The State of Free Maintenance 2026

Why this report exists

Every maintenance benchmark in circulation describes the same population: large plants, in rich countries, running paid software with a five-figure annual licence.

That is not what maintenance looks like for most of the world.

FreeMaint is a free CMMS with no user limit. That pricing model attracted a population no vendor has data on — two-person teams, plants in Algeria and Côte d'Ivoire, contractors managing fifteen assets. This report is what their data says.

Everything below comes from aggregated, anonymised production data. Nothing is modelled, estimated or extrapolated. Where the data is weak, we say so and publish the number anyway.

1. The planned maintenance gap is not a gap. It is a chasm.

The reliability profession has a target. Limble puts it plainly on its own blog: “High-performing maintenance teams aim for an 80/20 ratio — 80% planned, 20% unplanned.” The figure appears in every maturity model and every consultant's deck, usually attributed to SMRP Best Practices, where the metric itself (Planned Work, 5.3.1) is real even if the exact best-in-class number is industry convention rather than a published SMRP figure.

Here is the measured reality across 447 sites running a CMMS today:

Not 80%. Not 50%. Twenty-two and a half percent.

And the distribution is worse than the median suggests:

The gap between the weighted average (32.7%) and the median site (22.5%) is itself the finding: a handful of mature operations pull the average up. The median describes the actual shop floor.

More than one site in four has never recorded a single planned maintenance task. Not “does little preventive maintenance” — none at all.

A note on comparability. The SMRP metric is calculated on labour hours; ours is calculated on work order counts, because hours are not reliably logged in this population. Since reactive jobs typically consume more hours than planned ones, an hours-based ratio would likely be lower still. The direction of the gap is not in question; its exact magnitude is.

By region

Europe is roughly twice as planned as Asia. Africa sits ahead of Asia — which will surprise anyone who assumes maintenance maturity tracks GDP.

2. Priority labels have stopped working

We measured the median time between a work order being created and being closed, split by the priority the team assigned to it.

Read the middle two rows again.

A work order marked “High” is closed 53% SLOWER than one marked “Medium.”

This is not noise. It is 5,730 high-priority work orders against 20,535 medium ones.

Only “Critical” actually accelerates anything. Between “High” and “Medium”, the label has inverted — it now predicts a slower response.

The most likely explanation is priority inflation: “High” has become the default for anything that feels urgent but doesn't justify “Critical”. Once most things are High, High means nothing, and the queue sorts itself by something else entirely.

Priority inflation has a measurable cost, and this is the first time anyone has put a number on it.

3. The real maintenance team has two people

The CMMS market prices per user and designs for departments. Here is the size of the teams actually using one:

77% of maintenance teams running a CMMS have five people or fewer. 39% are one person alone.

Median team size: 2.
Median asset count: 15.

Two people. Fifteen machines. That is the median maintenance operation in this dataset — and it is a segment most CMMS vendors do not have a pricing tier for, because a per-seat model generates almost no revenue from it.

4. Maintenance software has a geography problem

Maintenance teams have signed up from 138 countries.

The rankings hold few surprises at the top — the United States leads. What matters is further down:

And then:

Africa accounts for more registered maintenance teams than Asia — 287 against 256.

By continent:

Nine African countries clear ten registered sites each. Not one of them appears in any published maintenance industry study. They were not overlooked — they were never counted, because counting them requires a product they can afford.

5. Four teams in ten do not work in English

Maintenance software is written in English and sold in English. Here is what teams do when given the choice:

40.2% of workspaces contain at least one user who explicitly changed the interface language.

Among those who made an explicit choice:

French leads Spanish. Portuguese outranks German. Arabic matches German.

This is not a localisation checkbox — it is four users in ten changing a setting within minutes of installing, because the default does not work for their team.

6. Maintenance requests come from the floor, not the office

Work request origin, across 6,803 requests:

One maintenance request in five is opened from a phone, standing in front of the machine. 17% come from simply scanning a QR code stuck on the equipment.

The QR code beats the mobile app five to one. No app to install, no account to create, no training — a sticker on the asset and a phone camera.

7. Free is not disposable — for those who deploy it

Among workspaces created more than six months ago that completed setup (at least one asset and five work orders):

44.3% are still active six months later.

The comparison that matters: among all registrations older than six months — including those that never opened the product — the figure is 4.9%. Over half (52.9%) never logged in a single time.

These two numbers describe two completely different populations, and conflating them produces nonsense in either direction. Free registration attracts enormous curiosity traffic. Free deployment retains at a rate that would be respectable for paid software.

Caveat: this measurement covers 70 workspaces. It is directionally solid but the base is narrow, and we will re-measure as the 2026 cohort matures.

8. One in five work orders is written after the fact

20.7% of closed work orders are completed less than five minutes after being created.

28.7% within the hour.

These are not jobs managed through the CMMS. They are jobs done first and recorded afterwards.

That has two consequences. First, it means a CMMS in a small team functions primarily as a logbook — a record of what happened — before it functions as a planning tool. Second, it means every median duration in this report understates real repair time, because a fifth of the sample was never timed at all.

We publish the number rather than filtering it out, because “the CMMS is a logbook first” is the more useful finding.

What this adds up to

Five conclusions, each measured rather than argued:

1. 1The 80% planned-maintenance target is disconnected from practice. The median site is at 22.5%, and 27% are at zero. A benchmark that nobody reaches is not a benchmark.
2. 2Priority fields have degraded. “High” now predicts a slower response than “Medium”. The taxonomy needs rebuilding, not more training.
3. 3The typical maintenance team is two people and fifteen assets. An industry that prices per seat and designs for departments has built for a minority of its own market.
4. 4Africa has more registered maintenance teams than Asia and appears in no published study, because studies count licences, not teams.
5. 5Free software retains when it is deployed. The failure mode is registration without deployment, not deployment without retention.

Methodology

Aggregated, anonymised data from FreeMaint workspaces, snapshot dated 10 September 2026.

Usage metrics (preventive ratio, time to close, team size, asset count) cover the 448 workspaces that have deployed the product — defined as at least one asset and five live work orders — representing 39,936 work orders and 51,906 assets. This threshold excludes empty registrations, which would otherwise distort every ratio.

Geographic and language data cover the 1,708 workspaces that declared a country at registration, from ISO 3166-1 country codes.

Privacy floor. No group of fewer than five companies is published. Any segment where a single customer represents more than half the volume is excluded entirely — this removed Africa's weighted figures and all of Oceania from the ratio tables. No company name, email or identifier appears anywhere in this report, and no individual record was examined.

Excluded populations. One cohort of 113 workspaces created within five weeks by a single training institution was removed from all usage metrics; these are student exercises, not plants. 942 work orders imported with a completion date preceding their creation date were excluded from duration calculations.

Durations are medians, not averages. Given that 20.7% of work orders are closed within five minutes of creation (retrospective logging), median durations understate actual repair time.

Reproducibility. The database is live; two runs minutes apart returned 518 and 519 workspaces for the intermediate population. All figures are frozen at the snapshot date.

Note on population. Section 1 reports 447 sites where this section reports 448. One deployed workspace has no computable preventive ratio; it is counted in the deployed population but excluded from the ratio distribution.