Parts Are Disappearing: A Decision Tree

Why this matters

The count came up short and your stomach dropped, because there is only one explanation an owner reaches for at that moment. Three of the four real explanations are not theft, and the commonest one by a distance is that the parts were used on jobs and nobody wrote it down. Treating a recording failure as a theft is the expensive error here: you spend a fortnight suspicious of a crew who did nothing wrong, and you still have the leak, because you fixed the wrong thing.

So this card forks on the explanation before it goes anywhere near a response. The thing that separates the branches is not how much is missing. It is the signature of what is missing.

Before you fork: is the gap real?

Do not start this until you have a count you would defend. A recount of the short lines by a second person, a check that nothing is sitting in a second location, a check that the system's expected quantity is not itself wrong because an old adjustment was entered twice. Roughly half of apparent gaps close here. The mechanics of running a count that produces a defensible number live in the inventory count SOP and the shrinkage card; this one starts after you have that number.

One more pre-check, and it saves whole investigations. Ask when the shop last changed how parts get recorded - a new tablet workflow, a move off paper slips, a new person on the counter. A gap that begins on a date is almost never a person.

The diagnostic: signature, not size

Pull the short lines and describe them three ways: what class of item, how they are spread, and what they correlate with. That description picks the branch.

Two cautions on reading that table. A count share is not a value share: a handful of short lines can be most of the loss if they are the expensive ones, and forty short consumable lines can be almost none of it. And branches are not exclusive. A real shop usually has A and B running permanently, with C sitting quietly underneath, and the question is whether D is also present.

Branch A: it was used and never written down

The default, and it is not carelessness so much as friction. A tech at height with gloves on will fit the part and record it later, and later is the end of a day with three more calls in it. The gap is largest on the parts that are fastest to fit and least memorable to log.

The fix is process, not discipline. Make the recording happen at the moment of use rather than at the end of the day, on the device already in the tech's hand, from a list short enough to be the truck's actual stock rather than the whole catalogue. Then close the loop that makes it stick: a job cannot be marked complete with parts issued to it and no parts recorded on it.

What tells you that you have picked this branch correctly: the gap stops growing within one cycle of changing the workflow, without anyone being spoken to.

Branch B: waste, damage and returns handled badly

Real material that left the building and produced nothing. Offcuts on cut-to-length stock, the fitting dropped in a crawlspace, the part opened and found wrong, the return credit nobody chased, the item that sat until it was out of date.

This one is legitimate and should be measured rather than eliminated. The number you want is a normal waste rate by material class, established from three or four clean jobs, so that next quarter's gap can be compared against something. Without that baseline every quarter's waste reads as a mystery.

The failure mode here is a returns pile. A shop that opens a part, finds it wrong and leaves it on a shelf is carrying a loss that looks exactly like shrinkage on paper. Give returns a named owner, a physical place and a weekly moment, and a share of your unexplained gap turns back into credit.

Branch C: the shop has always let people take things and never said so

The branch that needs the most care, and the one owners handle worst.

Somewhere in the history of your shop, somebody said "grab what you need" and meant a roll of tape. Nobody ever said where that ended. So a tech takes consumables for a relative's job, believing that is allowed, because as far as he knows it always was. That is a policy gap. It is not dishonesty, and treating it as dishonesty will cost you a good technician who will tell everyone, accurately, that he was accused of stealing for doing what he was told.

The tells: it is spread across several people rather than concentrated, it is consumables and offcuts rather than resaleable units, and when you ask directly you get an immediate, unbothered answer. Someone hiding a theft does not volunteer it in a corridor.

Handle it forward, never backward. Write the rule down: what may be taken, what must be asked for, who may say yes, and whether the shop sells to staff at cost. Announce it as a rule the shop never had rather than a problem the shop has. Apply it from the date it is announced, and take no action about anything before that date, because before that date there was no rule to break. Then watch the next cycle. Continued taking after a clear written rule is a different fact, and it belongs in a different conversation.

Branch D: it is being taken to be sold

The narrowest branch and the least common. Its signature is specific: resaleable items, the ones with a secondhand market and no serial-number trail, going missing in a concentrated period, correlating with a person, a truck or a route rather than spread across the shop.

What you do here is nothing visible. Do not change the locks, do not announce a count, do not have a quiet word to "let him know we are watching". Every one of those tips off a guilty person before you have anything, and destroys an innocent one on suspicion. The sequencing of detection and accusation is its own card in this group, and this is the branch that hands off to it. What matters from here is that the records you are going to need are in systems the person can edit, so they get preserved before anyone knows you are looking.

A worked case

A three-truck shop counts central stock quarterly and carries about 240 line items. This quarter, 14 lines count short. That is roughly 6 percent of line items, and it is worth saying out loud that 6 percent of line items is not 6 percent of value, because the two figures diverge sharply in a moment.

Describe the 14 by signature:

• 11 lines are consumables (fittings, fasteners, sealant, tape), each short by somewhere between 5 and 20 percent of expected quantity. Broad, small, no pattern by truck.
• 3 lines are high-value resaleable items, each short by one or two whole units.

Take the 11 first. Cross-reference usage records: on 9 of them, recorded usage thins out sharply from the week the shop moved parts logging off paper slips onto the tablet. That is branch A with a date on it, and the date is the evidence. The remaining 2 are cut-to-length material where offcut is never recorded, which is branch B and needs a waste baseline rather than an investigation.

Now the 3 high-value lines. Pull the issue records: all three were issued to truck 2, and none appears as consumed on any job in the following five weeks. Count truck 2. One of the three is physically on the truck, unrecorded, which is branch A again wearing an alarming costume. The other two are not on the truck and were never billed to anything.

So: 9 recording, 2 waste, 1 found on the truck, 2 unexplained. That is 12 of the 14 lines resolved without anyone being suspected, and it took an afternoon.

The remaining 2 are the finding. Both are from the resaleable class, both were issued in the same five-week window, both to the same truck. They are a small share of the short line count and the large share of the loss in value, which is exactly the divergence flagged above. That is a branch D signature, and the correct next move is not a conversation with the tech on truck 2. It is preservation of the issue records and the count history, then the sequencing card.

Note what the afternoon bought. Without it, the owner walks into that conversation holding "14 lines are short", which the tech can dismiss honestly, because 12 of them have nothing to do with him. With it, the owner holds two specific items, two dates and one truck, and has not said a word to anybody.

Getting the branch wrong: what each mistake costs

• Calling A or B theft: you investigate a crew who did nothing, the gap keeps growing because the workflow never changed, and the next real signal arrives into a shop that has already cried wolf.
• Calling C theft: you lose a technician who was following the only instruction he was ever given, and you keep the ambiguity that produced it.
• Calling D a process problem: you tighten the workflow, watch the consumable gap close, and conclude it is fixed - while the two resaleable units a quarter carry on, because no workflow change touches a deliberate act.
• Calling D correctly and acting immediately: the most expensive of the four. A confrontation on suspicion tips off the guilty, defames the innocent, and costs you the evidence that was sitting in a system the person still has access to.

References

• See related: The Inventory Count SOP; Catching and Preventing Inventory Shrinkage; Parts Inventory Shrinkage: Decision Tree; The Truck Stock Checklist
• See related: Before You Accuse Anyone: The Sequence, which owns the detection-to-confrontation ordering and the evidence standard for branch D
• See related: Separating Duties in a Shop With Four People, for who may issue, record and count stock