The Myth of the Rational Organisation
Data does not drive decisions. Decisions recruit data.
The decision that had already been made
The management dashboard was new, and rather beautiful. The transformation steering group had waited eighteen months for it — a single screen, refreshed weekly, drawing from the data warehouse that had cost more than anyone likes to remember, promising at last the thing every board now says it wants: decisions made on evidence rather than instinct. At the third meeting, the group used it to approve the closure of a distribution site. The model on the screen showed an annual saving of two point four million pounds. Everyone nodded. It was, we told ourselves, exactly the kind of clear-eyed, data-driven choice the whole programme had been built to enable.
The decision had in fact been taken months earlier, in a corridor, for reasons no one in the room would say aloud. The analysis had been commissioned afterwards, to support it. And when the closure finally went through, the saving arrived at a fraction of the headline — because three inconvenient figures, the transition cost, the service risk, and the true cost of the redundancies, had been quietly ruled out of scope before the model was ever built. The number on the beautiful dashboard had never been the reason. It had been the alibi.
I have watched versions of this happen too often to treat it as an accident, and it is why I have grown suspicious of the phrase currently doing the rounds — the promise that with enough data, and a good enough warehouse, we can at last build the rational organisation. We cannot, because the rational organisation does not exist, and the methodologies that assume it leave out the one thing every practitioner learns in their first hard programme: data does not drive decisions. Decisions recruit data.
The myth we are being sold
The current enthusiasm — business intelligence, the balanced scorecard, management by fact, the fashionable notion that a firm can compete on the sheer quality of its analytics — rests on a picture of the organisation as a single rational actor: it gathers information, weighs the options, and chooses. That picture is very old, very tidy, and mostly wrong. Herbert Simon told us half a century ago that real decision-makers do not optimise; they satisfice, under bounded rationality, with limited attention and limited time. The behavioural theory of the firm reminded us that an organisation is not one actor at all but a coalition of interests holding an uneasy truce. And the less flattering accounts — the ones that describe decisions as garbage cans into which problems, solutions and participants are tipped more or less as they happen to arrive — capture the texture of steering-group life far better than any flowchart in the delivery handbook.
None of this makes data useless. It means the organisation is not the kind of thing that can be made rational simply by feeding it better numbers, because its decisions are not, at root, calculations. They are settlements between interests, made under genuine uncertainty, about things that cannot be weighed in the same units.
What the textbooks leave out
Three omissions, in particular, separate the diagram from the room.
- Data is produced by interested parties, and is therefore political. The textbook treats data as a neutral input that flows in from the world. In a real organisation every important number has an owner, and the owner has a stake. The scope of a model is set by whoever commissions it, and scope is exactly where the politics hides — the inconvenient figure ruled out of bounds before the analysis has even begun. By the time the number reaches the dashboard it looks objective. The choices that shaped it have been washed away.
- The decision often precedes the analysis. More of our evidence-based choices than we care to admit are rationalisations dressed as inputs — the study commissioned to justify a direction already taken. The rise of audit and controls culture since Sarbanes-Oxley has, if anything, sharpened the appetite for a documented, evidence-shaped trail behind decisions reached on quite other grounds. We have not eliminated instinct and politics. We have professionalised the paperwork that conceals them.
- More data entrenches disagreement as often as it resolves it. When two directors disagree about closing the site, they are not, in truth, disagreeing about the size of the saving. They are disagreeing about whether short-term margin or long-term service capability matters more — two goods measured in different currencies, which no dashboard can convert into one another. Give them more data and each simply finds sharper ammunition. The argument gets longer, not shorter, because the argument was never about the facts.
The serious objection
There is a strong rebuttal to all this, and it deserves the best version of itself. Yes — organisations are political; and that, the objection runs, is precisely the disease that a single version of the truth is meant to cure. A shared, governed set of numbers removes the wiggle room. It stops every function arriving with its own flattering spreadsheet. It disciplines the debate and raises its floor. The whole direction of travel — warehouses, scorecards, dashboards — is right, even where a given implementation is clumsy. Are we not simply describing bad practice, and unfairly blaming the tools?
Half of that is correct, and I would defend it. A single version of the truth is a genuine advance — for operational facts. What were last quarter’s sales; how many orders shipped late; which sites carry which cost: on questions that have an answer, a governed number is worth more than a dozen rival spreadsheets, and I would fight to keep it. But the decisions that consume a transformation programme are not of that kind. They are contested precisely because they trade incommensurable goods under uncertainty, and a dashboard cannot make them rational — it can only make them look it. And there is a sting in the tail: the single version of the truth tends to become a single version owned by whoever controls the warehouse. We do not abolish organisational power by centralising the numbers. We relocate it — to the team that defines the metric, sets the scope, and decides what falls outside it.
“Data does not resolve a disagreement about values. It merely hands each side sharper ammunition, and makes the argument longer.”
The honest organisation, not the rational one
The move worth making is to stop chasing the rational organisation and start building the honest one. In practice that means three things. First, make the politics explicit rather than laundering it through a model — name the interests around the table before the analysis is scoped, not after it has conveniently concluded. Second, use data to narrow the argument rather than to end it: establish the things that genuinely are facts, and then say plainly that what remains is a judgement between competing values, to be argued and owned as such. Third, invest at least as much in the decision process — who decides, on what authority, accountable to whom, and on the record — as in the supply of data feeding it.
A steering group that knows the site closure is a judgement call, and says so, and writes down who made it and why, will make a better decision than one that hides the same judgement behind a two-point-four-million-pound number engineered to produce the required answer. The first can be challenged, learned from, and if necessary reversed. The second can only be defended, because to question the number is to question the people who commissioned it.
The goal is not an organisation that decides without judgement — there is no such organisation. It is one honest enough to show its judgement, put a name against it, and use its data to sharpen the argument rather than to disguise who is really choosing, and why.
The dashboards are not going away, and nor should they. But we should be clear about what we are buying. A warehouse buys us a firmer grip on what has happened. It does not, and cannot, buy us out of the political, uncertain, value-laden business of deciding what to do next. The programmes that pretend otherwise are not more rational than their neighbours. They are merely better dressed.