Spare Parts and MRO Inventory in a CMMS

In more than two decades of CMMS, CAFM and EAM work I have watched the same pattern repeat. The work order module goes live and works. The PM module goes live and works. The inventory module runs for about nine months on the data loaded at cutover, then slowly becomes fiction. Stock figures drift, technicians stop searching because search never finds anything, duplicate records multiply, and the parts that keep the site running end up in a locker rather than on a shelf. Nobody declares it a failure. It just stops being used honestly, which is worse, because the reports keep producing numbers.

The message up front: MRO inventory is not warehouse inventory with different part numbers. Demand is sporadic, volumes are tiny, the cost of a stockout on a critical spare is measured in lost production rather than lost margin, and much of what you must hold will never be issued. Any stocking policy imported from a production supply chain will be wrong here. What works is a disciplined item master, a stocking policy derived from asset criticality rather than turnover, and transaction discipline technicians actually believe in.

1. Why MRO inventory is unlike production inventory

MRO inventory breaks most of the assumptions classical inventory theory rests on, and that is the root cause of the trouble that follows. Hand the storeroom to a supply chain professional from a distribution background without explaining the differences, and they will optimise it into permanent stockout with perfectly defensible spreadsheets showing the working capital released.

• Demand is sporadic, not statistical. A production component moves a thousand units a month with a forecastable distribution. A pump seal moves twice in three years. Any reorder point calculated from two data points is numerology rather than mathematics.
• Volumes are tiny and variety is enormous. A mid-sized storeroom might hold eight thousand line items at an average on-hand of three. Production inventory is the opposite shape, and every technique built for depth fails on breadth.
• Stockout cost is asymmetric. Running out of a filter is an inconvenience. Running out of the only drive coupling for a chilled water pump serving a data hall costs you the building. The exposure has nothing to do with the value of the part, which is why ABC classification by spend misleads so badly here.
• Slow movers are not dead stock. In a warehouse, two years without movement means liquidation. In MRO it may be the most important thing on the shelf. Conflating slow movement with low importance is the most expensive mistake a well-meaning finance team can make.
• Demand is partly controllable. A large share of MRO consumption comes from your own planned work, so if you know the PM and project schedule you know much of next quarter's draw. That is a real advantage over production planning and almost nobody exploits it.
• Obsolescence is asset-driven, not market-driven. A part dies because the asset it fits was replaced or the OEM superseded the model. Both are knowable events, which makes MRO obsolescence largely preventable.

Hold those six in mind and the rest follows. For the counting and accuracy fundamentals MRO shares with other inventory domains, the inventory accuracy pillar covers them; this article assumes them rather than repeating them.

2. The item master: the discipline that decides everything

If I could fix only one thing in a struggling MRO implementation it would be the item master, because everything else is downstream of it. A storeroom with a clean catalogue and mediocre policies is recoverable. A storeroom with six records for the same bearing is not, because every number you calculate from it is wrong and nobody can tell you by how much.

The duplicate bearing explains the mechanism. A technician needs a bearing. Search finds nothing, because the record reads "BRG 6205 2RS SKF" and he typed "bearing 6205". So a new record appears: "Bearing, 6205-2RS". Months later another shift creates "SKF 6205 2RS Bearing", and procurement adds "6205 BEARING SEALED". Now four records each carry their own on-hand, reorder point and purchase history. Total on hand is eleven, which is plenty; the record the planner looks at shows two, which triggers an emergency purchase while nine bearings sit in bins nobody is watching. This is not carelessness. It is the default behaviour of any system where creating a record is easier than finding one.

The cure is structured naming plus a gatekeeper. Structured naming means noun-modifier discipline: primary noun, then modifier, then characteristics in a fixed sequence, with manufacturer and manufacturer part number as separate attributes rather than buried in a description. That principle underpins formal cataloguing conventions such as those in the NATO codification system , and you do not need the full apparatus to benefit from it. The framing above it, that spares holding is a lifecycle risk decision rather than a procurement one, sits in the ISO 55000 asset management family published by ISO . A gatekeeper means exactly one role, usually the storeroom controller, holds the right to create an item record. Technicians request; they do not create. That single permission restriction eliminates most duplication on its own.

Here is the minimum viable item master for an MRO catalogue. Less and you cannot run a stocking policy; much more and it will not be maintained.

