How to choose a CMMS: the method, the traps, and the spreadsheet nobody dares abandon

Almost every plant reaches the same question at the same moment: when the maintenance tracking spreadsheet becomes impossible to keep. It has grown campaign after campaign, it holds tabs nobody can read any more, and the person who built it has moved to another job. So you search for "how to choose a CMMS", you land on comparisons that rate thirty products against criteria nobody actually uses, and you leave with no idea where to start.

The real starting point is not the list of products. It is knowing what you expect from the tool, and what no tool will do in your place.

What a CMMS does, and what it does not do

A computerised maintenance management system has had a stable core for thirty years. It records an equipment fleet, receives intervention requests, turns them into work orders, schedules preventive maintenance, tracks the time spent and the parts consumed, and keeps the stores inventory. It is a tool of flow: it organises what the teams do, day after day.

That core is done well in most of the products on the market. If your need sits there, almost anything will fit, and the choice comes down to usability and price far more than to functions. Two products can carry the same feature list and feel completely different once a technician has to use them under pressure. The differentiator is rarely the function itself, it is how few taps it takes to record a real intervention on a real shift.

Where expectations break is on what a CMMS does not do natively. It knows a pump has been repaired three times this year, but it cannot tell you the pump body is thinning by two tenths of a millimetre per year. It records that an inspection took place, but it does not read the report of that inspection or pull the measurements out of it. It knows the date of the next preventive intervention, but it has no idea when the equipment is due for its statutory revalidation, which follows an entirely different logic.

That boundary is not a flaw in the software. It is a difference of trade between the management of work and the tracking of condition, a distinction we return to further down. Understanding it early is what separates a project that lands from one that spends its first year forcing a work-flow tool to behave like a condition record.

The spreadsheet nobody dares abandon

Before comparing any products, you have to look squarely at what you are trying to replace. In most plants it is not an old piece of software, it is a workbook.

The maintenance spreadsheet has genuine qualities that explain its longevity: it is free, nobody needs training, it bends to any organisation, and it works offline on any machine. The queries "CMMS Excel" and "free industrial maintenance software" are among the most frequent in the field, and they say something true. For a small fleet, a well-kept spreadsheet is enough for a long time, and telling people otherwise is dishonest.

It also carries three defects that become deal-breakers as it grows. It has no memory of access: there is no way to know who changed what, and no way to roll back. It does not handle concurrency: two people opening it at the same time create two versions. And above all, it rests on its author. The day that person leaves, the formulas become unreadable and the workbook dies slowly.

The right moment to move to a CMMS is not a number of assets. It is the moment your spreadsheet has become a single human point of failure: one person understands it, and the organisation depends on them. That tipping point often arrives well before anyone admits it, because the spreadsheet keeps working right up to the day its author is unavailable.

The criteria that actually decide

Comparisons rate CMMS products on dozens of criteria. In practice, five questions settle the matter, and they are about you, not about the software.

Who is going to enter data, and from where. A technician has no desk. If they have to walk back to a control room to record an intervention, they will not record it, and the CMMS will fill up with data that is false by omission. The real question is mobile data entry: on a tablet, on a phone, possibly offline in an area with no signal. A product that is excellent on a desktop and painful in the field loses its substance within a few months.

What you have to carry over from the existing setup. History migration is the most underestimated line item. Carrying over an equipment fleet is simple; carrying over three years of work orders and hundreds of inspection reports in PDF format is a project in itself. Many efforts succeed at the rollout and fail on the migration, because nobody had costed the time to re-enter the data. This point connects directly to turning PDF reports into usable data.

What the software will have to talk to. An isolated CMMS becomes one more data-entry chore. A useful CMMS exchanges with what already exists: the control system, the ERP for spare-part purchasing, the sensors where there are any. The question is not "does it offer connectors", it is "which flows will I actually plug in, and which will I carry on entering twice".

Where your data lives. Online hosting or installation on your own servers is not merely a technical preference. In food and beverage and in pharmaceutical plants, the real condition of the equipment is sensitive data, and some boards refuse to let it leave the company. It is a knock-out criterion for a share of sites, to be raised at the start of the project and not at the end.

Who answers when things go wrong. The software matters less than the contract that comes with it. Response time, the language of the support desk, the existence of someone who knows the industry and not only the product. An average tool that is well supported beats a brilliant tool delivered with nobody behind it.

