Why Your CMMS Sits Empty After Six Months

A computerized maintenance management system (CMMS) schedules and records machine maintenance. If technicians rarely use it, check work barriers: hard-to-type asset IDs, long forms, and poor connectivity. QR labels, short forms, and offline recording can help. Trial the workflow with technicians and compare the records against work actually performed.

Nobody buys maintenance software hoping for a blank database. The purchase gets approved on a promise: fewer breakdowns, planned parts, a repair history you can argue with a vendor about. Six months later the license renews and the history shows fourteen work orders — the tickets someone in the office typed in.

Is the problem the software or the data going into it?

Incomplete data is one possibility to investigate; survey benchmarks do not diagnose your installation.

UpKeep's State of Maintenance Report 2026, a survey of 214 maintenance and reliability professionals, found 90% who say preventive maintenance (PM) — servicing a machine on a schedule before it breaks — is worth doing, while only 25.5% describe their own operation as mostly or fully proactive. The same report found 70.9% rating their data readiness as fair, poor, or undeveloped, and only 21.8% who can demonstrate the return on their maintenance spend clearly.

Limble's CMMS Data Quality Benchmark Report, published 16 March 2026 across 200+ asset-intensive organizations, puts a number on the gap. Teams with strong logging discipline capture 82% of the maintenance work they perform; weak ones capture 45%. Across all organizations surveyed, only 20% say other departments trust the asset data enough to act on it.

Read those together and the picture is simple. Limble’s weak-logging group recorded less than half its work; that rate is not automatically representative of every factory. Every report built on top of that — PM compliance, cost per asset, the case for replacing a compressor — is built on a sample nobody chose.

Why does a technician skip the form?

Because the form is priced in minutes and paid for out of the only part of the shift that feels like real work.

Wrench time — the share of a paid shift a technician spends with hands actually on equipment — averages 25 to 35%, against a world-class target above 50% cited from Mobley and Smith's Rules of Thumb for Maintenance and Reliability Engineers (2007). Paperwork sits in the other 65 to 75%, next to travel, waiting for parts, and hunting for the store keys. A form that takes four minutes per job, forty jobs a week, is two hours and forty minutes. The technician does that arithmetic faster than any consultant.

Four common barriers are worth checking:

• Typing the asset ID. A nameplate behind a guard, in poor light, with an eleven-character code and gloves on. Mistyped entries can land on the wrong asset, making history unreliable.
• A login wall with no signal. Plant rooms, basements, and metal enclosures are exactly where the equipment lives and exactly where the phone has one bar.
• Approval before completion. When a technician must wait for a supervisor to sign off a small filter change, the job gets done anyway and never logged. The team quietly vetoes the system while keeping the machine running.
• Fields nobody reads. Ask for a cost centre, a fault subcategory, and a downtime estimate, and the technician learns within a week that none of it appears anywhere. What gets typed after that is whatever clears the screen.

At one industrial site we specified the hardware for, the maintenance budget was argued over for weeks. The line items that mattered were not software modules: outdoor-grade polyester asset labels with printed QR codes, two industrial barcode scanners sealed against dust and water and rated to survive a drop from three metres, and a large screen on the floor showing the team's own numbers. Roughly the cost of one mid-range server, spent on removing typing rather than adding features.

Design the record around the scan, not the keyboard

The first design rule: identifying the asset should cost zero keystrokes.

Print a tag for every asset — a QR code plus a human-readable number — and make scanning it the way a job starts. The scan opens that asset's screen with its history, its open work orders, and its PM checklist already loaded. A phone camera is enough for indoor equipment. Dirty, wet, or outdoor assets earn a dedicated scanner, the kind built to survive a drop onto concrete.

Spend money on the label stock, not on the printer. Paper labels near a chiller or in a wash-down area last about a month; outdoor-grade polyester with a thermal transfer print survives UV, abrasion, and cleaning fluid for years. A missing tag makes the record harder to find. Keep a searchable human-readable identifier and a replacement-label process.

How many fields is too many?

