What the Break Map Shows, and What It Decides

Because a map answers the wrong question. It tells you what the process is, and the question on the table is what to do about it, which needs two more artefacts. The first is a number: how much this process costs per week in rework, waiting and missed handoffs, calculated across the candidate workflows rather than across the company. The second is a ranking: which candidate returns the most for the least build risk. A vendor who hands you a beautiful diagram and a proposal has skipped the step where the diagram turns into an argument, and you are being asked to approve a build on aesthetics. The map is the evidence. The cost figure and the matrix are the finding.

Its construction, and the fact that it is written down before anyone knows what it will justify. It is built from loaded hourly cost rather than headline salary, because the hour a person spends on rework costs the business what that hour costs to provide, not what appears on the payslip. Volumes come out of the operational systems for a period long enough to include a bad month, not out of an estimate offered in a meeting. And it is scoped to the candidate workflows only, which keeps it small and specific and stops it drifting into the kind of company-wide inefficiency figure that cannot be checked against anything. The test of the number is simple: it is the baseline every later return claim gets measured against, so it has to be one you would be content to be held to.

Two axes, kept deliberately crude. Feasibility asks whether the inputs are structured enough, the systems reachable enough, and the rules stable enough that software can do this reliably rather than mostly. Return asks how much of the measured weekly cost this candidate would actually recover, at conservative assumptions rather than optimistic ones. Scoring both matters because the highest-return candidate is frequently the least feasible one, and a plan that ignores that ships a project which is exciting for three weeks and abandoned in the fourth. The output is a queue, and the point of the queue is that the first item is defensible to someone who was not in the room.

Often enough that the gate is real, which is the only property that matters. If the arithmetic at conservative inputs does not clear positive, the diagnostic deposit is refunded, the reasons are written up, and the engagement stops there. This sounds like a marketing line and it is the opposite: a filter that never rejects anything is not a filter, it is a formality, and every consultancy that has one which never fires is running the second thing while describing the first. The client still leaves holding the map, the baseline and the matrix, which are worth having on their own and which make the next automation attempt, whoever runs it, materially cheaper.