Two fields carry disproportionate weight. The asset linkage, the parts bill of materials tying an item to the equipment it fits, turns a catalogue into a decision tool: what to stock when you commission an asset, what to dispose of when you retire one, and what an asset has genuinely cost. The criticality class lets you stock rationally rather than by spend. Most implementations I review have neither populated, which is why their reporting can only describe what is on the shelf, not whether it is the right thing. The governance around this sits in the master data management for assets pillar.

3. Criticality-based stocking policy

The default stocking policy in most deployments is inherited: somebody classified items A, B and C by annual spend and moved on. For MRO that is close to backwards. Spend-based classification concentrates attention on expensive fast movers, usually the easiest items to buy quickly, and ignores the cheap slow movers that will shut a building down.

What works is criticality of the assets served, crossed with lead time to obtain. A part is worth holding in proportion to the consequence of not having it and the time it would take to get it. That means stocking policy is downstream of asset criticality, so if you have not done that work the storeroom cannot be rationalised. See the asset criticality classification pillar for the ranking method.

The row that generates the most argument is the long-lead insurance spare, because it is pure carrying cost with no measurable turnover. Section five deals with defending it.

4. Min/max and reorder points when demand is sporadic

Every CMMS has min/max fields, and in most implementations they were populated once at data load and never touched, so the replenishment suggestions are politely ignored. That is how a replenishment engine dies: not switched off, distrusted.

The classical reorder point, average demand during lead time plus safety stock for variability, works well on parts with regular consumption and should be used on them. In a typical catalogue that means routine PM consumables: perhaps fifteen to twenty percent of line items but most of the transaction volume. For the other eighty percent, the parts with two issues in five years, there is no distribution to sample. What works instead:

• Derive demand from planned work, not history. If the PM schedule replaces forty filter sets next year, demand is forty, known in advance, no forecasting required. This is the biggest lever in MRO planning and it needs only the PM task lists to be connected to the parts they consume. It is also what makes predictive inventory planning viable in maintenance at all.
• Use criticality and lead time instead of demand. For a critical single spare the correct min is one, forever, and "how often is it used" is irrelevant. Set it by policy and record the policy reason in the item record so the next reviewer does not delete it.
• Pool demand across similar assets. One pump seal has no pattern; forty identical pumps give a population failure rate that means something. Standardising on fewer equipment models is an inventory decision as much as an engineering one.
• Use intermittent-demand forecasting where volume justifies it. Croston's method and its variants handle lumpy series far better than a moving average. Worth applying to the middle tier, not catalogue-wide.
• Set safety stock from lead-time variability, not demand variability. On long-lead parts the dominant risk is the supplier taking fourteen weeks instead of the quoted eight. Track actual receipt dates against promised and let observed variability drive the buffer.

5. Critical spares and insurance spares

The two terms get used interchangeably and should not be. A critical spare is a part whose absence stops something important; it may move often. A drive belt on a single non-redundant kitchen extract fan is a critical spare you might replace annually. Criticality is about consequence of absence, not rarity.

An insurance spare is a high-value, long-lead item held because failure without it is unacceptable, and which you may never issue in the asset's life: a spare transformer, an HV switchgear panel, a pump rotating assembly, a chiller compressor. The logic is not inventory logic but risk transfer. The correct comparison is not against carrying cost but against an extended outage plus the emergency procurement premium you would pay in a crisis.

That reframing matters because insurance spares are exactly what a well-intentioned working capital review kills. They look like dead stock in every standard report: high value, zero turnover, years on the shelf. If the only defence is "we might need it", the item goes, and eighteen months later it is needed. What holds up instead:

• Classify them separately. A distinct stock category, excluded from turnover and slow-moving reports by design, so they never reach a disposal candidate list.
• Document the risk decision in the item record. Which asset, what the failure consequence is, the replacement lead time, who approved holding it and when. One paragraph, where the next reviewer will find it.
• Review annually with engineering and finance together. That converts a recurring argument into a scheduled decision, and catches the genuine releases, usually where the covered asset has been replaced.
• Consider alternatives to holding. A contractual response-time guarantee, a spares-sharing arrangement with a nearby operator running identical plant, or a manufacturer-held unit under a service agreement can deliver similar protection without the capital.
• Maintain them. A motor stored eight years without shaft rotation or bearing lubrication, or a board stored without humidity control, may fail on installation. Stored spares need their own PM regime, and this is missed almost universally. If you have never raised a work order against a stored spare, you probably have spares that will not work.

