Your transformation doesn’t need a new strategy. It needs a diagnosis.
A doctor who prescribed before examining you would be struck off. Yet prescribing before examining is the default operating model of organisational transformation.
A change programme stalls. Engagement is flat, the new ways of working haven’t stuck, the steering group is quietly losing faith. And the response, almost every time, is to reach for a cure. More training. A reorg. A new framework. A culture programme. A fresh set of OKRs. The cure is usually whatever the person in the room already knows how to sell.
Nobody asks the obvious question first. What, precisely, is wrong here? And how would we know if we were right?
Medicine has spent a century building a rigorous, teachable answer to exactly that problem: how to reason well under uncertainty, with incomplete information, when being wrong is expensive. It is called differential diagnosis. Almost all of it transfers to change work, and the parts that don’t are worth understanding too.
Generate a differential, don’t pattern-match
The first discipline is the hardest, because it runs against how experts actually behave. A good clinician does not settle on the first plausible explanation. They generate a differential: a list of competing hypotheses that could each account for the symptoms. Then they work through it deliberately.
Most transformation diagnosis skips this entirely. The consultant has a favourite diagnosis. For some it is always culture. For others it is always leadership, or psychological safety, or a lack of agile maturity. They have seen it before, they know the cure, and so every organisation that walks through the door has the same disease.
The fix is to force the differential out into the open before you fall in love with an answer. Why has this stalled? Unclear goals. No real ownership. Incentives that quietly reward the old behaviour. Change fatigue. A capability gap. A sponsor who does not actually believe in it. Write the competing explanations down, all of them, and only then start narrowing.
Rank by likelihood and by lethality
Medicine ranks hypotheses on two axes, not one. The leading hypothesis is simply the most likely given the symptoms. But running alongside it is a separate and vital category: the must-not-miss diagnoses. The conditions that are rare but fatal if overlooked. A doctor will actively test to exclude a pulmonary embolism even when it is unlikely, because the cost of missing it is death.
Transformation has must-not-miss diagnoses too, and they are almost always the ones nobody probes for, because they are uncomfortable. The executive sponsor does not actually want the change and is backing it for show. The change threatens the power base of someone senior who will quietly strangle it. The money to finish what was started is not there. These are rarely the most likely explanation. They are routinely the fatal one. If you only ever chase the comfortable, common diagnosis, you will miss the thing that was always going to kill the programme.
So rank twice. What is most likely, and separately, what would be fatal if you ignored it. Probe for both.
Stop running discovery that tells you nothing
Not all evidence is equal. In medicine, a test with a high likelihood ratio meaningfully shifts what you believe. A weak one barely moves the needle whatever the result. Clinicians choose tests for their power to confirm or exclude, not for thoroughness.
Most organisational discovery fails this test badly. The all-staff engagement survey, the dozens of near-identical stakeholder interviews, the maturity assessment scored out of five. Reams of data, almost none of it diagnostic, because whatever the result it does not change the diagnosis or the action. It feels like rigour. It is mostly theatre.
Look instead for pivot points: the few questions that split the space fast. “Have the last three initiatives also stalled?” If yes, you are looking at something systemic, and half your differential just collapsed. One good question can be worth a month of surveys.
Know when to stop
Clinicians work on a probability line with two markers on it. Below the test threshold, a cause is too unlikely to be worth chasing, so they stop investigating it. Above the treatment threshold, they are certain enough, so they stop investigating and act. The skill is knowing where you are on that line and behaving accordingly.
This single idea cures the two opposite diseases of change work. One is the endless discovery phase that never commits, forever gathering more input because nobody will call it. The other is the leap straight to a solution before anyone has understood the problem. Both are failures to read the probability line. You do not need certainty to act. You need to be past the treatment threshold.
Treatment is a test, not a conclusion
A diagnosis is provisional until the response confirms it. You treat the leading hypothesis and you watch what happens. If it does not resolve, that is not failure, it is data. You re-rank the differential and try again with what you have learned.
This is why a fixed eighteen-month transformation plan is a category error. You cannot pre-decide the treatment for a condition you have not finished diagnosing. Real change work is iterative: a strong working hypothesis, a deliberate intervention, an honest look at the response, and a willingness to re-rank when the body tells you that you were wrong.
The biases are a mirror
Medicine is honest about how this reasoning fails, and it catalogues the failures. Read the list and you are reading the transformation industry’s own pathology back to it.
Availability: treating the diagnosis you saw most recently as the most likely. “It worked at the famous tech company, so it must be a squad-model problem here.”
Chasing zebras, or base-rate neglect: reaching for the exotic diagnosis. The bold reorg, the culture revolution, when the dull truth is that nobody owns the work and the goal was never defined.
Representativeness: ignoring the inconvenient evidence that does not fit the story you have already chosen.
Confirmation bias: running discovery to confirm the diagnosis you wrote into the proposal, rather than to disprove it. The answer was pre-baked, and the evidence-gathering is set dressing.
Premature closure: stopping at the first plausible cause. “It is a culture problem.” Five words, and quite possibly the most expensive sentence in change work.
You will recognise all five, because you have watched them happen, and if you are honest, you have done at least three of them yourself.
Where the analogy breaks
The comparison is not perfect, and pretending it is would be its own diagnostic error. Medicine’s likelihood ratios come from large studies. Organisational diagnosis has no comparable evidence base, so its probabilities are informed judgement, not numbers. Treat them as numbers and you have invented false precision.
More importantly, the patient is not passive. An organisation is political. It hides symptoms, it resists, and it changes the moment it knows it is being observed. You are diagnosing a system that has opinions about its own diagnosis, and the treatment alters the disease as you apply it.
None of that breaks the method. It just means the method is a discipline for thinking clearly under uncertainty, not a machine that hands you certainty. Which is exactly what it is in medicine, too.
The first move
Here is the practical shift, and you can make it on the next stalled thing that lands on your desk. The first question is not “what is our strategy” or “which framework.” It is “what exactly is wrong here, what are the competing explanations, and how would we know which one is true.”
Write down three hypotheses for why it has stalled. Mark the one that is most likely, and separately, the one that would be fatal if you ignored it. Then ask what single piece of evidence would most change your mind, and go and get that first. You have just done more genuine diagnosis than most programmes manage in a quarter.
Treat before you diagnose and you are not transforming anything. You are just prescribing.
Tim Robinson, 27 June 2026