Keep fields required for safety, warranty, and mandatory records. For other fields, ask whether they appear in a report that a named person reads at least monthly. Unused reporting fields can become optional or be removed after reviewing their purpose.

What remains sorts into three tiers.

Fault codes deserve their own argument. Keep the list to about seven entries per asset class, written the way the team talks — leaking, won't start, noisy, worn out, user error — rather than a standards taxonomy borrowed from a manual. A long dropdown gets answered with whatever is first or last on screen, and inconsistent codes are exactly what makes asset history unreadable later.

The photo is the escape hatch. When the form cannot express what the technician saw, the camera can, and a photo costs one tap. Plenty of teams get more value from four thousand photos tied to assets than from four thousand text notes nobody wrote.

Keep fields and approvals needed for safety, warranty, and required records even if no monthly report reads them. Simplifying a form must not discard mandatory evidence.

What happens when there is no signal?

The job finishes anyway. Anything else trains the team to keep a paper notebook, and the notebook always wins.

Build the app offline-first: every action writes to a queue on the device first, and syncing to the server happens in the background once there is connectivity. The technician sees the work order close instantly, walks to the next machine, and the phone catches up in the car park. We covered the mechanics of that pattern — local queue, conflict rules, safe handling of duplicate submissions — in offline-first field apps, and a maintenance app has the same shape as a metering app.

Approval works the same way. Review after completion may suit routine jobs that do not require prior authorisation. Retain pre-approval where safety, warranty, or other rules require it, including isolation, hot work, and permit-controlled jobs.

Measure how much work gets logged, before anything else

PM compliance — the share of scheduled jobs completed on time — is the number every dashboard shows first, and it is meaningless while half the work is unlogged. A team that logs only the easy jobs will report excellent compliance right up to the failure.

Measure capture first, by reconciling the CMMS against records the maintenance team does not control:

• Parts issued from the store. A part may be used, returned, or transferred. A part issue with no matching work order needs reconciliation, and the store ledger is the cleanest audit you have.
• Contractor invoices. Externally performed work is still asset history.
• Downtime logged by production. Production counts stopped minutes for its own reasons. Stops without work orders need investigation; production downtime can also have non-maintenance causes.

Once capture is above roughly 80%, the standard targets start to mean something: 90% PM compliance, reactive work under 20% of total jobs. Limble's benchmark found organizations with strong data discipline running 36% unplanned work against 56% for the weak ones — a difference between surveyed groups, not evidence that logging alone caused the lower rate.

Show the result on a screen the technicians walk past. Team numbers, not individual ones. A per-person leaderboard produces exactly the behaviour you do not want: fast closes, empty notes, fault codes chosen for speed. The board's job is to tell the team that the typing they did last week turned into something.

A rollout that survives contact with the floor

The rollout order matters more than the product choice. This sequence works:

1. Tag the fifty assets that cost the most downtime, not all three thousand. A partial asset register that is accurate beats a complete one that is stale.
2. Write one PM checklist per asset class, not per asset. Ten checklists covering four hundred machines is maintainable; four hundred is not.
3. Pilot with two technicians for three weeks. Pick one who likes the idea and one who does not. The second one finds the real problems.
4. Count the taps. Time the record entry for a routine job on the actual phone in the plant room. As a pilot target, aim for under ninety seconds and ten taps while preserving required evidence.
5. Review unread fields after thirty days. Open the reports, see which fields were referenced, and remove unnecessary ones while retaining safety, warranty, and required records. Test the effect on adoption.
6. Triage history before importing. Bring open work orders and the asset register first. Clean important records for new reports and retain a searchable original archive. Do not discard safety, warranty, or required maintenance records simply because their format is inconsistent.

Treat the first three months as a project with a named owner and a weekly check-in, the same way any internal system needs one — the failure modes are the ones described in software project management for SMEs, and a maintenance rollout hits them all.

The short version

The software is rarely the problem. A technician with a scannable tag, a four-field form, and an app that works without signal has fewer barriers to logging the job. The same technician, asked to type an asset code twice and wait for approval, will fix the machine and tell nobody.

Cut the typing. Cut the fields. Cut the waiting. Then the reports are worth reading.