6. Kitting, staging and reservations against work orders

The link between storeroom and work order has three stages, and most implementations configure only the last.

Reservation happens at planning. Parts added to a work order should be soft-allocated: still on the shelf, no longer available to other demand. Without it, the planner schedules Tuesday's job believing the parts are there, a corrective job consumes them on Monday, the technician stands down, and the planning exercise is discredited. Reservations are the cheapest reliability improvement available in most deployments and are frequently left unconfigured because they add a step.

Kitting happens before execution. The storeroom picks the reserved parts, puts them in a labelled container on a staging shelf, and marks the kit ready. The technician collects one kit rather than hunting six line items, and the planner can see which jobs are parts-ready. Kit readiness is one of the more useful scheduling signals a maintenance operation can have, because it turns "we think we can do this job" into a binary fact.

Issue happens at execution, and it is the transaction that updates stock and charges cost to the asset. It must be the only way parts leave the storeroom.

For any planned-work-heavy operation I would run a weekly parts-readiness review alongside the scheduling meeting: the planner presents next week, the storeroom confirms what is kitted, and anything not parts-ready comes out of the schedule rather than being sent out hopefully. That one meeting does more for schedule compliance than any configuration, and it puts honest pressure on the catalogue, because parts that cannot be found get raised as blockers instead of absorbed as friction. The work order types that drive different parts behaviour are covered in the work order types pillar.

7. Issue and return discipline

Every part that leaves a shelf without a transaction is a permanent error in the record, and errors do not cancel out. This is where MRO inventory becomes a behaviour problem rather than a software problem, and where most implementations lose the argument.

• Issue against a work order, always. Never a bare stock issue with no cost destination. If a technician cannot name the work order, the work order does not exist yet, and that is the real problem. Parts issued to nothing are cost you can never analyse.
• Return unused parts the same day. Over-picking is the commonest single source of stock error: five gaskets drawn, two used, three left in the van. Those three are recorded as consumed, so asset cost is inflated and shelf quantity understated, which eventually triggers an unnecessary purchase. A same-shift return rule with a visible return bin fixes most of it.
• Handle rotables and repairables properly. A failed motor coming back is neither scrap nor new stock. It needs a repairable status, a repair order, and a route back to serviceable stock. Treating rotables as consumables is how organisations lose track of assets worth more than the parts around them.
• Make the transaction faster than the workaround. If issuing takes four screens and a supervisor approval, technicians will take the part and tell someone later, or not. Mobile issue with barcode or QR scanning at the point of pick is the most effective single intervention, and the barcode inventory management for MRO storerooms pillar covers the mechanics.
• Design out-of-hours access deliberately. Most storerooms are unmanned at night, so either the night shift cannot get parts or there is an unrecorded route. Neither is acceptable. Choose one: controlled access with mandatory scan-to-issue, or a small emergency stock with next-morning reconciliation. If you do not design it, one will emerge and it will be invisible.

8. Cycle counting for MRO

An annual wall-to-wall stocktake is the wrong instrument here. It is disruptive, pulls people off maintenance, produces one large adjustment that explains nothing, and by the time eight thousand lines are counted the first ones have moved. Cycle counting suits MRO better, with these adjustments:

• Count by criticality, not value. Critical spares need quarterly counts regardless of cost; a low-value consumable annually is fine.
• Count on trigger events. Any issue that takes an item to or below min is a free chance to verify while someone is at the bin. Negative on-hand, a failed pick, and any adjustment above a threshold should trigger a recount of surrounding bins.
• Measure accuracy by line item, not value. Value-weighted accuracy flatters you: get the expensive items right and the number looks excellent while half the catalogue is wrong.
• Investigate causes, not variances. An adjustment fixes the number and changes nothing. The useful output is a cause: unrecorded issue, wrong unit of measure, mis-pick from an adjacent bin, duplicate record.
• Do not let counting become the accuracy strategy. Counting detects error, it does not prevent it. Steady variance means the upstream transaction discipline is the problem, and no amount of counting resolves that.

