Sixteen factories, sixteen ways to fix a machine

Every plant had a way of fixing things. None of them was written down.

An 8 to 10 week sprint. One senior analyst and two juniors. A €400 million global chemical company.

Maintenance had grown, not been designed. Sixteen factories had each arrived at a way of keeping their equipment running, and each way worked, in the sense that the plants ran. What did not exist was any version of it that could be compared, taught, resourced or improved.

Three conditions were true at once, and they compound rather than add.

The practices varied by site. Sixteen versions of the same job means sixteen versions of what good looks like, which means no site can learn from another and central functions cannot tell a real problem from a local habit.

Assets and spare parts were only partly tracked. Without a reliable picture of what is installed and what is on the shelf, planned maintenance is guesswork. You cannot schedule against equipment you cannot list.

Maintenance staff were short. This is the one that closes the trap. A team that is short of people spends every hour it has on the machine currently stopped, which is exactly the condition in which nobody has time to build the system that would stop machines stopping.

So repairs arrived as surprises, and each surprise consumed the capacity that would have been needed to make the next one predictable.

The people who could standardise it were the people keeping the plants running

It is worth being blunt about this, because the polite version is useless. Nobody at this company lacked the ability to write a maintenance standard. Maintenance engineers know what good maintenance looks like better than any outsider does, and they would have produced a better standard than an outsider would, if they had been able to start.

A maintenance team short of people does not stop to redesign maintenance. The work that would create capacity is the work there is no capacity for.

There is a second problem, and it is the one that actually needs an outsider. Comparing sixteen sites requires somebody who is not from any of them. Every site believes its way is the way, and often with good reason, because local conditions genuinely differ. Somebody has to sit in all sixteen without a stake in which one wins, and then say which differences are real and which are habit.

And selecting a system on top of that is a specific, unglamorous skill. Screening a market of maintenance management software, defining what to score it on, and holding the line when a vendor demonstrates something impressive that nobody asked for, is a thing you get good at by doing it repeatedly. A plant does it once a decade.

Two figures, and a question worth asking about both

Both are stated flatly, as things that happened. We are going to be awkward about our own page here, because the alternative is to let a reader assume something we cannot support.

An 8 to 10 week sprint cannot measure a 25% reduction in downtime. Downtime is a rate. Establishing that it fell by a quarter means comparing a meaningful period before against a meaningful period after, and a period long enough to average out the one compressor that failed badly in March. That window is quarters, not weeks. What the sprint can honestly claim is a standardised workflow, a tracked asset picture and a chosen system, which together are what a 25% reduction would come from.

The same applies to the 30%. Coverage is a design property, how much of the estate falls under a defined maintenance regime, so it is more defensible than the downtime figure, but the page does not say whether it is coverage achieved or coverage the new regime is built to reach.

Read both as what the design was built to deliver. That is a real and useful thing to have built. It is not the same as a measurement, and we should not let the wording imply that it is.

Our internal record of this engagement also carries a figure the public page does not: €4 million of profit opportunity identified. It is not on the page, so it is not a published figure, and this article does not lean on it. It is noted here because a reader comparing our materials would otherwise find it and wonder which of us to believe.

The senior decided what to score. The juniors opened forty systems and twenty conversations.

What the senior decided. That the workflow gets standardised before any software is chosen, which is the single decision that determines whether the rest of the project is useful. What the eight benchmarking criteria are, and what weight each carries, because that choice quietly makes the shortlist. Which differences between the sixteen sites are genuine constraints and which are habit, which is the judgement no site can make about itself. And when to stop screening, because there is always a forty-first system.

What the juniors did. Twenty interviews across the sites. More than forty maintenance management systems screened and scored against the criteria, one at a time. That is weeks of patient, unglamorous comparison, and it is the reason the shortlist of five means anything. A criteria set applied to five systems somebody had heard of is not a screen, it is a preference.

What stayed behind. The standardised workflow, the criteria set with the scoring, and the shortlist with the reasoning attached. If the chosen system disappoints in two years, the client can rerun the screen against the same criteria rather than starting from nothing. That is the part that is worth more than the recommendation.

Three things, and the first one is about our own page

Whether downtime actually fell by 25%. Not measurable inside an 8 to 10 week sprint, for the reasons above. Our page states it flatly and should not. We are treating it here as the design target, and the honest version of the sentence is that the regime was built to deliver it.

How many factories were studied in depth. Our record says practices varied across sixteen and separately reports four analysed. Sixteen is almost certainly the estate and four the sample, but we do not state it, so this article does not claim more than the estate figure.

Whether the chosen system was the right one. A screen against eight criteria narrows forty to five defensibly. It does not tell you how the winner behaves in year three, under a maintenance team that has changed. The criteria set was handed over precisely so that question can be asked again with the same yardstick.

Where every figure above comes from

From the published case study page. The 25% maintenance downtime reduction and the 30% increase in maintenance coverage, quoted with the page’s own wording, and questioned above rather than restated as fact. The client’s Head of Operations Excellence described the result as structure and visibility brought to maintenance operations, and an accelerated shift toward preventive maintenance.

From our internal engagement record. The €400 million revenue scale, the sixteen factories, the shortage of maintenance staff, the incomplete asset and spare parts tracking, twenty interviews, more than forty systems screened, eight benchmarking criteria and the shortlist of five.

Deliberately not used. The €4 million profit opportunity, because it appears in our internal record and not on the published page. And nothing from the capacity block of our internal template, which is boilerplate printed on most pages regardless of engagement and is not a measurement of anything.

Clients are described by revenue scale and sector, never named.