Predictive maintenance doesn't start with sensors, it starts with the tickets you already have

Last week you stood in front of a tenant with no hot water, while the engineer couldn't come until the day after tomorrow. You logged it as bad luck, the kind of breakdown that happens every week. Then you scrolled back through your inbox and saw that the same boiler had already come up twice in the past two months: once for lukewarm water, once for a unit that switched itself off. Three signals about the same appliance, scattered across separate emails, and nobody had put them side by side.

That's not sloppiness on your team's part. Each ticket was small on its own and was handled properly. The problem runs deeper: your admin is organised around closing tickets, not around the assets they concern. As long as a ticket is an email rather than a line in the history of a specific boiler, the pattern doesn't exist anywhere. Not because the data is missing, but because nobody can see it in one place.

Why you didn't see the breakdown coming

You already had the information to predict this failure. What was missing was the link: three messages about the same unit sat in three separate threads, with two different colleagues, filed under the address rather than the installation. And every emergency callout you miss this way is an expensive one. An engineer called out on the spot costs more than a scheduled service, a tenant without hot water calls three times, and a boiler that finally packs in gets replaced under time pressure instead of on your terms.

Those costs add up across the sector. This spring the Autoriteit woningcorporaties (the regulator for woningcorporaties, Dutch housing associations) launched a thematic investigation into the control of maintenance costs, because maintenance spending by housing associations rose from 4.1 billion euros in 2018 to more than 6.7 billion now, an increase of over 60 percent while construction costs rose by around 30 percent over the same period. The Woonbond points to the accumulation of causes behind that rise. There's little you can do about contractor prices. There's a lot you can do about the share of emergency work in your own workflow.

Predictive maintenance without sensors or a big budget

Predictive maintenance sounds like sensors in every home, a digital twin and a year long implementation project. But the parties who work with this daily say something different: predictive maintenance doesn't start with technology, it starts with data that's accurate. Complete, consistent and structured, in that order. The same article shows what happens when that isn't in place: a line item of 10,000 euros in the long-term maintenance plan (MJOP) that ends up costing 15,000 euros or more, not because it was unexpected, but because it was flagged too late.

For your ticket flow, that means three changes, and none of them require a budget:

• Link every ticket to the asset, not just the address. A home has a boiler, a mechanical ventilation unit, a lift in the building. The ticket belongs to the appliance.
• Use a fixed naming convention. If the same boiler is logged once as "cv", another time as "heating" and a third time as "no hot water", neither a person nor a system will see it's the same unit.
• Record how each job was closed. What did the engineer actually do: reset it, replace a part, top it up? A reset that's needed again two months later tells a different story than a replaced part.

What a pattern actually gets you

Two complaints in two months about the same boiler aren't two incidents, they're a warning. If your history is organised by asset, the third ticket immediately shows you the previous two, and you decide differently: not another reset, but a full service or replacement, scheduled at a time that suits you, with a proper heads up to the tenant instead of an emergency callout.

The numbers back this up too. According to Aedes's exploration of the maintenance process, a tenant submits roughly one repair request per home per year on average. With a portfolio of 500 homes, that means you build up 500 data points a year on the state of your installations. You already have those data points. The only question is whether they add up to a history per asset or evaporate into closed tickets.

So start small. Pick the installations with the highest emergency risk, usually boilers, lifts and pumps. From now on, log new tickets against the asset and once a month, review the list of assets with two or more tickets in the past quarter. That list is your predictive maintenance, before a single sensor comes into play. The boiler from last week would have been on it three weeks earlier.