Bringing Clarity Back to a Company Growing Too Fast

By ten past eight, the workshop already smells of coffee and damp cardboard. Élise Garnier runs Lisière, a small condiment manufacturer near Angers. Twelve people prepare, package and ship short production runs for delicatessens and restaurants.

The day looks organised. The schedule is displayed, orders are printed and everyone knows their station. Yet three people are waiting for the same confirmation: should the hotel batch be packaged before the wholesaler’s?

The answer exists in a message sent the previous evening. Nobody knows that it has become a decision.

Marc Delmas does not begin by proposing software. He asks to follow one order from the first sales conversation to the parcel leaving the building.

At Lisière, sales live in Pipedrive, production in several Google Sheets and purchasing in Élise’s inbox. Urgent problems arrive through WhatsApp. Important procedures are explained verbally, often by the person who knows them best.

None of the tools is unreasonable. It is the way they fit together that forces the team to keep checking whether information is still valid.

Marc photographs the boards, listens to operators and notes every moment when someone says a variation of ‘I would rather ask first’.

‘We thought we had a discipline problem. Marc showed us that the real issue was how decisions moved between people.’

Élise Garnier, founder of Lisière

For two days, Marc builds the flow in Miro with the people who carry it out. He separates what triggers an action, what requires a decision and what exists only as a record.

The result is not a perfect diagram. It is a map of hesitation: the approved quote that does not always open a production run, the quantity change that never reaches the workshop, the quality check that remains in a notebook.

Marc’s system keeps the useful tools and gives each one a precise role.

Marc refuses to automate a step until two people can explain in the same way what should happen.

Make now creates a production run when an order meets the right conditions. It flags quantity changes and prepares shipping information. It never chooses the priority between two customers, approves a non-conformity or changes a promised date.

‘A useful automation removes duplicate entry. A bad automation removes the moment when someone should have stopped to think.’

Marc Delmas

Metabase brings together four indicators: late orders, blocked production runs, likely shortages and time spent on rework. Marc deliberately limits the dashboard. A metric that changes no decision does not belong there.

Every Monday, the team spends twenty minutes reviewing these four signals. They no longer look for whoever owns the latest version. They discuss what is preventing the work from moving forward.

Élise keeps an overview without becoming the compulsory route for every decision again.

Some requests still arrive by phone. Operators sometimes annotate a paper document. An experienced person can change the order of a production run when they spot a risk the system cannot see.

Marc is not trying to eliminate these exceptions. He wants them to become visible and capable of improving the shared way of working.

The system is structured enough to pass on. It remains open enough to learn.

Three months later, Lisière handles more orders with the same team. Yet the most important gain is not measured in volume.

Decisions no longer depend on the memory of whoever happened to be present. The tools no longer tell four competing stories. They form a path that everyone can follow, challenge and improve.

‘I no longer spend my days redistributing information. I can finally look at where the company needs to go.’

Élise Garnier

See how to map an operational stack in ToolTrim →