Interests, Not Positions

“We cannot go below 95 percent fill rate” is a position.

The interest behind it might be that the plant manager was blamed for a stockout two years ago and has no intention of being blamed again. It might be that a contract carries a service-level penalty nobody has read recently. It might be that 95 was in a board deck and walking it back requires a conversation they do not want to have.

These are three different interests, and they have three different resolutions. None of them is addressed by demonstrating that 94 percent is optimal.

Positions feel like requirements

To the person holding one, and frequently to everyone else in the room, a position is indistinguishable from a constraint. This is why analyses get built inside boxes that need not exist.

The cost is not merely a suboptimal answer. It is that the real disagreement never surfaces, so the meeting argues about your model while the actual objection sits unexamined.

Asking the question

The question is some version of: what would go wrong for you if this number moved?

Asked neutrally, in private, and before the meeting, it works far more often than technical people expect. People will generally tell you what they are protecting, because most positions are not strategic. They are compressed history, and nobody has asked them to decompress it.

Asked in the room, in front of colleagues, the same question is an attack and will be answered defensively. The setting does most of the work here.

When the position is real

Sometimes it is a genuine constraint: a contract, a regulation, a physical limit. Then you plan around it, and you have lost nothing by checking.

The check is cheap and the alternative is expensive, which is the entire argument for doing it every time rather than only when you suspect something.

When interests genuinely conflict

Sometimes they do. The parts organization is measured on holding cost and the service organization on stranded jobs, and no amount of clarification dissolves the tension because the tension is real and structural.

The honest move is to name it rather than to arbitrate it. Put the trade in front of the person who owns both scoreboards, with the exchange rate stated explicitly and the flip point marked. That person exists, usually one level up, and the decision is theirs rather than yours.

Analysts who try to resolve a structural conflict by finding a cleverer model are solving the wrong problem and will be worn down by it.

Being right is not enough

This is the sentence technical people most resist, so it is worth being precise about what it claims. It does not claim that merit is irrelevant, or that persuasion substitutes for correctness. It claims that correctness is one term in a product, and a product with a zero in it is zero. That is arithmetic rather than cynicism.

The useful version names the other terms, because technical people tend to pour everything into one of them and then find themselves puzzled that it did not work.

Competence is whether you appear to know what you are talking about. This is the term analytical training supplies in abundance, and it is almost always already sufficient by the time you are invited to the meeting. Adding to it has diminishing returns immediately. Nobody is declining your recommendation because they doubt you can compute a holding cost.

Reliability is whether the thing you said would arrive on Thursday arrived on Thursday, whether last month’s number survived contact with the closed books, and whether you flagged the problem yourself before somebody else found it. It accumulates only by repetition, which is exactly why it is undervalued by people who think of influence as something that happens during presentations. A recommendation from someone who has been right about small things for two years is heard differently from the same recommendation by a stranger with a better model, and that is a reasonable inference on the audience’s part rather than an irrational one.

Safety is whether people feel able to tell you something awkward. This term is load-bearing for everything else on this page, because the interest underneath a position is usually slightly embarrassing: a manager who does not want another Saturday call, a director who was burned publicly the last time service slipped, a function that does not trust its own data. Nobody says any of that to someone who might use it. If nobody has ever told you something off the record, you are not being told the interests, and your model of the decision is the one a stranger would build.

Self-orientation is the fourth term and it works against you rather than for you. It is how much of your attention is visibly on yourself: on your model, your framework, your cleverness, whether you are seen to be right. Whatever competence, reliability and safety you have accumulated is divided by it.

That structure explains something otherwise puzzling, which is how a brilliant and dependable analyst gets quietly excluded from the decisions that matter. It also says where the remedy is. The first three terms take years. The divisor you can change in the next meeting, by being the person who says the recommendation is close, or that a colleague’s objection is the right one, or that the number they are asking about is the one you are least sure of.

Contempt is the specific failure to guard against

Technical people fall into it by temperament, and it is the most efficient way to destroy the other three terms at once.

Telling a room that 98 percent is arbitrary wins the argument and loses the meeting. Telling them that 98 is standing in for stranded jobs, and that here is a rule aimed at the jobs instead, is also a fight, but it is a fight about something both of you want.

The same logic is why you do not announce the method. A note that explains its own framework is a note about its author. A note that ranks alternatives on consequences, names who loses, and says what would change its mind is a note about the decision, and people extend credit to the second kind even when they disagree with it.