9. Obsolescence, supersession and dead stock

MRO obsolescence is mostly self-inflicted and therefore mostly preventable. Stock becomes dead for a small number of knowable reasons:

• The asset was replaced and nobody told the storeroom. The largest source. A chiller is swapped for a different model and eleven line items sit on the shelf for a decade. Asset decommissioning must trigger a spares review, which requires the asset linkage field. Without it this is undetectable.
• The OEM superseded the part. Supersession pointers prevent old stock becoming invisible and prevent both records being replenished in parallel.
• A project over-ordered. Capital projects buy spares generously and hand over a pallet, so a five-year supply enters stock with no policy behind it. Project handover should include a spares list reviewed against the stocking policy.
• Duplicate records split the usage. One record gets used, its twin never does, and the twin looks like dead stock. Deduplication often shows that a chunk of apparent obsolescence is a cataloguing artefact.
• Shelf-life expiry. Elastomers, adhesives, batteries, greases and some filters degrade in storage. These need a shelf-life field, first-expiry-first-out picking, and an expiry report somebody reads. Most storerooms hold expired seals they believe are serviceable.

Obsolete items should be written off promptly once identified, because holding them costs space, count effort and attention. The caution is the one raised earlier: be very sure the item is genuinely obsolete and not simply slow-moving and critical. The asset linkage is the test. If the record points to a live asset it is not dead stock however long it has sat there; if it points to nothing or to a retired asset it is a disposal candidate. Without the linkage you are guessing, and guesses tend to follow whoever is under pressure that quarter.

10. Vendor lead times, consignment and vendor-managed stock

Lead time determines how much risk you must carry yourself, and it is almost always recorded wrong. The value in most item masters is the quoted lead time from the last time somebody asked. What you need is observed lead time: measured purchase order release to goods receipt, over the last several orders, with its variability.

That is mostly a matter of connecting the purchasing cycle to the item record so receipt dates feed back automatically. If your CMMS handles requisition to purchase order to goods receipt the data is already there and only needs surfacing; the procurement workflow pillar covers that chain and the three-way matching pillar the control side. If purchasing lives in a separate ERP, this is one of the integration points that genuinely matters, discussed in the CMMS and ERP integration pillar.

• Consignment stock. The vendor's stock on your shelf, paid for on consumption. Good for moderately-moving items from one reliable supplier. The catch is that your consumption transactions become the billing trigger, so consignment on top of poor issue discipline creates a commercial dispute rather than a saving.
• Vendor-managed inventory for consumables. The supplier owns bin-stock replenishment for fixings, tape, lubricants and PPE. This removes a large volume of low-value transactions and lets the controller spend attention on items that matter. One of the better returns available in MRO.
• Framework agreements with committed response times. For items you would rather not hold, an enforceable delivery commitment can substitute for stock. Test it before relying on it.
• Warranty-covered components. Parts under warranty should not be replaced from your own stock at your own cost, which happens more often than anyone admits because the technician does not know the warranty status at the point of repair. Surfacing it on the work order is the fix; see the warranty management pillar.
• Documented emergency supply routes. For every critical spare you decided not to hold, record a named source, a contact and a known expedited lead time somewhere findable at three in the morning. A decision not to stock is only defensible if the alternative actually exists.

11. Closing the loop: parts usage back to asset cost history

The point of all this transaction discipline is not tidiness. Every part issued against a work order against an asset becomes a line in that asset's cost history, and that history is what lets you make lifecycle decisions with evidence instead of opinion. Once parts cost lands reliably on the asset, a set of questions becomes answerable:

• Which assets consume disproportionate parts cost? A Pareto of parts spend by asset usually surfaces a handful of units quietly absorbing budget, and they are normally candidates for replacement or a root cause investigation nobody has done.
• Is repair still cheaper than replacement? That needs cumulative maintenance cost against replacement cost. Without parts cost on the asset you are arguing from memory.
• Which failure modes consume the most parts? Consumption cross-referenced to failure codes points at recurring problems worth engineering out rather than repeatedly fixing.
• Which stocking decisions were right? Consumption against policy tells you where you are over-stocked and where you had a stockout on something you should have held. It is the only honest feedback loop on the policy.
• What does this asset class cost to run? Aggregated across a fleet, parts cost per class informs the next procurement specification, which is where maintenance data finally influences capital decisions.

None of it works if parts leave without a work order, if the work order is not linked to an asset, or if half the consumption comes out of a locker. For where inventory sits among the other modules see the core modules explained pillar, and for the wider system context the CMMS buyer's introduction.

12. The honest part: the two failure modes

Everything above is the design. What actually goes wrong reduces to two failure modes that reinforce each other.

