It is 4:40 pm. The technician closes the electrical cabinet, wipes their hands and takes out a phone. They open the work order, type “motor replaced” and close it. Thirty seconds, standing, between two emergencies.
In the database, the job looks complete: an asset, a date, a name, a description. Nobody will read it again for months. The day someone does, trying to understand why that motor failed for the third time this year, they will find “motor replaced” three times. And nothing else.
I did not make this scene up for effect. I see it in the numbers, with minor variations, on every continent.
As of October 28, 2026, more than 11,700 people in 158 countries have signed up for FreeMaint, the CMMS I build. A sign-up is not a user: many people open an account and never come back. So for this article I only kept the 525 workspaces that have actually deployed the tool, with at least one asset and five work orders. That is roughly 54,000 work orders over the last twelve months, 84,000 assets and just under 1,500 people active in the past month.
Here is what these teams taught me. Not what they should do: what they do.
What gets written: the what and the where
Let's start with the good news, because it is real. Teams write. At the median site, almost every work order created by hand has a description. Across all work orders, about eight in ten are linked to an asset, and as many to a person. The log exists, and it is kept.
The problem shows up at close-out. On completed corrective work, a root cause is recorded in about 4% of cases at the median site. Nearly half of all sites have never filled it in, not once in twelve months. Failure codes fare worse: six sites in ten never use them. Seven in ten have never recorded a single part used on a completed work order. And more than half have never attached a photo.
In other words, we record what was done and on what. We almost never record why it happened, or what it took to fix it. Yet that is exactly what ISO 14224 expects from failure data before it is of any use: the failure mode, its cause, its mechanism, then the resources used to restore the equipment. Without that information, the history tells a sequence of events. It explains nothing.
The asset record follows the same pattern. Location is almost always filled in, category in about nine cases out of ten. Manufacturer six times in ten, model four in ten, serial number one in ten. Then come the fields that would support a decision. Half of all sites have never set the criticality of a single asset. About two in three have never entered an acquisition cost, a purchase date or a parent asset. Four in five have never recorded a warranty end date. The typical record is a label: a name, a place, a family. It is an inventory, not a technical file, and criticality, the field that should rank emergencies and steer preventive work, is empty one time in two.
I do not think this is negligence. The fields that get filled have one thing in common: they serve the person writing, at the moment of writing. The asset, to find the machine. The assignee, to know who goes. The empty fields serve someone else, later: the reliability engineer, the buyer, the auditor. We ask technicians to work for a reader who never writes back.
The defaults have won
A second pattern is quieter, and more awkward for KPIs.
In the CMMS, a work order is born with a “medium” priority and a “corrective” type. These are defaults, not decisions. About three work orders in four keep the medium priority. On work orders created by hand, the median site changes the type on only one in seven, and more than one site in three never changes it. Estimated duration is filled in on one work order in four. As for the “emergency” type, it accounts for half a percent of work orders: in twelve months, only 15% of sites used it even once.
The consequence is simple. When a dashboard shows a preventive-to-corrective ratio or a breakdown by priority, it is partly measuring the form. EN 13306 carefully distinguishes corrective, preventive and condition-based maintenance. Someone still has to tick the right box.
What gets abandoned: whatever depends on someone else
To measure abandonment, I isolated the sites that are still active, those that created at least one work order last month. Among those that started using a module more than two months ago, I counted the ones that stopped within a month of first use. So this is not the tool they left; it is a feature.
The contrast is sharp. About one site in three abandons stock movements. One in four abandons work requests, and as many stop logging hours. By contrast, preventive maintenance, once it is running, is abandoned by only one site in seven. Photos on work orders, the same. Failure codes, among those who adopt them, also hold: one site in five drops them.
The dividing line is not technical. What sticks directly serves the person entering it: the photo saves them explaining again, the preventive plan tells them what to do on Monday. What breaks relies on someone else's discipline: a storekeeper booking out every part, a requester using the form instead of stopping the technician in the corridor. A module survives when the person who enters the data is also the one who benefits from it.
What works everywhere: the log, and the phone
Some things do not depend on culture. In each of the six regions I compared, description, asset and assignee are filled in on at least two work orders in three at the median site. And in each of them, parts used stay at zero. The minimal log is universal. So is the gap.
The other constant is the phone. Over the past month, nearly one connected person in two used the mobile app at least once, and one in four used nothing but mobile. The database cross-check gives the same proportion: 46% of active users have the app installed. On the work-request side, about one in three comes from a phone: roughly one in five through a QR code stuck on the machine, one in six through the app. At more than one site in three, most requests already arrive that way.
