Running a Diagnostic Is an Intervention, Not a Survey

A diagnostic is usually planned as though it were a measurement. A fixed thing is out there, a set of instruments is pointed at it, and a reading comes back. The plan survives until the first interview invitation goes out. From that moment the organisation knows it is being examined, and it starts behaving like an organisation that knows.

The behaviour is not deception. People prepare. They reread the numbers they expect to be asked about, they check what their peers were asked, and they arrive with an account of their function that is tidier than the one they hold in an ordinary week. Two teams already in tension both assume the diagnostic will settle that tension in someone's favour, so both make a case rather than describe their work.

What the invitation list announces

The list of people invited is read as a statement about where the problem is thought to be. Interview eight people in operations and two in commercial, and the building concludes that operations is the subject. That conclusion travels faster than any finding and is harder to withdraw than one. The first output of a diagnostic is not a report but the signal sent by its own design, and that signal lands weeks before anything is written.

Nothing about this is a reason to run the work in secret, which produces a worse version of the same problem. It is a reason to treat the design of the diagnostic as the first thing the organisation will read.

Most diagnostics that go wrong go wrong on scope rather than on method. The work begins with a date for the readout and no agreement about what is being diagnosed, so the question widens with every interview, and every widening is reasonable at the moment it happens. Somebody mentions pricing. Pricing is obviously relevant. Pricing joins the scope, and the readout now has to cover a territory no three weeks of work can cover.

Three boundaries hold a diagnostic together, and all three cost less to set on the first day than on the twentieth.

• The question. One sentence, naming the outcome under examination rather than the function suspected of producing it.
• The period. The stretch of time whose evidence counts, chosen so that it contains at least one clean instance of the thing going wrong.
• The exclusions. What will not be examined even if it comes up, written down, so that declining it later is not read as protecting someone.

Access is a boundary, not a logistics problem

Access belongs with those three rather than with scheduling. A diagnostic that can interview the leadership team but not the people doing the work can describe intent and nothing else. One that can read the systems but not sit in the meetings can describe the record and nothing else. What can be reached decides what can be concluded, so the limits of access belong in the scoping conversation rather than in a caveat at the back of the report.

A diagnostic that runs well has three passes, and they differ in kind rather than in depth. Running them in order is most of what separates a diagnostic from a round of interviews with a summary attached.

The first pass collects without a hypothesis. Interviews stay deliberately loose, documents are read for what they assume rather than what they assert, and the material is not yet sorted into a story. The discipline is to resist the first plausible explanation, which tends to arrive early and tends to match whatever the person who commissioned the work already believed.

The first plausible explanation usually arrives in the third conversation.

Where the hypothesis is allowed in

The second pass forms a claim and traces it. A cause is written as a mechanism — this policy produces that queue, which produces the delay the sales team reports — and the pass is spent following the mechanism forward and backward to see whether each link holds. The tool in use decides where the cause is allowed to sit, which shapes this pass more than most teams notice while they are inside it.

The pass most diagnostics skip

The third pass tries to break the claim. It looks for the period when the cause was present and the symptom was not, the team with the same structure and a different outcome, the person whose account contradicts the mechanism. It is the pass that gets cut when the readout date moves closer, and it is the one that decides whether the work was worth commissioning. A diagnosis that has not survived a deliberate attempt to disprove it is a hypothesis with a cover page.

The readout is where most of the value is lost. A finding lands in a room for the first time, the function it implicates hears it in front of its peers, and the next forty minutes go to defending rather than examining. The finding may be correct. It is now attached to a defeat, and the person who has to act on it is the person who just lost the argument in public.

Returning a finding before writing it down

The alternative is unglamorous. Each finding goes back to the people it concerns before it is written, in private, framed as a mechanism to be checked rather than a verdict to be accepted. Some findings break at this stage, which is the point of the step: a correction from the person closest to the work costs far less before the readout than after it.

What survives arrives at the readout already known to everyone it names. The room's attention goes to what to do about the finding rather than to whether the finding is fair, and that shift is the difference between a diagnostic that changes something and one that is filed.

The pattern across the diagnostics we are asked to rescue repeats itself. The method was sound, the scoping was verbal, and the question widened until the readout had to cover everything and could commit to nothing. By then the organisation has spent three weeks of senior attention and holds a document that describes it accurately and tells it nothing it did not already suspect.

The repair is at the front of the work rather than at the end of it. One written question, one period, one list of exclusions, one map of who can be reached, agreed before the first invitation goes out — because the first invitation is already an intervention. The same discipline decides how the method set is chosen and what has to be settled before a proposal is written.

Run this way, a diagnostic produces fewer findings and more decisions, which is the only ratio that matters by the time the work is done. Where that thinking becomes work is consulting, and the neighbouring notes in diagnosis before solution cover the sequence on either side of it.