Shadow Maintenance: The Work That Happens Without Work Orders

Key takeaways

See the work order types this bypasses.

• Every plant has a layer of maintenance work that happens without a corresponding work order, a technician swapping a worn part during a changeover, an operator tightening a loose bolt, a quick adjustment between batches. The work is real, the time is real, the parts consumption is real. None of it shows up in the CMMS.
• Most plants assume the unlogged share is small. When a plant actually measures it, by comparing its OEE stop stream to its work-order record, the gap is routinely much larger than expected, often a substantial fraction of total maintenance effort. That gap distorts every analysis that depends on work-order data: MTBF, failure rate per asset, parts consumption, cost per machine hour.
• The fix is not eliminating shadow maintenance, much of it is operationally appropriate. The fix is creating a low-friction logging mechanism so the work is captured without becoming bureaucratic enough to discourage the work itself.
• The single biggest lever is reducing the time cost of opening a work order. If logging a 5-minute fix takes 4 minutes of CMMS clicks, the technician will skip it forever. A 30-second mobile capture changes the calculation.

Where the shadow comes from

Shadow maintenance is not a discipline problem. It is the predictable result of a system where the cost of logging exceeds the perceived value of being logged. A technician performing a 5-minute adjustment between batches has three options:

• Open a work order, document the symptom, document the fix, close the work order, 4 minutes of overhead on a 5-minute task.
• Skip the CMMS entirely, do the fix, move on, 0 minutes of overhead, 0 record.
• Promise to log it later, forget, 0 minutes of overhead, 0 record, accidentally.

Options 2 and 3 are operationally rational. The technician's job is to keep the line running, not to feed the CMMS. If the CMMS makes feeding it cost more than the technician thinks it's worth, the CMMS stops getting fed.

Most plants underestimate how much of their maintenance work falls in this category. The OEE event stream shows the stops; the CMMS shows the work orders; the gap between them is where shadow maintenance lives. The article on the work order management system covers the data structures that make the gap visible.

What the unlogged work distorts

Failure rate per asset

If 30% of small failures get fixed without a work order, the asset's recorded failure rate is 30% lower than the actual rate. The PM schedule is calibrated against the recorded rate. The schedule under-protects the asset by the gap.

Parts consumption

Parts pulled from the storeroom without an associated work order do not get attributed to the asset that consumed them. The asset looks cheaper to run than it is. The retirement calculation, which depends on operating cost per asset, mis-ranks assets.

MTBF and MTTR

Both metrics depend on the population of failure events being complete. Shadow maintenance inflates recorded MTBF, the real failures are more frequent than the records show, so the time between recorded failures looks longer than reality.

Its effect on MTTR runs the other way and is easy to miss: because the quick fixes never enter the data, the recorded average is computed only from the longer repairs, so recorded MTTR is overstated. Either way, the recorded numbers drift away from reality once a meaningful share of work goes unlogged.

Technician workload

The team appears less loaded than it actually is, because a meaningful share of their work is invisible. Capacity-planning conversations about hiring or rebalancing run on the wrong numbers.

The piece on manufacturing KPIs covers how the related metrics depend on complete event data.

What the fix looks like

The fix is not "make everyone log everything." That promise has been made in every CMMS rollout and broken in most. The fix is reducing the friction of logging until the technician's calculation flips.

1. Mobile capture at the asset

The technician scans an asset QR code, picks a failure mode from a short list, taps a parts consumed indicator, and the entry takes 30 seconds. No keyboard, no laptop, no walking back to the office. The article on the work order management system covers the mobile patterns that make this feasible.

2. Voice or photo capture

For asset classes where the failure mode is hard to type, a 5-second voice memo or a single photo gets stored against the asset. The CMMS doesn't need a perfectly structured record; it needs the data to be captured at all. Structure can come later.

3. End-of-shift batch log

For very short interventions, the technician can log the day's batch of small fixes in a single end-of-shift screen. 90 seconds total, 6 fixes captured. Compromise on real-time accuracy, win on completeness.

4. Auto-prompts from OEE events

When the OEE system detects a stop above some threshold that did not generate a work order, the supervisor's screen surfaces a prompt: "Line 3 had a 6-minute stop at 14:22. Was it logged?" If the answer is no, the prompt offers a fast-capture path to log it after the fact. The piece on root cause analysis covers the related event-driven prompts.

What not to do

Don't mandate without measuring

A directive to "log everything" without measuring whether logging actually increased usually produces theatre, technicians log the easy items and continue to skip the awkward ones. The audit that catches this is comparing OEE stops above a threshold to work orders for the same window; the gap is the unlogged work.

Don't punish gaps

If unlogged work is treated as a discipline issue, technicians learn to suppress the discussion rather than improve the logging. The framing should be "the system is making it too hard to log; let's fix the system." Almost always true.

Don't over-engineer the capture form

Each additional field on the capture form reduces the share of work that gets captured. The minimum viable capture, asset, failure mode (from a short list), elapsed time, is enough to materially improve the analysis. Optional fields can be added later.

What the math looks like after the fix

Plants that successfully reduce the logging friction typically see:

• Recorded failure rate rises materially in the first quarter as the previously-shadow events become visible.
• Recorded MTBF falls correspondingly, not because reliability got worse, but because the measurement got more honest.
• PM schedules tighten where the rise concentrates (the previously-shadowed events were on a specific asset class).
• Headcount conversations get more honest (the team was carrying more load than the numbers showed).

The first reaction is usually alarm: "did our reliability suddenly get worse?" No. The measurement got more honest. The piece on the preventive maintenance schedule covers how to recalibrate PM cadences against the corrected failure rate.

How Fabrico fits

The mobile-first capture is what makes the fix possible at scale. Where a unified OEE + CMMS platform helps is that the OEE-event-to-work-order audit is automatic, and the mobile capture goes into the same database the analysis runs against. Fabrico is built so 30-second logging is the path of least resistance. To see what your shadow-maintenance gap would look like against your live OEE-vs-work-order data, book a demo.

Frequently asked questions

Is some shadow maintenance always going to exist?

Yes. Some 5-second fixes will never be worth logging individually. The target is not zero; it is reducing shadow to under 10% of total maintenance effort, which is enough that the analyses are reasonably trustworthy.

How do we measure shadow maintenance in our plant?

Compare OEE stops above 2 minutes against work orders for the same time window over a month. The gap is your floor estimate of shadow maintenance. Add another 30-50% to account for sub-2-minute interventions that never show up in OEE either.

What about parts consumption that bypasses the storeroom?

Some plants have unofficial spare-parts caches at the line maintained by individual technicians. These are pure shadow, no log, no consumption record, no tie to the asset. The fix is acknowledging the cache and bringing it into the storeroom as a "line stock" sub-location with simplified replenishment, not pretending the cache doesn't exist.

Will reducing shadow maintenance cost extra time per shift?

Marginally. The 30-second capture is real time. The offset is that the better data makes future PM schedules and capacity decisions more accurate, which saves more time downstream than the capture costs upstream.

What is the most common implementation mistake?

Trying to capture shadow maintenance by mandating more detailed work orders on the existing form. Adding fields to the form increases logging cost, decreases logging rate, and produces less complete data than before. The right direction is shorter capture, not longer.