The Rebate Season That Cost More Than It Returned

Why this matters

Most rebate programs fail a shop quietly. There is no incident, no angry call that names the problem, no single claim you can point at. The office is simply busier than it should be, a handful of customers are annoyed about money that never showed up, and something unrelated slips. By the time anyone asks the question, the season is over and the only way to answer it is backwards, out of whatever got written down. This is a reconstruction of one such quarter, and the useful part is not the diagnosis. It is which findings the record could support and which ones it could not, because the gaps turned out to be the finding.

The signal

A three-truck shop running a busy replacement quarter, 13 weeks. The owner's complaint was not about rebates at all. Two things had happened together: the office coordinator, who also ran collections, said she was underwater in a quarter that was not their busiest by job count, and receivables aged noticeably. Rebates came up only because when the owner asked what she was spending time on, the answer was "chasing rebate stuff," with no number attached to it.

Nobody had been tracking rebate handling time, so there was no report to open. What existed was a submission spreadsheet, the job records, the customer email history, and the receivables aging. The reconstruction ran off those four.

What the record could tell us

The spreadsheet was thin but it had rows, and rows are enough for counts.

Sixty-two claims submitted in the quarter, across three programs: 34 on the first, 19 on the second, 9 on the third. Seventeen of the 62 came back rejected, about 27%. Of those 17, twelve were refiled and five were never touched again. That last number was the first thing that stopped the room. Five customers had been told a claim was in flight, and nothing further ever happened on those claims, in either the spreadsheet or the email history.

Final outcome across the quarter, reconstructed by matching payments and customer emails to rows: 51 of 62 claims paid, 11 did not. Of the 11, five were the never-refiled rejections, five were refiles that came back rejected a second time and stalled, and one was a claim that had been submitted, was never rejected, and simply could not be located anywhere.

What the record could not tell us, and why that was the finding

Confirmation references were present on 41 of the 62 rows, about two thirds. Twenty-one rows had none.

That single missing field is what made most of the rest of the quarter unknowable, and it explains the one claim nobody could locate. Without a confirmation reference there is no way to distinguish a claim in review from a claim that never arrived, no way to chase one, and no way to prove to a customer or a program that you filed anything. The coordinator had not been careless about it in the way it sounds; she filled the field whenever the portal displayed a reference on screen, and skipped it on the two programs where the reference arrived later by email into a mailbox she checked irregularly.

The second gap was that rejections were re-filed in place. When a claim came back and was corrected, the row was updated rather than appended, so the rejection left no trace. The 17 rejections were only recoverable at all because the rejection notices themselves sat in the shared email. Had those been deleted, the record would have shown a quarter with 62 submissions and 51 payments and no visible rejection problem whatsoever, which is exactly what the owner believed going in.

Neither gap was a failure of diligence. Both were fields that nothing forced, in a spreadsheet nobody reviewed, maintained by the only person who knew what any of it meant.

Reconstructing the handling hours

The planning assumption, when the shop took on the second and third programs, had been about 1.0 office hour per claim. Nobody had ever tested it.

Rebuilding an actual number took the coordinator's calendar, the portal submission timestamps, and the outbound email log. Sessions were identifiable: a block of submissions on a given afternoon, a run of status emails, a cluster of customer replies. Counting session time against claims touched came out around 2.8 office hours per claim including refiling and customer status traffic, so about 2.8 times the assumption.

Across 62 claims that is roughly 174 office hours in the quarter. Over 13 weeks that is about 13.4 hours a week, roughly a third of a 40-hour week, spent by one person on work that had never appeared in anyone's plan or in any job's cost.

Treat that figure as an estimate with a real error bar, because it was reconstructed rather than measured. It is not precise and it did not need to be. The assumption was 1.0 and the reality was somewhere close to three times that, and no plausible refinement of the method brings those back together.

One program produced just over half the rejections

Splitting the 17 rejections by program is where the quarter's largest single defect surfaced. Nine of the 17, just over half, came from the second program, which had carried 19 of the 62 claims, just under a third of the volume. Within that program's own claims the rejection rate was 9 of 19, about 47%, against 8 of 43, about 19%, across the other two combined.

