A WORKED EXAMPLE
The question before the options
Most of these studies examine a decision after it has gone wrong. This one comes first – taking an instruction that arrives already settled, and turning it back into a real choice by asking what it was for.
The studies that catch a failure work backward, on something already decided. This one works forward, on something not yet decided, because the instrument is not only a way of seeing what went wrong. It is, first, a way of framing what is being chosen.
A decision usually arrives pre-broken. It comes as a goal that sounds like a conclusion: we need to cut X, adopt Y, restructure into Z. Stated that way it is not a decision at all; it is an answer with the question removed, and the work – the part where it could still go well or badly – has been quietly skipped. The instrument’s first move is to put the question back.
It does this with one deceptively plain question, asked before any option is weighed: what is this for, and would each thing we might do build that up or feed on it? Take “cut cost by fifteen per cent.” The question asks what the cut is in service of, what capability has to survive it, and which way of cutting serves that while another way hollows it.
The goal as handed down – the number, the tool, the new structure – is never the thing it is for; it is a means that has been promoted to an end by being stated first. The question demotes it back to a means, and in doing so re-opens the decision the goal had closed.
Then the options become live. Not a menu of pre-approved variants, but genuine forks: the ways that build up the thing it is for, the ways that feed on it while appearing to deliver the goal, and the non-obvious one that changes the scope of the question rather than the method of answering it. The instrument does not choose among them – that is the work that belongs to the people who must live with the result – but it sorts them by the only question that matters: which of these serves what we are for, and which quietly consumes it.
And it keeps the measures honest from the start, instead of discovering too late that they went wrong. Any number used to track the decision is kept a servant: chosen for what it stands for, paired with a counter-number that lights up if the first is being bought at the expense of the thing it was meant to serve, and never allowed to become the goal in its own right – which is exactly the failure the worked examples elsewhere watch happen after the fact, and which is cheaper to prevent than to diagnose.
None of this predicts the outcome or removes the risk; it is not that kind of instrument. What it does is make sure the decision actually gets made – that the question hidden inside the handed-down goal is asked while there is still room to answer it, rather than discovered, unasked, in the wreckage. A surprising amount of what later needs diagnosing was never decided at all. It was assumed, and the assuming is the first thing the instrument refuses to let pass.
So the forward use and the backward use are one thing seen at two times. The question that catches a failure afterwards – what was this for, and did it serve that? – is the same question that, asked beforehand, keeps the failure from forming. The only difference is whether there is still time to act on the answer.
This is one of the moves the programme makes visible. The full account of what it is and how it works is here.