Every service business tracks the money customers owe it. Almost none of them track the money their suppliers owe them. That second pile is real, it is usually four figures a year for a small shop and considerably more for a fleet, and it walks out the door one failed part at a time.
The pattern is always the same. A capacitor fails at eleven months. An opener logic board arrives dead and you swap it on the spot from truck stock. A key fob will not take a program no matter what you do. A circulator pump quits ninety days after install. In every one of those cases the part is almost certainly covered by the supplier or the manufacturer — and the labor to go back out almost certainly is not. The credit exists. It just never gets claimed, because six weeks later nobody can say which vendor sold you that part, on which invoice, or when.
As of September 2026 this remains one of the least automated corners of field-service operations. Most shops have a system for what goes out and no system at all for what comes back. This is a walkthrough of how to build the second one: linking a part to the purchase that bought it and the sale that installed it, handling cores, opening a return authorization and driving it to a credit memo, reconciling what the supplier actually gave you against what you were owed, and calling a defect honestly when the failure was your own install.
The credit is small; the leak is not
Take a single failed run capacitor at $14. Nobody files an RMA for $14, and reasonably so — the paperwork costs more than the part. But that is not the number to look at. Look at the year: a two-truck HVAC shop that replaces one covered part a week at an average of $60 is leaving roughly $3,000 on the table annually. A garage-door outfit replacing logic boards and operators recovers considerably more per unit. An automotive locksmith who gets a bad batch of aftermarket fobs can be sitting on several hundred dollars of unclaimable parts in a single month.
None of that shows up as a loss anywhere in your books. It shows up as cost of goods sold on a job you did for free, which is exactly why it stays invisible. The SBA's guidance on managing business finances makes the general point that you have to track income and expenses continuously to know where you stand — and a warranty part you paid for twice is an expense that quietly doubles without ever announcing itself.
The right mental model is that a covered failure has two costs, and you can only control one of them. The labor is gone; you are going back out regardless, and you should not be charging the customer for it. What you can refuse to absorb is the second purchase of a part you already bought once.
The whole problem in one sentence: which invoice bought this part?
Everything else in this article is downstream of a single question. When a technician holds a failed component in the driveway, can your system tell you where it came from?
Answering that requires two links most operations never make:
The purchase link. The part has to be tied to the supplier and the purchase order or invoice that brought it in — vendor name, their invoice number, the date, the unit cost, and the serial or lot number if it carries one. If you already run purchase orders with reorder alerts, you have most of this; what is usually missing is that the receipt of goods never gets connected to the individual units that go on a truck.
The sale link. The same unit has to be tied to the customer, the job, the date installed, and the technician who installed it. This is the part most shops do have, because it lives on the invoice — but only if the invoice recorded the specific part rather than a generic line that says board.
Connect those two and a claim becomes a lookup. A customer calls about an opener that quit; you search their name or address, see the exact board with its serial, and one click away is the vendor, the PO, and the purchase date that determines whether you are still inside the window. Without the connection, you are scrolling through eight months of supplier statements guessing, and the honest answer is that you will not bother.
Serial-level tracking is also what makes ordinary warranty tracking for a service business work in the customer-facing direction. The same record that lets you tell a customer their install is still covered is the record that lets you tell a supplier the same thing about the part inside it.
Cores and returns: a separate pile of money
Before the defect claims, clear the easier money. Core charges are refundable deposits suppliers add on rebuildable units — compressors, starters, some motors, certain control boards. There is no defect argument to make and no approval to win. You return the old unit inside the window and the deposit comes back.
Cores get lost in exactly two ways. The old unit rides around in the truck bed until it is unrecognizable and past the return window, or the credit comes back on a statement and nobody checks that it matches what was owed. Both are tracking failures, not disputes.
Treat a core as inventory the moment the tech pulls the old unit. It has a value, a location, and a deadline. Handle plain overstock returns the same way — the box you over-ordered and never opened is cash sitting on a shelf, and the return window on it is shorter than most people assume. The mechanics are close cousins of customer-side refunds and returns in a service business: the same discipline of a documented reason code, a dated record, and a reconciliation against the money that actually came back.
From RMA to credit memo, without a shoebox
A workable return authorization process has five steps and no ambiguity about who owns each one.
- Capture at the job. The tech records the failed part, its serial, the failure mode in plain language, and a photo or two, attached to the job record while standing in front of it. Anything reconstructed at the shop three days later is a guess.
- Decide and file. Someone in the office looks up the purchase, confirms the part is inside the vendor's window, and requests the return authorization. Record the RMA number the vendor gives you against the original part record — that number is the only thing linking your paperwork to theirs.
- Ship and log. The failed unit goes back with a tracking number, recorded on the same record. A defective part sitting in a bin behind the counter is a claim that will expire.
- Track the pending credit. The claim is now a receivable. It needs an owner and a date, exactly like an unpaid customer invoice, because a claim nobody follows up on gets quietly dropped by both sides.
