The Case for Writing Python on the Factory Floor

Every large manufacturer has the same problem. Sensors everywhere, readings every few minutes, data piling up in CSVs that nobody opens. Plenty of capture. Almost no insight.

That was the situation at a Rolls-Royce precision facility where, at nineteen, basic VBA and a few macros were enough to dig into energy monitoring data and surface things nobody had spotted. Not because of any particular brilliance, because asking questions of data directly, without raising a ticket or briefing an analyst, removes the bottleneck between "the data exists" and "someone acts on it."

That small ability has compounded across every role since.

Lean production systems are everywhere in manufacturing. Visual management boards. 5S checklists. SPC charts. Quality documents. The concepts are sound. The execution is almost always terrible.

Walk into most factories and you'll find lean tools running on printed checklists, spreadsheets nobody opens twice, and paper audits that exist because the quality system demands them, not because anyone uses them to make decisions. The tools aren't bad in concept. They're bad in form.

The gap between "printed and filed" and "captured and visible" is enormous. And closing it doesn't require a six-figure software contract. A basic CRUD application, nothing complicated, nothing fancy, can take a paper checklist and turn it into something that captures data, shares it, visualises it, and makes it available for people to actually act on. An afternoon's work, if you know enough code to build it yourself.

That's the key phrase: build it yourself. Not pitch a project. Not wait for the digital team's roadmap. Just fix it.

At AUAR, where the team builds robotic construction systems, there's no SAP. No ERP. No enterprise infrastructure. But there's still a need to know where parts are, what they cost, and what they've been allocated to.

So the answer was a custom inventory system. Not flashy. Not enterprise-grade. But it does the job: a part arrives, gets booked in, location tracked. Booked out, allocated, costed. The same level of control a large manufacturing facility gets from its systems, without the complexity, the cost, or the twelve-month implementation.

The pattern is the same at every scale: spot a problem, build a tool, ship it to the team. What's changed recently is the speed.

A few years ago, the constraint was time. The problems were visible. The solutions were often obvious. But building the fix, the dashboard, the tracker, the tool, was a project in itself.

AI has collapsed that cycle. Today it's realistic to have a problem at lunchtime and share a working tool with the team by end of day. That's not aspirational. That's a normal Tuesday.

In lean thinking, there are different levels of problem-solving. Root cause analysis, countermeasures, verification, all still essential. But the solution cycle, the part where you actually build the fix, used to be the bottleneck. If the countermeasure required a new tool or a new way of capturing data, that was weeks of work. Now it's hours.

No inventory system means parts get lost, which means delays on the build line. The traditional response: requirements document, vendor evaluation, implementation timeline. The new response: there's an inventory system by tomorrow morning. That's the step change.

The pushback on all of this is predictable: "That's IT's job. That's what the digital team is for."

And it's not wrong. Not every ops leader needs to write code. The fundamental skill of someone running a factory hasn't changed and won't change: it's interpersonal. The best factory managers know everyone's name, people trust them, they can walk a shop floor and feel what's off before the data confirms it. That's the core. That will always be the core.

Those interpersonal skills turn a factory from bad to good. They might even make it great.

But the people who really optimise, who find those last few percentage points, who turn good into genuinely great, are the ones who can harness data. Not replacing the human skills. Multiplying them.

Here's the thing about running a factory: the leader delivers zero value as an individual. The entire job is helping teams be more effective. Anything that cuts bureaucracy and increases floor time is leverage. Anything that helps the team spend more time on value-adding work is leverage.

Building the technical competency to solve problems directly, without waiting for another team, without a project plan, without permission, that's the highest-leverage skill available. It's not complicated. It's not flashy. But it compounds every single day.

Tom d'Arcy is Head of Manufacturing & Deployments at AUAR, where he builds robotic construction systems. Previously Rolls-Royce, Edwards Vacuum, and BCG.