The shop floor did not wait for someone to install a terminal in the workshop. It is already in the tool, standing in front of the machine. That is also where data quality is decided: a work order filled in on the spot, right away, carries the right information. One reconstructed on Friday evening loses it.
What changes from one country to another
Regional differences exist, and they do not always point where you would expect.
On planning, Europe leads: at the median European site, about three work orders in ten are preventive. Latin America follows, at around one in four. North America and Asia sit around one in five, sub-Saharan Africa around one in eight. In North Africa and the Middle East, preventive work is still rare, around 6%, on a narrow base of about twenty sites.
On diagnosis, the order flips. In Latin America, more than one completed corrective job in five records a root cause. It is around 16% in North Africa and the Middle East, 14% in North America and about 6% in Europe. Photos follow the same slope: about one completed work order in eight in Latin America, North Africa and the Middle East, almost none elsewhere.
One last gap: after-the-fact logging. In Europe and Asia, more than one completed work order in five was created and closed within five minutes, which means it was written once the job was over. In the Americas and sub-Saharan Africa, it is closer to one in eight.
I have no definitive explanation, and I distrust stories that are too neat. One possible reading: where planning has long been established, the CMMS is used to follow a plan, and corrective work becomes a deviation you log quickly to move on. Where preventive maintenance is just starting, every breakdown counts, so it gets documented. What is certain is that no region does everything well. Europe plans better than it explains. Latin America explains better than it plans.
Three lessons for anyone deploying or relaunching a CMMS
1. Require one diagnostic field, at the right moment. Not fifteen mandatory fields: they will be filled in with anything. Just one, when a corrective job is closed, picked from a short list of ten or fifteen failure codes or causes, drawing on the ISO 14224 classification. It is the field missing everywhere, and the only one that turns a history into lessons learned.
2. Never hand a KPI to a default value. If your preventive-to-corrective ratio relies on a pre-filled type, it is measuring the form. Run the test this week: count last quarter's work orders still at “medium” priority and “corrective” type. If that is the vast majority, remove the default, or compute the ratio from what was actually planned.
3. Deploy first what serves the person entering the data. The phone, the QR code on the machine, the photo, the preventive plan that says what to do on Monday. Leave for later the modules that rely on another department's discipline, or bring that department on board from day one. And give technicians back what they write: having the machine's history in front of them when they intervene is the best reason to fill in the cause next time.
Method: anonymous aggregates computed read-only on October 28, 2026, over the 525 workspaces with at least one asset and five work orders (work orders from the last twelve months; training and test workspaces excluded). No group of fewer than five companies is published, nor any segment dominated by a single account. The 11,700 users and 158 countries are sign-ups: usage is measured on those who actually deployed. Mobile figures come from product analytics and exclude people who declined tracking.
More figures, measured on the same population, are published free to read in The State of Free Maintenance 2026.
What about you?
Which field in your CMMS stays empty even though everyone swears they fill it in? And if you finally got it filled in, how did you do it?
References
- • ISO 14224:2016 — collection and exchange of reliability and maintenance data for equipment: failure modes, causes and mechanisms
- • EN 13306 — Maintenance terminology: corrective, preventive and condition-based maintenance
- • FreeMaint Research — The State of Free Maintenance 2026, method and reference population
In the same series
Record it on the spot, while the information is accurate
FreeMaint runs on the web and as a mobile app, with work requests through a QR code stuck on the machine. Free to start.
Frequently asked questions
Which work order fields should be mandatory?
As few as possible. Teams fill in on their own what helps them in the moment, the description, the asset, the assignee, and leave empty what helps later. Piling on mandatory fields produces values picked at random. The one requirement that really pays off is a single diagnostic field when a corrective job is closed, a cause or a failure code, picked from a short list.
Why do technicians not record the failure cause?
Because that field serves someone else, later, and it is often asked at the wrong moment, in a list that is too long, with nothing given back. At the median site, about 4% of completed corrective jobs record a root cause. For that to change, the list has to be short and the data has to come back to the person who entered it: the machine history in front of them when they intervene.
Does a CMMS need a mobile app?
Over the past month, nearly one connected person in two used the mobile app, and one in four used nothing but mobile. About one work request in three comes from a phone, often through a QR code stuck on the machine. Mobile is no longer a bonus: it is where data gets entered on the spot, while the information is still accurate.