutal: one page. If the thesis does not fit on a single page, it is not a thesis yet. The rest becomes an appendix, which is exactly what it is: evidence, available to whoever asks for it.
03 — The corridor test
Before the readout, stop someone in a corridor and tell them the conclusion in forty seconds. If they interrupt you with a precise question, you have something. If they nod politely, you have written a summary of the situation, not a recommendation. The difference between the two is measured in risk: a recommendation can be wrong.
That is where the value of the craft is decided. We are not paid to describe an organisation to itself — it knows itself better than we do, and has done for longer. We are paid to carry, for a few weeks, the political cost of a sentence that nobody on the inside can say without leaving their position behind.