How to Prioritize a Maintenance Backlog by Risk, Failure Cost and Repair Time

Why order age is a poor ranking tool

The most common shortcut is to work the oldest orders first. It is easy to understand but dangerous: a ten-day-old cosmetic paint request can outrank a fresh bearing replacement on the single compressor that feeds the whole line. Age only tells you how long a job has waited; it says nothing about what happens if it waits longer.

Deferred maintenance usually begins with minor problems pushed aside because resources are limited, and the longer tasks sit, the larger the repair, downtime, risk and energy cost become. Mature teams therefore separate strategic deferral, when a low-priority job is consciously moved forward so a critical one runs, from accidental deferral, when the queue simply grows from emergencies and staffing gaps.

Replacing “oldest first” with “most dangerous first” needs a model. At minimum a planner needs two inputs: the asset's criticality from a systems criticality analysis and the operational criticality of the specific job. Multiplying the two gives a more accurate priority. The sections below add cost of failure and repair time to that base.

Factor 1 — Risk: criticality and probability of failure

Risk is best expressed as probability of failure multiplied by consequence of failure. Consequences are scored across several impact categories — safety and environment, production stoppage, repair cost and quality — in the same spirit as FMEA/FMECA criticality reviews, where you take the highest consequence score and multiply it by the probability score.

Criticality is a property of the asset, not of one work order, and it is fairly stable. Before scoring, build a clean asset hierarchy (site, area, unit, equipment, component) and assess criticality at the equipment level, not the bearing level. Refresh the ranking every 12–24 months or when the process, regulations or demand change.

A common mistake is treating expensive equipment as automatically critical. If a plant runs three identical compressors but needs only two, each one's criticality drops because redundancy softens the consequence of failure. Conversely, a low-cost sensor that is the single point of failure for a multi-million-dollar line is a highly critical asset.

• Probability of failure: failure history, MTBF, age, operating environment, vibration and temperature trends.
• Consequence of failure: injuries and violations, line stoppage or available backup, scrap volume, replacement cost.
• Criticality is assigned to the asset once; priority is the dynamic attribute of each order placed on it.

Factor 2 — Cost of failure and cost of deferral

The second factor answers a business question: what does the company lose if the work is not done now but postponed for a week or a month? This is not the repair bill but the price of inaction: lost output and margin, overtime, excess energy, scrap, penalties and reputation. Comparing “do it now” versus “do it later” yields the return on investment of each job in the backlog — the larger the gap, the higher the priority.

In practice, estimate a weekly deferral cost in money for each order in the upper part of the queue. A job on a non-redundant packaging line costs an hourly stoppage rate if postponed; a lubrication job on a ventilation fan costs far less. The model automatically lifts the former above the latter.

The cost of failure also includes the full repair if neglect turns into a breakdown. A complete failure is almost always more expensive than prevention, so all else being equal you close a job on a critical asset with a high probability of failure before cosmetics on a secondary one. This is the heart of maintenance economics: directing scarce resources to where risk and loss are greatest.

Factor 3 — Repair time and logistics reality

The third factor is how long the job realistically takes: labor hours, availability of the right skill, parts lead time and the shutdown window. Mean time to repair and mean time between failures are core reliability metrics that inform both criticality and planning. A job with a long parts lead time must start early even if its “pure” risk is lower, or the failure will arrive before the part does.

This factor also captures windows: some jobs need a line stoppage and production sign-off. They cannot run tomorrow regardless of score; they must be planned into the next available window with a complete package — spares, tooling, permits. The planner assembles materials before the order reaches the crew, not after.

Logistics is why ranking by risk and money alone fails: a high score on a job that physically cannot be done this week without parts is only an illusion of control. Its right home is the next real window with a ready kit, while freed hours go to executable medium-priority work in the meantime.

Combining the three factors into one score

A practical scheme is two-stage. First run a gate: any safety, compliance, legal or breakdown work goes to the top regardless of score. Everything else is scored and sorted. The same logic appears in asset criticality programs, where a high ranking dictates a more sophisticated maintenance strategy.

For scoring, use an additive-multiplicative model. Assign the asset a criticality of 1–5. For the order, rate the probability of failure before the next window (1–5) and the consequence of failure (1–5). Multiply probability by consequence for a risk sub-score from 1 to 25. Then add a weighted cost-of-deferral term and a correction for repair time and parts lead time.

Example: a seal job on a non-redundant feed pump — criticality 5, probability 4, consequence 4 — gives a risk of 16; high weekly deferral cost and parts on hand for a two-hour job push it into the top ten. Painting a pipe rack — criticality 2, probability 1, consequence 1 — gives a risk of 2 and low deferral cost; it falls to the end of the queue but is not forgotten.

• Gate first: safety, compliance and legal work go to the top with no arithmetic.
• Risk = probability of failure × consequence of failure (1–25).
• Add weekly deferral cost in money and a repair-time/lead-time correction.
• Sort descending, then verify the top ~20% is executable with current skills and spares.

Refresh cadence, data quality and limits

The model is only as good as the data. First purge duplicates, obsolete and merged orders from the queue, and review older open orders alongside new ones so crews do not run an alignment this week and a motor repair with no alignment next week. Without decent failure history in the CMMS, probability estimates are guesses — acknowledge it and start capturing failure codes and running hours.

Recalculate priorities weekly or after a material event: an incident, a demand shift, an arrival of spares. Field evidence shows that cleaning up criticality and removing over-scheduled work can cut backlog substantially and free time for genuinely critical jobs.

Keep the limits in view. The score supports a decision; it does not replace the planner and supervisor who see site context, crew skill and agreements with production. Do not promise a fully empty queue — a small healthy backlog is normal; uncontrolled growth is the danger. The model orders work and makes the trade-off among risk, money and time explicit and defensible, but the final call stays with a person.