A storeroom nobody trusts. The system says four. The technician finds one, or nothing, or four of something else. After two or three times he draws the only rational conclusion: the storeroom cannot be relied on. From then the behaviour changes permanently. Parts get over-drawn, because taking six when you need two is cheap insurance against a second trip. Parts get squirrelled away in lockers, vans and the back of the workshop, because a personal cache is more dependable than the official one. Hoarding is self-reinforcing: every part in a locker is a part missing from the shelf, which makes recorded stock less reliable, which justifies more hoarding. I have walked into workshops where informal locker stock was a significant fraction of recorded inventory value. Nobody was stealing. They were compensating rationally for a system that had stopped working.

A catalogue nobody maintains. At go-live the item master is reasonable, because somebody was paid to clean it for migration. Then maintenance stops. Nobody owns item creation, so duplicates accumulate. Nobody owns supersession, so obsolete records linger and new ones appear alongside them. Nobody reviews min/max, so replenishment suggestions drift into irrelevance and get ignored. Nobody maintains the asset linkage, so the storeroom cannot tell which parts serve which equipment. Three years on, the catalogue holds the same records plus twenty percent noise, search does not work, and the data cannot support the analysis it was bought for.

The two feed each other directly. A bad catalogue means parts cannot be found, which destroys trust, which produces hoarding and unrecorded movement, which destroys stock accuracy, which makes the catalogue less usable still. Once the loop is running it does not stop on its own, and no software upgrade interrupts it, because the problem is not in the software.

13. A recovery checklist

If you have inherited a storeroom in that state, this is the sequence I would advise, and the order matters:

• 1. Appoint a named storeroom controller with exclusive rights to create item records. Nothing else works until this exists.
• 2. Amnesty the hoards. A no-questions return period, collection bins, everything received back without blame. Punishing it guarantees it stays hidden.
• 3. Deduplicate the top of the catalogue. Not the whole master. Start with the highest-movement and highest-criticality records, merge duplicates, fix descriptions so search finds them.
• 4. Give every stocked item a real bin location and label the bins.
• 5. Make issuing faster than not issuing. Mobile or barcode issue at the point of pick. If the honest route is slower, the workaround wins.
• 6. Populate asset linkage on critical spares first, then backfill. This unlocks obsolescence control and cost history.
• 7. Reset stocking policy by criticality using the section three matrix, starting with critical assets.
• 8. Start cycle counting on critical spares and high movers, investigating causes rather than only posting adjustments.
• 9. Switch on reservations and kitting and run the weekly parts-readiness review.
• 10. Publish line-item count accuracy monthly. One visible trend, owned by the controller, is what keeps the discipline alive after the initial push fades.

Nine of those ten are organisational. Exactly one, step five, involves technology. That ratio is a fair summary of MRO inventory work in general.

The idea to walk away with

MRO inventory management is not a module you configure, it is a discipline you staff. The technical design is not hard: a structured item master with controlled vocabularies, stocking policy from criticality and lead time rather than spend, reservations and kitting against planned work, a same-day return rule, criticality-weighted cycle counting, and supersession and asset linkage kept populated. Any competent CMMS, from IBM Maximo and SAP PM at the enterprise end through Infor EAM and Hexagon EAM to MaintainX, Limble, Fiix, eMaint and UpKeep in the mid-market, can support all of that. The software is not the constraint.

The constraint is that the discipline must be owned by someone with authority and maintained indefinitely, and that is a headcount decision made by people who see the storeroom as a cost centre rather than as the thing that makes planned maintenance possible at all. Every storeroom I have seen work well had a named controller who was allowed to say no. Every storeroom I have seen fail had the function spread across people who all had more urgent things to do.

Final thoughts

This side of the CMMS fails quietly because there is no single moment of collapse. The work order module fails visibly: jobs stop getting done. The inventory module degrades gradually, and the degradation is absorbed by technicians who find workarounds, so what reaches management is not "the storeroom is broken" but a slow rise in emergency purchases, a schedule that keeps slipping for reasons nobody can name, and asset cost figures everyone privately knows are incomplete. By the time anyone traces it to the catalogue, several years of history are unusable.

If you want one diagnostic, ask a technician where they go first when they need a part. If the answer is the storeroom system, you are in reasonable shape. If it is a colleague, a locker or a van, both failure modes are running and the reported inventory numbers are decoration. That conversation takes five minutes and tells you more than any report the system will produce.

Related reading: Barcode inventory management for MRO storerooms, Inventory accuracy, Predictive inventory planning, Asset criticality classification, Procurement workflow: PR, PO, GRN, Master data management for assets.