Designed for prevention, not reporting
Three hundred days, four continents, one lotion — and a loss no single team could see.
Practice — Design research · Decision-intelligence product design · Data visualisation
Role — Led the design engagement — direction, and the calls that followed
Team — Supply-chain domain, data engineering and design, as one team
Period — September 2022 – February 2023 · 22 weeks
Client — A global prestige beauty company (Fractal, CPGR)
Three hundred days. Four continents. One lotion.
The network optimisation modeller decides where the company manufactures next. In 2022 he was assembling that picture out of staggered regional reports, one Power BI view at a time. He was not short of data. He was short of a path.
In 2021 a high-demand lotion sat unavailable for over two months in the company’s biggest market. Their own estimate of the missed opportunity was around $60M — an estimate, not a realised loss.
The brief was never “build us a dashboard”. It was: make sure a disruption on the scale of COVID cannot do this to us again. What came with it was a requirement — don’t give us another Power BI dashboard, we need beauty.
Design research with the category organisation’s product owners followed one lotion end to end. Source APAC to manufacture in North America, 188 days. Manufacture to distribution in EMEA, 90 days. Distribution to sale back in APAC, 15 days. Roughly three hundred days across four continents — and no single team could see the whole of it. The path was the cause.
The rules we designed by
Answer the decision, not the data request.
The network is one object, not six reports.
Argue in low fidelity, before the data is ready.
What we did
01 · Designed one decision, not six audiences.
Six teams called themselves the user. We designed the decision the modeller was making, and let everyone else read the same object from their own side. Instead of six report families, each true on its own.
02 · Made the network one object.
Dependencies and exceptions sit in the same view as the metric. Lead time, sufficiency and continuity are answerable in one place, with drill-down and geo view in a single frame.
03 · Argued in wireframes, not decks.
Low fidelity went in front of the client before engineering feasibility was settled, so the reframe could be agreed while it was still cheap to change — and only against options engineering had already called feasible. Instead of waiting for data readiness.
04 · Answered “beauty” with a system.
A palette with colour assigned by meaning rather than mood, chart types chosen for the decision each view answers, and a density rule. Beauty defined as a system, not a finish — which is why it held as scope moved and the data changed underneath it.
How the work ran — FRAME · EXPLORE · SYNTHESISE · CO-CREATE · LEARN · PILOT · LAUNCH. Fractal’s decision-backwards process — the behaviour-change path, not the traditional one.
Project highlights
The challenge — Two months blind
A lotion out of stock for two months in the biggest market, and a modeller deciding future manufacturing from staggered regional reports. The cause sat in a three-hundred-day path no single team owned.
The approach — One window, one decision
Research followed a single product end to end. The product that followed organised itself around lead time, sufficiency and continuity — the decisions — rather than restyling the reports that already existed.
The outcome — Prevention, not reporting
One window replaced six report families. Questions that once needed a request cycle are answered directly. Signed off by the client; never instrumented afterwards.
Gallery
Three hundred days, mapped — 188 days to manufacture, 90 to distribute, 15 to sell. The cause, drawn as a path.
Low fidelity was the argument — the whole network as one object, agreed in wireframes months before a build could prove it.
Beauty, defined as a system — one hue per state, chart types chosen for the decision each view answers.
One decision per view — the drill-down answers, it does not just display.
“It is with great pleasure, I say APPROVED. Thank you for your diligence, embrace of feedback as a gift and challenger spirit.” — Director, client · on sign-off
Results — “Approved. Never measured.”
What changed for the user
One window replaced six report families. Lead time, sufficiency and continuity became answerable in the same view. Questions that once needed a request cycle got answered directly.
What it was worth
~$60M — the client’s own estimate of the missed opportunity this was built to prevent. An estimate, not a realised loss.
The programme set targets of 30% end-to-end lead time, 30% NPI time-to-market, 80% sourcing in region of manufacture and 70% manufacturing in region of sale. These are targets the programme set, not measured outcomes.
The honest line
The client signed it off and nobody instrumented what happened next. The only verified evidence this case study carries is the client director’s approval — which is why the claim is about how it was designed, not what it returned.
Three hundred days. Four continents. One window.