The Number You Accrue Is a Guess. ASC 606 Just Requires You to Defend It.
Every quarter, a controller at a manufacturer books a rebate accrual. It is, by design, an estimate — a number that will almost certainly turn out to be wrong by some margin. That is not a failure of the process. Under ASC 606, it is the process. The standard doesn't ask you to be right. It asks you to make a defensible estimate at the point of sale, and to update it every reporting period until the uncertainty resolves.
The problem isn't the estimate. The problem is that most manufacturers can't reconstruct how they got to it when someone asks.
Why a rebate is variable consideration, not an expense
The most common accounting error with customer rebates is treating them as an expense booked when the cash goes out. Under ASC 606, that's wrong. A customer rebate reduces the price the customer effectively pays — so it's a reduction of the transaction price, not a separate cost of doing business.
Rebates, discounts, refunds, credits, price concessions, and volume incentives are all variable consideration — the portion of the transaction price that can change based on future events. ASC 606 pulls variable consideration into Step 3 (determining the transaction price) and Step 5 (recognizing revenue), which means it affects both how much revenue you recognize and when. Before ASC 606, companies could wait until the uncertainty resolved before recording. That option is gone. You estimate at contract inception and true up each period.
Concretely: if you invoice a customer $10,000 and expect to owe them a $500 rebate, you recognize $9,500 in revenue at the point of sale and park the $500 as a liability — your best estimate of what you'll owe. When the rebate is claimed, you clear the liability against the receivable with no revenue impact. Book the liability at the sale, not when the cash settles.
The two estimation methods — and you don't get to pick the flattering one
ASC 606 gives you two ways to estimate variable consideration:
Expected value
The probability-weighted sum across all possible outcomes. Best when you have a large portfolio of similar contracts or a wide distribution of possible rebate outcomes.
Most likely amount
The single most probable outcome. Best when there's one dominant result.
The standard is explicit that you use the method that better predicts the consideration you'll ultimately be entitled to — not the one that happens to produce a lower liability. In its SEC correspondence, Philip Morris International described using the most-likely-amount method for its variable consideration, estimating the transaction price based on the amount it expected to be entitled to. The method is a judgment you have to defend, not a lever you get to pull.
The constraint — the part auditors actually push on
Then there's the constraint, and this is where the judgment gets heaviest. You may only include variable consideration in the transaction price to the extent that it is probable a significant reversal of cumulative revenue will not occur when the uncertainty resolves. In practice, "probable" is generally read as roughly 75% or more likely.
The constraint requires the most judgment precisely when the estimate sits near that probable threshold. And per KPMG's revenue recognition guidance, one of the most common errors is applying the constraint at the portfolio average level rather than the individual contract level — using a blended assumption where the standard expects a contract-specific assessment.
This is the mechanical heart of why rebate accounting is hard: it's not arithmetic, it's a chain of judgments — which method, what probability, what constraint, applied at what level of granularity — each of which an auditor can ask you to justify.
Where it actually breaks: you can't reconstruct the judgment
Here's the operational reality underneath the accounting. Manufacturers running rebate and trade-spend programs commonly hold agreements across many customers, tiers, and periods. Retroactive tiers are common — a customer who crosses a $500K threshold may earn the higher rate on their entire volume, not just the incremental purchases. Estimates get reassessed quarterly. True-ups adjust revenue in the period you learn about them.
Spreadsheets tend to break down as the number of active agreements grows — the calculation logic outpaces what a manually maintained workbook can hold reliably. And when the spreadsheet becomes the real system of record, the problem isn't that the number is wrong. It's that when the auditor asks how you got to it — which method, which constraint, which transactions rolled into which accrual — the honest answer is often a workbook nobody fully trusts.
The failure isn't the guess. The failure is not being able to show your work.
ASC 606 doesn't penalize you for a wrong estimate — it expects estimates to move. What it requires is that the judgments in assessing variable consideration be documented and disclosed.
The takeaway
A rebate accrual is a guess with a paper trail requirement attached. The standard is comfortable with the guess. It is not comfortable with a number you can't trace back to the transactions and judgments that produced it.
The teams that struggle at quarter-end aren't the ones with imprecise estimates — every estimate is imprecise. They're the ones who can produce a number but not the reasoning behind it. Reconciliation, in the end, isn't a calculation problem. It's a provability problem.