A level of repair analysis is easy to run and hard to trust. The model is arithmetic, the inputs are estimates, and the output is a procurement commitment that lasts as long as the fleet does.
| Pitfall | What it looks like | The guard |
|---|---|---|
| Comparing repairs, not capabilities | Shop labour rates argued for a week | Compare the annual cost of owning the option, fixed costs included |
| Test equipment left out | A workshop option that looks cheap | Per site, purchase plus calibration, and multiply by the number of sites |
| One site assumed | The base option costed as though there were one base | Every fixed cost at base level is multiplied by the site count |
| Pipeline spares ignored | Turnaround time treated as free | demand × turnaround ÷ 365, plus a safety level, at that item's price |
| Demand taken as certain | A single failure-rate estimate, one answer | Sweep it; report the break-even and whether the answer flips |
| Removals counted as repairs | Every arrival at the shop assumed to be a fault | Carry the no-fault-found rate: the shop is loaded by removals, the model is driven by repairs |
| Condemnation left out | Every repair assumed to succeed | Some arrivals are beyond economic repair and buy a new unit anyway; that stock has to be in the repair option |
| Marginal costs treated as absolute | A tester the base already owns, charged at full price | Charge what the decision actually adds: the marginal cost |
| Fixed costs treated as marginal | A tester charged only to its first item | Shared equipment is shared: allocate it, or evaluate the items jointly |
| Non-economics left out | The cheapest option is one the policy forbids | Screen first, and record every constraint against its item |
| Constraints never priced | A note that says depot only, and no number | Compute what the constraint costs a year; that is what relief is worth |
| Blanket policies | Everything repaired at the depot, or everything discarded | Item by item; the blanket answers lose on this example by 6 to 50 per cent |
| Indenture inheritance | Boards repaired because the unit is | Each indenture is its own decision, and usually a different one |
| Single-point answers | A winner named, with no margin | Report the margin; a two per cent difference is a coin toss |
| The horizon quietly chosen | Fifteen years in one study, twenty-five in the next | Horizon and discount rate stated on the front page |
| Discounting misused | Capital costs entered as though paid every year | Annualise one-off costs; keep recurring costs recurring |
| Disposal forgotten | Discard costed as the purchase price | Add disposal, and check whether the waste is controlled |
| The analysis run once | A study dated at contract award, never revisited | Carry the break-evens; re-check when fleet, usage, price or rate moves |
| Results that stay in the report | A recommendation nobody turned into a purchase | The output is a code on the item record and a line in a procurement plan |
| Optimising the wrong thing | The cheapest support system, on paper | Availability has a price too: a long pipeline is cheap and slow |
Four of these deserve more.
Fixed costs are the mechanism, not a detail. On the worked example, the repairs themselves are seventeen per cent of the cost of repairing at the depot. Test equipment, calibration, training, technical data and pipeline spares are the other eighty-three. Every real disagreement about a repair level is a disagreement about how those fixed costs are divided by a demand that nobody is sure of, and an analysis that presents itself as a comparison of repair costs has hidden its own argument.
Marginal against absolute is where most disputes actually live. A pneumatic rig already at the base for another system costs the new item its calibration and its fixture, not its purchase price. The same rig, if this item is the only reason to buy it, costs all of it. The two treatments can put an item at two different levels, and the honest report says which one it used and why.
A repair level is not inherited by the parts inside the item. It sounds like an inconsistency to repair a unit and discard everything in it, and it is the normal answer: repairing the unit is replacing a module. What the parent decision does fix is where the child decision happens and who has to stock the child.
The answer is a property of the support system, not of the hardware. The same valve is a discard on a fleet of six and a workshop repair on a fleet of sixty, at the same price, with the same reliability, in the same design. When somebody asks whether an item is repairable, the answer is another question: repairable where, for how many, and how often.