The Decision Factory

Supply chain · A series of six papers

Most supply chains optimise only the half of the error they can see.

Overstock is counted. The order that quietly went to a competitor is not. That asymmetry, rather than forecast quality, is what makes a planning process drift, and it is why a decade of diligent improvement so often ends with the same people reconciling the same spreadsheets on the last Thursday of the month, in automotive plants, defense supply chains, pharmaceutical distribution, retail and grocery alike.

The series
Six papers
Reading time
59 minutes
Written for
Operations leadership
Drawn from
Auto, defense, pharma, retail
Start with Paper 01 See all six

Three findings that run through the series

Each is uncomfortable, each is testable against your own data within a quarter, and each is developed in full in one of the papers below.

01

The error you can see is the smaller one

Write-offs, markdowns and expedited freight all reach a report and acquire an owner. Lost demand does not. A process charged for one direction of error and not the other will drift toward the unpriced one, every cycle, and look disciplined the whole way.

Paper 02 →

02

A service level is not an objective

Ninety-five, ninety-eight, ninety-nine point five: percentages nobody derived, applied uniformly across items whose economics differ by an order of magnitude. On the same data, a blanket target routinely costs a third more than the economic optimum it was standing in for.

Paper 02 →

03

Accuracy is the wrong contest

Two forecasting models can score identically on error and differ by a third in what their decisions actually cost. Error metrics are symmetric; the business is not. Most model bake-offs are structurally unable to see the difference that matters.

Paper 03 →

The Papers

Six papers in sequence. Each completes one part of the argument and none repeats another, so read in order they are a method, and read alone each still answers a question you can act on this quarter. The series is closed: there is no seventh.

01 The Case for ChangeWhy the process cannot be improved 9 min → Planning processes are not underperforming versions of good ones. They are the stable outcome of a system with no priced objective, which is why fifteen years of diligent improvement has made them worse. 02 What Being Wrong CostsThe economic ledger 10 min → Service levels and fill rates are targets nobody derived. Until each direction of error carries a price your finance director will sign, there is nothing for a planning system to optimise. 03 Forecasting Is Not the PointUncertainty, priced 9 min → A forecast is not a number, and accuracy is not the objective. The organisations that win are not the ones with the smallest error. They are the ones whose errors cost least. 04 The Decision FlowThe architecture 11 min → What the machine actually looks like: five stages, four of them unattended, and the one interface where human judgment enters and is captured rather than lost. 05 Who Does What NowThe organisation after 10 min → Automation does not thin the planning organisation so much as repoint it. Three functions appear that almost nobody staffs today, and skipping them is how the spreadsheets return. 06 Proving ItEvidence and decay 10 min → How to make the case in money on history you have already lived, run the thing in parallel before betting on it, and detect the slow decay that returns a working system to committee.
Cover of The Decision Factory by Adam DeJans Jr. and John Brandon Elam

Also by the authors

The Decision Factory

A novel in the tradition of The Goal, drawing on many of the same ideas as this series without dramatising any single paper in it. It follows a team that rethinks its policies and its simulations to make better decisions under uncertainty, which is the short way of saying what the whole book is about: how to decide well.

By Adam DeJans Jr. and John Brandon Elam. Foreword by Warren B. Powell, Professor Emeritus, Princeton University.

Get the book Start with Paper 01

Paperback/Hardcover Audiobook, early 2027