How a stated problem becomes a decision

The usual shape of consulting work is a study, then a recommendation, then a decision. It works where the decision is a single event and the client is free until the moment they take it. A hospital is not that. The client is spending options continuously from the day they look at a parcel, and most of those options are spent without anyone noticing that a decision has been taken.

By the time a board is presented with a design to approve, the jurisdiction is chosen, the land is optioned, the massing is drawn, and the walking distance every nurse will cover for the next sixty years is already fixed to within a few metres. The approval feels like the decision. It is the ratification of about nine earlier ones.

Which means an answer has a shelf life. The same analysis, identical in every word, is worth a great deal before the option on the land is taken and almost nothing afterwards, because afterwards it can only tell you what you have already bought. Value here is a function of timing, not of quality, and work that is excellent and late is indistinguishable from work that was never done.

So the stream is organised around the gates rather than around the chapters of a report. Each gate gets the answer it needs, in the state it needs to be in, before it closes. Some of those answers are provisional and say so. A provisional answer that arrives in time beats a certain one that arrives after the concrete.

Six gates. Read the last column of each one, because that is the part that matters: it is not what the gate decides, it is what the gate removes from everybody who comes after it.

Two of them do most of the damage when they are rushed. Land, because it fixes the jurisdiction, the catchment and the ceiling on everything, and it is routinely taken on price and availability while the questions that decide the hospital are still open. And massing, because floor area is what departments are cut out of, and a department cut down at massing stage is a service the place needs and will not have for sixty years.

The sixth gate is the odd one out, and it is the reason this group is worth watching rather than only worth hiring. It never closes. A hospital in operation is the only instrument that can tell you whether the first five gates were answered correctly, and Meridian intends to build again. The measurement taken in the first hospital is not housekeeping. It is the input to the next one, which means the first site earns twice: once as a hospital, and once as the thing that made the second one better.

Every answer that goes into a gate carries the same four things, and the discipline is not decoration. It is the commercial requirement of the business model.

Meridian sells its hospitals, and only once they are fully mature and best in class. That means every figure this group has ever relied on will eventually be defended in a room containing a buyer and a buyer's technical adviser, both of whom are paid to doubt it. A number without a basis does not survive that room, so it is not worth writing down now.

The third of those four is the one people find strange, because it means publishing figures marked as schematic and saying plainly that their magnitude should not be trusted. That is the point. The alternative is not more accurate figures. The alternative is the same figures with the uncertainty removed from the page and left in the reader's head as false confidence.

The first five gates all close before a patient exists, and all five are questions about behaviour. How long will people wait. Where will the queue form. What happens on the worst Monday rather than the average one. When does this building stop being able to discharge before noon.

Those questions have a shape that spreadsheets cannot hold. A spreadsheet computes a value from other values, which makes it excellent at capacity arithmetic and structurally incapable of producing an emergent failure, because an emergent failure is a property of a chain and a spreadsheet has no chain in it. The lift bank one car short is the example the instruction uses: every component passes its own review, the building fails, and nothing that examines components one at a time will ever find it.

So the instrument is a model of the whole thing running together. Land, building, staff, policy and patients, in one simulation, with the patients arriving the way patients actually arrive rather than at their mean. You change the massing, or the rota, or where the store room sits, and you watch what happens to the corridor at six in the evening. It is the only way to test a chain before the chain exists.

Two properties of it are worth stating because they are not obvious. It runs the whole life of the asset rather than the period of ownership, which is the only way to see the decision that is fine at year eight and bites at year twenty five. And it runs faster than harm, which is the entire point: today the alternative test costs one patient per finding.

Every solution in this stream ends with a grade, and the grading has one rule that does all the work: the criteria and their weights are published before the scoring, not after it.

The reason is the same reason Meridian's sale condition needs an external definition. A gate a company grades itself against drifts, gently and without anybody lying, toward whatever the company was going to do anyway. Weights chosen after the options are known are not weights, they are a justification with arithmetic on top. Published first, they are a constraint, and they occasionally produce an answer nobody wanted, which is the only evidence that they were ever real.

The same rule has to cover what the answer is scored against, and this is the half that usually goes missing. The comparator set belongs in the brief, named, before any scoring happens. Choosing the alternatives after you already know your preferred answer is the second way to rig a published rubric, and unlike the weights nobody audits it. Weights get argued over in the room. Almost nobody asks why the three options on the chart were those three.

Which gives the comparator set two tests of its own. It has to contain at least one option that somebody credible actually prefers, because a rubric scored against alternatives chosen for their weakness produces a number that is arithmetically correct and worth nothing. And it has to name what was left out, including the option of not proceeding, so that a reader can challenge the boundary of the comparison rather than only the scoring inside it.

The grade is scored against named alternatives rather than against an abstract standard, because best in class is a comparison and a comparison needs something on the other side of it. And the places the chosen answer falls short are published in the same document as the recommendation, at the same size, rather than in an appendix.

That last part is not humility. A shortfall that is written down is a thing somebody can be assigned to close. A shortfall that is left out does not stop existing, it simply stops being anybody's job.

Three things, stated here so that no individual solution has to keep apologising for them.

It does not remove uncertainty. It moves it. Before the run you are uncertain about how the hospital will behave; after it you are uncertain about whether the parameters you fed it are right for this place. That is a much better position, because the second kind of uncertainty can be reduced by measurement and the first kind can only be reduced by opening.

It is worth no more than its parameters. A model carrying a fatigue curve borrowed from another country, a walking distance from another building and an arrival pattern from another population will produce a confident and precisely wrong answer. Where a parameter has not been measured here, the honest move is to run the thing across the plausible range and report which decisions change inside that range and which do not. Most of them do not, which is the genuinely useful finding and the one nobody expects.

And it cannot see what nobody put in it. A simulation contains the mechanisms somebody thought to model. The failure that arrives from a direction not represented in the model is invisible to it, and will stay invisible right up until it happens. This is why the sixth gate never closes, and why the first hospital in a programme is worth more as an instrument than as an asset.