The Good Meeting Is the Most Expensive Failure

Of the three ways a decision fails, two announce themselves. A tour is visibly directionless, and everybody in the room knows it by slide twenty. A sales document draws an objection, because someone always notices that the reasons could not have come out any other way.

The good meeting does neither. Everyone agrees, the recommendation is accepted, the sponsor thanks the team, and the project is marked complete. Nothing in the room signals failure.

Then a quarter passes and the operating process is unchanged.

Why it happens

Because the decision was never specified to the level where it could run, and nobody in the meeting had the job of noticing.

The sponsor heard a recommendation and agreed with it. The analyst heard agreement and recorded it. Between those two facts sits a specification that does not exist, and specifications do not create themselves from goodwill.

The gap is usually not the headline. It is the unglamorous part underneath: who owns the parameter, what the process does when an input is missing, who may break the rule and at what price.

What has to be specified

Seven things, and every one of them corresponds to a way rules die.

An owner of the running rule, which is not the sponsor of the study. The person who can change a parameter without convening a task force. If analytics retains the parameters, the rule dies when the analyst changes seats, and this is by some distance the most common way good work is lost.

A cadence, meaning when the rule fires and separately when humans review it. “Ongoing” is not a cadence. Ongoing is how freeze dates pass.

The rule itself, including what the system does when it is unsure. Unsure is a case. If you do not specify it, a planner will specify it with a safety pad, and the pad will be invisible to everyone including you.

An exception path, naming who may break the rule, how the break is priced, and how it is recorded. An unpriced exception path is already a shadow rule, and it will widen until the exception is the rule.

Feedback, meaning what you will look at to see whether the ranking is still right. A scoreboard for this rule, not a tour of last quarter.

Revisit triggers, as dates and flip conditions rather than monitoring. “The third of July, or sooner if bypass exceeds eight percent for four weeks” is a complete thought. “Monitoring” is a label.

A check that the rule can know what it needs at the moment it fires. This one is the cleanest separator between a rule that runs and a rule that was designed. A stocking rule requiring next quarter’s demand cannot run, however well it performed in backtest, because backtests are computed with information the rule will not have.

The other half: who undoes this

A specification is necessary and not sufficient, because a rule lands inside a system of local scoreboards.

Before writing the recommendation, find the adjacent team that will dislike it if it works as designed, and go ask them what they will do about it. Not whether they agree. What they will do.

If they say they will escalate, you have a disagreement, which is workable and belongs in the open. If they say they will comply, ask what it costs them and whether anyone is watching that cost, because uncompensated compliance decays quietly. If they say they will work around it, you have not finished, and you now know it before implementation rather than after.

None of these people is being obstructive. A plant manager measured on utilization will run long batches. A parts organization measured on holding cost will thin the tail. Everyone in the system is being reasonable, which is precisely why the system produces an unreasonable aggregate.

Legibility beats optimality more often than people expect

The effective value of a rule is what it delivers when followed, multiplied by the probability that it is followed. The second term is almost never estimated, because compliance is assumed.

It should not be. A planner who does not understand a rule will route around it, and the routing will not be reported. A simpler rule with a higher compliance rate frequently beats a better rule with a lower one, and the comparison is rarely made because it feels like an argument for doing worse work.

The practical test is whether the person running the rule can explain it to a colleague in one sentence. If they cannot, expect a workaround and plan for it rather than being surprised by it in the quarterly review.

When you are not allowed to install

Sometimes the organization will not let you specify the operating process. It is not your remit, or the owner will not commit, or the change needs a system nobody will fund.

Say so in the write-up, explicitly, as a limitation of the decision rather than as a footnote. “This recommendation is not installable until Parts Planning owns the class parameters” is a finding. It names the actual blocker and it puts the cost of inaction where it belongs, rather than leaving the analyst to absorb it privately as a failure to influence.

That sentence is uncomfortable to write and it is the difference between a project that ended in a good meeting and one that ended in a diagnosis.