- Match the credit memo. When the supplier's credit lands, match it to the specific claim and close it. This is the step that gets skipped, and skipping it is how partial credits go unnoticed for a year.
The IRS's recordkeeping guidance for small businesses is clear that electronic records satisfy the same requirement as paper — which matters here because vendor credits reduce your cost of goods and you want the supporting documentation searchable rather than stapled inside a folder.
Here is how the three common approaches actually compare:
| No tracking | Shoebox or spreadsheet RMA log | Serial-linked purchase and sale records | |
|---|---|---|---|
| Credits recovered | Only the large, obvious failures — most small parts never claimed | Whatever someone remembered to write down that week | Every covered failure, because the vendor and purchase date are one lookup away |
| Finding the source invoice | Manual scroll through supplier statements, often abandoned | Depends on the note being complete and legible | Instant from the customer, job, or serial number |
| Staff time per claim | 30-60 minutes of reconstruction, so most are skipped | 15-20 minutes plus chasing the tech for details | 2-3 minutes; the data was captured at the job |
| Pending claims followed up | Not visible anywhere | Only if someone rereads the sheet | Tracked like a receivable with an owner and a date |
| Credit memo reconciliation | None — the statement is taken at face value | Rough, by memory | Line-by-line against what each claim was worth |
| Callback cost visibility | Invisible; buried in job costs | Anecdotal | Reportable by part, by vendor, and by technician |
Vendor defect or your install? Call it honestly
Not every failure is the supplier's fault, and a shop that files everything as a defect burns its credibility with the one vendor rep who can approve a borderline claim next quarter.
The two useful signals are failure mode and time to failure. Dead on arrival, or several units from one batch failing the same way, is a vendor problem and you should say so firmly. A part that ran fine for months and then failed in a way consistent with a loose connection, a wrong-sized replacement, contamination, or over-torqued hardware is usually yours — and finding that out is worth more than the credit, because it points at a training or process gap that is generating repeat visits.
That is the same signal your callback and rework rate is supposed to surface. Tagging each callback with a cause — vendor defect, install error, or customer-caused — turns a pile of annoying return trips into two separate management problems: one you take to your supplier, and one you handle in your own shop. Without the tag they blur together and you address neither. Reviewing that split by technician, alongside the job costing numbers for true profit per job, tends to make the pattern obvious within a quarter.
What the customer pays, and what you stop absorbing
The customer pays nothing. A part you supplied failed inside its warranty and you are going back out to make it right — charging for that visit converts a recoverable cost into a lost customer and a public review.
What changes with tracking is the internal accounting. The visit gets written as a zero-dollar warranty job with the parts and labor still recorded at cost, so it lands in your reports instead of vanishing. The replacement part comes out of stock properly rather than being quietly untracked, which keeps your counts honest and your parts markup and margin numbers meaningful. And the failed unit becomes an open claim with a dollar value attached.
Do that for a quarter and you can answer questions you previously could not: which supplier's parts fail most often, what warranty callbacks cost you in labor hours, and whether the cheaper aftermarket line is actually cheaper once return trips are priced in. That last one reverses buying decisions more often than people expect — a part that saves $22 and fails at three times the rate is a bad deal the moment you count the truck roll. It also cleans up your stock accuracy, since untracked warranty swaps are a quiet contributor to inventory shrinkage between stock counts.
Reconcile the statement, not just the claim
The last step is the one that separates a shop that files claims from a shop that gets paid on them. Supplier credits arrive on a statement, in a batch, often weeks apart from the claim and sometimes for less than the full amount — a restocking fee, a prorated credit on a partially used item, a core deposit netted against something else.
Reconcile line by line against the claims you opened. Three things routinely turn up: claims credited at less than expected, claims approved verbally and never issued, and duplicate credits you should flag before the vendor finds them. Per Intuit's small-business cash-flow research, the operators who stay solvent are the ones watching money in both directions continuously rather than at year end — and vendor credits are money in, arriving on someone else's schedule.
A short recurring habit is enough: once a month, pull the open claims list and the supplier statements side by side and close out everything that matched. Anything still open after sixty days gets one email. Anything still open after ninety gets written off deliberately, so at least you know what the number is.
The bottom line
Vendor parts credits are not a big program. They are a small habit that depends entirely on one piece of infrastructure: knowing which supplier invoice bought the part that just failed, and which job installed it. Build that link and claims take three minutes; skip it and every credit under a hundred dollars is effectively uncollectable.
Start where the money is. Cores and unopened overstock first, because those need no argument. Then serials on anything that carries one — boards, motors, compressors, fobs, openers. Then the monthly reconciliation, which is the step that tells you whether any of it is working. The parts will keep failing either way. The only question is whether you buy each one once or twice.
Related reading: Purchase orders and parts reorder alerts · Warranty tracking for service businesses · Cutting your callback and rework rate. For a complete machine-readable feature and pricing reference, see our LLM reference page.