The Case for Change

What you are actually looking at

Ask a supply chain leader what is wrong with their planning process and you will hear a list. The data is scattered. The forecast is overwritten before anyone sees it. Every region runs its own spreadsheet. The consensus meeting takes four hours and settles nothing. Planners are firefighting. The list is usually accurate, and it is usually treated as a list of separate problems requiring separate remedies.

That reading is the expensive mistake. The items on the list are not independent. They are the visible surface of a single structural condition, and any programme that attacks them one at a time will spend two years and leave the condition intact. This paper sets out what the condition is, why it reproduces itself, and what the alternative to patching it actually requires. It is the first of six papers, and it is deliberately the one with no architecture diagrams in it. Nothing downstream matters if the case for change is not made honestly first.

Six symptoms that travel together

Across engagements in different industries and different functions, the same six conditions appear together. Their reliable co-occurrence is the first clue that they share a cause.

1. The number is assembled before it is judged. Skilled and expensive people spend the majority of the cycle extracting, reconciling and formatting, then apply judgment to whatever time is left.
2. The number is constrained before it is an estimate. What the business expects to sell has been trimmed to what it has agreed to sell, and the two can no longer be separated, which means forecast error can never be measured honestly.
3. Nobody can state the cost of being wrong. Ask what one unit too many costs and what one unit too few costs. If the answer is a pause, every downstream calculation is unpriced.
4. Every unit runs its own version. Each region, plant or category has its own file, sequence and vocabulary, so consolidation is monthly translation performed without written rules.
5. Disagreements are settled by seniority. When commercial and supply chain disagree, the outcome depends on who is more senior or more persistent. That is not a decision procedure with any known accuracy.
6. Overrides are constant and unrecorded. People routinely change the system's number, and none of the reasoning is captured, so the system never learns what its operators know.

Four or more of these describe most large organisations most of the time. They are not evidence of a badly run team. They are the predictable output of a process that was never designed, only accumulated.

Why these are one problem, not six

The instinct on seeing that list is to fix the items individually. Better templates for the assembly problem. A cleaner extract for the data problem. A standard agenda for the meeting problem. This instinct is wrong, and understanding precisely why is what separates a programme that works from one that consumes a budget cycle.

The six conditions are the equilibrium of a system with no priced objective. Trace the loop. Because nobody knows what error costs, there is no arbiter between speed and caution, so assembly expands until it fills the available time. Because assembly fills the time, judgment is compressed to the final hours, so experienced people override at the end rather than reason at the start. Because the overrides are unrecorded, the model never improves, which justifies more overriding next cycle. Because each region has learned to survive this on its own, each has built a private variant, which makes any central standard feel like an imposition. Every symptom is load-bearing for the others.

Remove any one of them and the system reconstitutes it within two cycles, because the condition that produced it is untouched.

This is why organisations can buy a planning platform, run a data programme, hire a forecasting team, and find three years later that the planners are still reconciling spreadsheets on the last Thursday of the month. Nothing was done wrong. The wrong thing was done well.

Where the margin actually goes

The financial consequence of an unpriced process is not evenly distributed, and this is the part that most reliably surprises boards. Exhibit 1 sets out the shape of it.

Three of the four leakages are recorded somewhere. Excess and obsolescence appear as write-offs. Expedited freight and rework appear as cost variances. Capital tied up in cover appears, at least in aggregate, on the balance sheet. Each has an owner who is asked about it and who therefore works to reduce it.

The fourth has no such owner. Lost sales are not a line item. The customer who ordered elsewhere this quarter does not send a note explaining why, the trial that was never run generates no data, and the account that quietly reduces its share does so over a horizon longer than any planning cycle. The result is a systematic and entirely predictable bias. A process charged for one direction of error and not the other will drift toward the unpriced direction, every cycle, and will look disciplined the entire way, because the metrics it is judged on are precisely the ones that improve as it drifts.

The improvement paradox

There is a second argument for replacement rather than repair, and it is the one that tends to change minds in the room, because it explains a frustration the leadership team has already lived through.

Incremental improvement makes this class of problem worse rather than better. Every improvement is designed and applied locally, because that is where the pain is felt and where the initiative is funded. A better spreadsheet in one region is a larger reconciliation burden at the centre. A smarter local adjustment rule is one more undocumented translation between the local number and the group number. Local optimisation raises global divergence, and divergence is the thing that makes the whole process expensive.

Fifteen years of diligent, well-intentioned, individually justified local improvement is precisely how an organisation arrives at the condition described above. Nobody was negligent. The incentives worked exactly as designed and produced a result nobody wanted.

Read Exhibit 2 as a statement about capacity rather than about efficiency. The organisation is not short of analytical talent. It has bought a great deal of it and is spending most of it on clerical reconstruction of numbers that already exist somewhere in its own systems. The prize from redesign is not headcount. It is the recovery of scarce judgment for the minority of decisions where judgment actually changes the outcome.

What replacement does and does not mean

Replacement is a loaded word in an organisation that has already survived two transformation programmes, so it is worth being precise about the scope of the claim.

It does not mean replacing systems. In most engagements the enterprise resource planning system, the data platform and the existing forecasting models all stay exactly where they are. It does not mean starting the analytics work over; in our experience the data science team is usually building something competent that has simply never been connected to a decision. It does not mean removing human judgment, which as Paper 04 argues is the one thing the architecture is specifically designed to capture rather than eliminate.

What is replaced is the process by which a recurring decision gets made: the sequence, the artifacts, the definition of who decides what on what basis, and above all the objective the decision is ranked against. That is a governance change. Its first three steps require no software, no vendor selection and no capital request, which is the single most useful thing to know when deciding whether to begin.

The three commitments that get this funded

Change of this kind fails at the point of sponsorship far more often than at the point of execution. What follows is the minimum that has to be true before anything else is worth starting, expressed as three commitments a leadership team either makes or does not.

1. One objective, in writing, applied consistently. A single economic measure that every planning decision is ranked against, stated once and honoured from the annual budget through to the monthly review. An objective restated annually and renegotiated monthly is not an objective, it is a preference.
2. A named owner for the cost of being wrong. Someone in finance, by name, who will sign what one unit too many and one unit too few cost the business. Paper 02 is entirely about this artifact, because it is the highest-leverage and least-built component in the discipline.
3. An explicit mandate to retire the shadow process. If people are still judged on producing a defensible number by hand, they will keep the spreadsheet, and the spreadsheet will win, because it is the artifact their performance is measured against.

Notice what is absent. There is no platform decision, no vendor, no headcount request and no multi-year roadmap. Those come later and are comparatively easy. The three commitments above are difficult precisely because they are cheap: they cannot be delegated to a procurement process, and they require an executive to write down a number that someone will later be measured against.

The next paper takes that number apart. It is the shortest artifact in the entire architecture and the one that decides whether any of the rest of it produces value.