The cause was in the rejection notices, which all said a version of the same thing: equipment not qualifying under current program terms. The program had revised its qualifying tier effective the start of week 6 of the season. The shop found out in week 13, from the notices.

Every claim submitted for work completed after that effective date was filed against terms that no longer existed. Nobody was checking, because the eligibility check happened at the estimate, sometimes weeks before install, and nothing in the process re-read the terms at submission. The loss scaled with submission rate, which is the property that makes a mid-season terms change the most expensive single event in rebate administration: a shop that files more claims loses more, faster, and the feedback arrives an entire processing cycle late.

Worth being exact about the blame here. Nine claims were lost to a rule change the shop could not control. They were lost for seven weeks longer than necessary because of a check the shop could have run in two minutes per claim.

The claims nobody could prove were submitted

The five never-refiled rejections all shared a shape. Each arrived as a notice, was read, was judged to need something the coordinator did not have immediately to hand, and was set aside. None had a cure end date written anywhere. By the time the reconstruction found them, every window had closed.

The customer side of that was worse than the claim side. Reviewing the email history, three of those five customers had followed up at least once and received a reply saying the claim was still processing, which was true when it was written and false within weeks. Nobody had lied to anyone. The information simply stopped being accurate and no process refreshed it.

What it cost on the other side of the desk

The coordinator owned collections as well as rebates. Over the same quarter, receivables aged past 60 days moved from 11% to 19% of open receivables, an 8 percentage point rise, in a quarter with no unusual customer mix and no change to terms.

That is a correlation reconstructed after the fact, not a proven cause, and it should be read that way. What is not in doubt is the direction of the constraint: roughly 13.4 hours a week of unplanned work landed on the one person whose other job was getting money in the door, and the money got slower.

This is where the season's title comes from. Against that cost, the return was 51 paid claims of money that went to customers rather than to the shop, plus a close-rate lift and a goodwill effect that nobody had measured and nobody could produce a number for. The shop could demonstrate the cost precisely and could not demonstrate the return at all. That asymmetry, rather than any single failed claim, is the actual verdict on the quarter.

What changed, and what the next quarter measured

Five changes, each aimed at a specific finding above rather than at rebates in general.

The third program, 9 claims for a small incentive, moved to refer-only. The shop still tells those customers the program exists and hands them the documents, and no longer files. The second program was suspended pending a fresh read of its revised terms and a decision about whether it clears the shop's handling ceiling at all.

The confirmation reference became a required field: a row without one is treated as not submitted, and the claim is either resubmitted or escalated the same week. Rejections stopped being edited in place and started being logged as their own event with a cure end date entered on the day the notice arrived. And a single line was added to the submission routine: re-read the program's terms on the day of submission and record the version date, which is the two-minute check that would have caught the tier change in week 6 instead of week 13.

The following quarter, on one program: 38 claims, 3 rejections, about 8%, against about 19% for those same two programs the quarter before. Compare like with like: the 27% all-programs figure included the suspended program's 47%, and benchmarking the survivor against a blend that contained the program you removed overstates your own improvement. Measured handling time, this time actually recorded rather than reconstructed, ran about 1.6 office hours per claim, so about 61 hours across the quarter, roughly 4.7 hours a week. One of those 3 rejections was a premises already claimed on a home that had recently changed hands, which no process on the shop's side prevents, and it is counted here rather than quietly excluded because a rejection rate you have curated is not a rate.

The lesson the owner took was not that rebate programs are a trap. It was that they had been running a standing administrative line with no measured unit cost, no required fields, and no trigger for a terms change, and that any of the three would have surfaced the problem within weeks instead of at season end.

References

• The current published terms, effective dates and cure windows of the specific utility, manufacturer or state program, which govern every clock described here
• See related: How to Decide Which Rebate Programs Are Worth Running; The Rebate Submission SOP; How to Track a Submission You Do Not Control; How to Handle a Rejected Rebate Claim