Free, paid, spreadsheet: what hides behind the price

The query "free CMMS" deserves an honest answer, because it covers three different realities.

There are genuine open-source CMMS products, whose code is open and whose use carries no licence fee. They are real and in use, but their being free simply shifts the cost: you have to install them, maintain them, update them and back them up yourself, which assumes an in-house IT skill. The software is free, the time it demands is not.

There are entry-level versions of commercial products, free up to a number of assets or users. They are fine to start with, provided you know you will pay as you grow, and provided you check that you can get your data out the day you leave.

And there is the spreadsheet, which remains the most widespread free CMMS in the world. We have set out its limits; the important thing is not to leave it for a paid tool without first checking that the new tool solves precisely what was jamming in the old one.

The real cost of a CMMS is never read on the "licence" line. It is made up of the setup time, the migration of the existing history, the training of the teams and the daily data entry. That last item is the heaviest and the least visible: a CMMS is only worth the data you put into it, and putting it in takes time from people who have none to spare.

CMMS or asset integrity software: the real divide

This is the point where projects go wrong most often, because the two families look alike from a distance.

A CMMS answers the question "what do we do, and when". It is organised around the intervention. Its basic unit is the work order.

An asset integrity management software answers a different question: "what condition is this equipment in, and how is that condition changing". It is organised around the equipment and its measurement history. Its basic unit is not the intervention but the condition datum: a wall thickness, a finding, a deadline. This is the subject developed in tracking the condition of industrial equipment.

The two are complementary, not competing. A CMMS on its own knows you have inspected a vessel, but not that its wall thickness is dropping dangerously. Integrity software on its own knows the equipment is degrading, but does not raise the work order that will get it repaired. Mature sites use both and make them talk to each other: integrity detects, the CMMS executes.

The classic mistake is to ask a CMMS to do the work of integrity software, rigging custom fields to note thicknesses in them. It works for a while, then collapses when you have to plot a trend over five years or rebuild a history per condition monitoring location. A spreadsheet grafted onto a CMMS is still a spreadsheet, with all the same weaknesses, only harder to see.

Where artificial intelligence genuinely changes something

You have to be precise here, because "CMMS with AI" has become a sales pitch that covers a great deal of emptiness.

AI does not replace the CMMS and does not magically improve scheduling. Where it concretely changes the work is upstream of data entry, on the tasks nobody wants to do.

It reads the documents the CMMS cannot read. An inspection report in PDF, a contractor's write-up, a manufacturer's notice: an AI can extract the equipment tag, the measurements, the findings and the deadlines, and prepare them for the CMMS or for the integrity software. That is exactly the friction point where history is lost today, for lack of time to re-key it. Recovering even a fraction of that lost history is often worth more than any single feature on a comparison sheet.

It finds in the history what a keyword search cannot: every piece of equipment that has shown the same type of defect, every intervention linked to a common cause. On a fleet of several hundred assets, that is the difference between half a day of digging and an immediate answer.

What it does not do, though, must be said just as plainly. It does not decide to keep equipment in service. It does not revalidate. It does not sign. Human validation remains necessary, and it is the human who carries the responsibility. An AI that reads three hundred reports will get some of them wrong, and the skill lies in knowing where to look to check. The frame around that boundary is the same one described in reducing unplanned downtime: the tool raises the signal, the human decides. Choosing the AI solution itself, between building, buying and repurposing a consumer tool, is a separate decision, detailed in choosing an industrial AI solution.

A five-step method for choosing

• Choosing on the feature list.Every CMMS ticks the same boxes. What will set them apart is real use in the field, not the length of the comparison table.
• Forgetting to cost the migration of the existing history.This is the item that sinks projects that are otherwise a success. Three years of history does not re-key itself for free.
• Asking the CMMS to do integrity software's job.Noting wall thicknesses in custom fields works for a while, then collapses the moment you have to plot a trend.
• Believing a paid tool beats a spreadsheet on principle.A well-kept spreadsheet beats a badly filled CMMS. Only change if the new tool solves what was really jamming.
• Taking "AI" for a feature in itself.Ask what it reads, what it extracts, what it checks. Without a concrete answer, it is a pitch, not a function.
• Neglecting reversibility.Software you cannot get your data out of is a trap, whatever its price.