What It Looks Like

How to Catch This Book Being Wrong

An argument that insists every claim should decompose into something testable does not get to exempt itself. The diagnosis and the correction rest on six claims, and each is falsifiable in your organisation within ninety days, without consultants and without anyone's permission. Here is each claim, and the observation that would break it.

Coordination outweighs implementation. If the diagnosis is right, the elapsed time of a typical cross-team change is dominated by negotiation rather than engineering. Take the last three changes that crossed team boundaries and separate the implementation days from the coordination weeks. If the median ratio is close to one to one, the structural diagnosis does not describe your organisation, and the coordination arithmetic running through these pages is someone else's problem.

Precision is selected against. If the selection argument is right, the people who speak in mechanisms and failure modes advance more slowly than the people who speak in narratives. Take the last two promotion rounds and sort the promoted by which language they speak; the sorting is uncomfortable and usually takes about ten minutes. If both populations advance at the same rate, the selection pressure does not operate where you work, and the human cost described in the early chapters is borrowed from somewhere else.

No artefact is permitted to fail. If the avoidance argument is right, your most important documented process cannot be contradicted by anything the system observably does. Ask the owner of that document what runtime behaviour would prove it wrong, and what happened the last time that behaviour occurred. A specific answer, with a date and a consequence, breaks the claim.

A definition is cheaper than the meetings it replaces. The economic claim is that a binding process definition takes roughly three weeks to write, well under an hour a week to maintain, and surfaces discrepancies whose cost exceeds its own. Write one, for one process you already own. If after ninety days it has cost more to maintain than the coordination it displaced, or has surfaced nothing the team did not already know, the economics are wrong, and the thesis that clarity is now cheaper than pretence has failed at the only altitude that matters, which is yours.

The reader's noise is your artefacts' noise. The mechanism claim is that a reconciliation cycle starts noisy and that the noise decays when the artefacts are repaired, not when the models improve. Run the same reconciliation twice without touching the artefacts; if the noise falls on its own, the mechanism is wrong in one direction. Then repair the artefacts the flags point at and run it again; if the noise does not fall, it is wrong in the other.

The first unit moves leading indicators within a quarter. The lagging outcomes (cost, settlement time, customer satisfaction) take two quarters or more, and a sceptic is entitled to wait for them. But contract compliance, drift resolution, and the time a new engineer needs to ship unaided should all move within ninety days of the charter being signed. Baseline them in week one. If none has moved by day ninety, the structural claim has failed its cheapest test, and you should not fund the more expensive one.

The six tests fit inside a quarter and need no budget line. If the argument loses, you will have spent ninety days learning precisely where your organisation differs from the pattern, recorded in artefacts you now own. If it wins, those artefacts are the first quarter of the correction, already done.

...

See also: Full contents · Preview chapters · Illusions in the Boardroom