One engagement, written out end to end.

Situation

Order volume had grown by roughly half over two years. Nothing about how the order was promised had changed with it.

Purchasing, the warehouse and customer service each held a defensible view of what was on order and what was available, and none of the three matched. Customer service quoted from the order system. The warehouse picked against what was physically on the floor. Purchasing worked from supplier confirmations that reached the order system only after goods were booked in. On any given morning the three numbers differed by enough to matter, and the difference was resolved the same way every time: someone walked down to the floor, or picked up a phone.

That worked while volume was small enough for everyone to know the exceptions by name. At the new volume it produced late promises, duplicated counting and a steady stream of credits. The commercial cost was visible in the accounts as service failures and expedited freight; the operational cause was not visible anywhere, because the checking that held the process together happened outside every system that recorded it.

We were asked in after a quarter in which on-time dispatch fell below the level the largest two accounts had contracted for. The board wanted to know whether the answer was a new system. It was not, and saying so early was the first useful output of the work.

The constraint across functions

No single function owned the answer to the only question the customer asked.

Following the work rather than the documentation, the order path had two points where it stopped moving. Availability was confirmed by conversation, not by record, so the promise depended on who answered. Stock was then re-checked in a second system before picking, because nobody trusted the first check enough to act on it. Both points were compensating for the same missing thing: an agreed source of record for what could be sold.

This is why it was a cross-functional constraint rather than a departmental one. Each function had optimised locally and each was right to. Purchasing protected supplier terms. The warehouse protected pick accuracy. Customer service protected the customer relationship in front of them. The cost of the three positions landed in the gaps between them, where no manager had authority and no measure was kept.

Scope was drawn deliberately narrowly: order capture through to dispatch, for standard and contract orders only, with purchasing included where it drove availability and excluded everywhere else. Project and bespoke orders were left out. Two functions argued for a wider boundary; the narrower one produced an answer inside six weeks that the wider one would not have.

Decisions leadership took

Analysis narrowed the options. Four decisions could only be taken by the people accountable for the operation.

1. One source of record

   The order system became the single authority for availability, and the warehouse view was reconciled to it nightly rather than consulted in parallel. This was contested: it moved a daily judgement away from the people closest to the stock.

2. Rules in place of escalation

   Exceptions were given thresholds — quantity, customer tier, promised date — with a named decision owner at each. Anything below the threshold stopped being escalated at all.

3. A promise the operation could keep

   Standard lead time was extended by one day for contract orders, against the sales director’s initial position. A date that holds proved worth more to the two largest accounts than a shorter one that did not.

4. Where the work stopped

   Bespoke orders were explicitly left outside the redesign for the current year, with the reasoning recorded so the decision could be revisited rather than rediscovered.

Each of these was taken at a gate, in a meeting with the sponsor and the three function heads present, and recorded with the evidence it rested on. Two of the four went against the initial recommendation. That is the point of the gate: the analysis is ours, the decision is the operation’s.

Handover

The engagement closed on a working set the management team runs without us.

Nothing in the handover depends on our presence to keep working. The order-flow view marks the two points where work used to wait, so a recurrence is visible rather than absorbed. The rule set runs to a single page and is held by the operations manager, who also owns any change to it. Ownership is named against each step of the path, and each boundary has one accountable role rather than a shared one.

Measurement was reduced to three figures reviewed monthly: promises met against promises made, exceptions handled below threshold, and the reconciliation variance between the two stock views. The review slot sits in the existing management meeting rather than in a new forum, and the decision record travels with it, so the reasoning behind the contested decisions survives the people who took them.

Six weeks of framing, diagnosis and redesign; a further period alongside the team while the rules were put into the line. The engagement ended when the management team ran two review cycles on its own.

The five stages behind this field note →