Ask an integrator why they bother writing a recommended spares list at all, and the honest answer is not efficiency. It is so that eighteen months later, when a line is down and a component nobody stocked has a fourteen-week lead time, the customer cannot turn around and ask: "Why didn't you tell us this?"
That is risk transfer, not housekeeping. And it is the reason the parts record matters more to the firm that assembled the system than to anyone who supplied a part for it.
The usual way to draw this line is ownership: an OEM owns its bill of materials, an integrator does not. That line is legal, not operational. The better question is how much of the delivered system came from somebody else, and across how many companies.
That share is measurable. For motor vehicle body and trailer manufacturing, where work-truck upfitting sits, the US Bureau of Labor Statistics puts purchased inputs at 0.797 of output value in 2023, up from 0.710 in 1987 (BLS Industry Productivity, via FRED). It is an industry figure and an upper bound, since it includes purchased services. Across equipment industries the share runs from roughly 0.52 to 0.80 with no gap. It is a level, not a category.
So the useful test is whether most of what you delivered came from elsewhere, and whether you answer the phone when it fails. Plenty of integrators and machine builders that own a bill of materials pass both.
Here is the structural oddity. Read what the industry says a complete machine documentation package should contain and you find controlled drawings, CAD models, schematics, assembly procedures, inspection records, software versions, maintenance guides. Read the canonical control-systems reference and you find taglists, loop sheets, P&IDs, field wiring diagrams, junction box drawings.
Neither names the purchased-component parts list. The turnover package has a slot for almost everything except the record of what is in the machine and who made it.
So it lives wherever it landed. Supplier PDFs in a project folder. A vendor's application engineer's email in one person's inbox. Redlines from commissioning that may or may not have been folded back in. And the largest repository of all: the engineer who was on the install and remembers what actually went in, as opposed to what the drawing said would.
That is not negligence. It is what happens when fifteen suppliers each publish for their own customers and nobody publishes for the firm that has to combine them.
Numbers in this part of the industry are less trustworthy than most, so here is only what is sourced.
Parts availability breaks first visits. In the Service Council's Service Parts Management study, one in five service leaders reported that more than 30% of unsuccessful field service visits were caused by the unavailability of the appropriate part, and 43% named poor visibility into available parts as their biggest challenge (Service Council, June 2023; no sample size published).
Identification failure has been measured once, rigorously. A Ricardo study for the European automotive suppliers' and aftermarket distributors' associations found 85% of respondents had experienced multiple or wrong deliveries caused by inadequate identification. Around 10% of parts in incoming orders were not unequivocally identifiable, and the average cost of running the business rose 11% (CLEPA and FIGIEFA, December 2017). It is automotive, European and trade-association-commissioned, so the number won't transfer to your shop. The mechanism will.
The rest you already know. The median wage for industrial machinery mechanics and millwrights was $30.82 an hour in May 2025 (US Bureau of Labor Statistics); your loaded rate is higher. Multiply it by hours spent working out what a part is rather than replacing it, add return freight on the wrong part and the second visit, and the number is built from your operation rather than somebody's marketing page.
Parts software as a category assumes one manufacturer publishing its own bill of materials to its own dealer network. For a firm at 0.797 purchased content, every part of that assumption fails: there is no single manufacturer, the parts list references other companies' numbers, and there is often no dealer network because the integrator services direct.
You can force it, inventing an internal number for every purchased item and maintaining the cross reference by hand. But you have then created a second documentation job to manage the first.
For one delivered system:
None of this makes fifteen suppliers publish in one format. It moves the work from the moment the customer calls to the moment the system ships, when it is cheap and the engineer who did the install is standing there.
Not "should we document better." Ask: when a customer sends a photograph of a failed component on a system we delivered three years ago, how many places does someone have to look, and how many of them are a person's memory?
If any of them is a person, you already own the phone call. What is left is what it costs to answer it, and whether you can show you told them.
Interactive Parts Schematics (IPS) builds one parts structure per delivered system from supplier PDFs, photographs and parts lists, with each component's manufacturer and number kept attached. How many places would your team look for the last part a customer asked you about?