Under-provisioned warranty is a quiet, repeatable EBITDA overstatement in product businesses. Here is how to test provision adequacy against a real claims trend.
Every time a business sells a physical product, it books revenue today and takes on an invisible obligation for tomorrow. Some fraction of what just went out the door will come back - broken, faulty, or simply returned under a warranty promise - and the cost of honouring that promise is as real as the cost of making the product. The accounting question is deceptively simple: has the business set aside money for that future cost at the point of sale, or is it just paying the bills as they land? The answer separates a clean earnings picture from a flattering one, and testing it is among the more mechanical but genuinely consequential jobs in a product-business FDD.
Accrual accounting has a firm view here. A business that sells a product carrying a warranty is supposed to recognise the expected cost of future claims at the moment of sale, not when each claim eventually arrives. It does this by booking a warranty provision - a liability on the balance sheet with a matching charge in the P&L - so that the cost of standing behind the product lands in the same period as the revenue that product earned. That is the matching principle doing exactly what it exists to do.
Now picture a business that ignores this and simply expenses warranty costs as claims come in - pure cash accounting dressed up as accrual. In any period where claims happen to run below their true long-run rate, reported costs are artificially low and EBITDA is overstated. In any period where claims spike, costs jump and profit is understated. The trend an analyst is trying to read cleanly is now distorted by the timing mismatch between when products were sold and when they happened to fail.
The insidious part is that under-provisioning does not announce itself. It is not a dramatic one-off or a restated number. It is a quiet, period-after-period sliver of profit that was never really there - the kind of overstatement that survives a casual read of the accounts precisely because nothing looks wrong.
This sits in the same conceptual family as the wider treatment of provisions in FDD, but warranty deserves its own scrutiny in any hardware, equipment, automotive, or consumer-product target - and it is exactly the sort of judgemental accounting a proper quality of earnings review is built to interrogate rather than accept.
The central test is simple in concept and disciplined in execution. Compare the warranty provision balance against a reasonable estimate of expected future claims, built from the business's own history: the claim rate - claims as a percentage of units sold or of revenue - applied forward to the population of products still within their warranty window.
If the provision the business carries is materially below what its own historical claim rate implies, you have a candidate EBITDA adjustment: the reported profit was flattered by the shortfall in provisioning, and normalising it reduces the earnings a buyer should pay for. This is core EBITDA adjustments territory - a genuine sustainability question about whether reported profit reflects the real cost of doing business.
Three things deserve specific checking:
One subtlety worth carrying: the right claim rate is not always the simple historical average. Products fail on a curve - some fault early ("infant mortality"), most fail late in life, and a warranty that has just been extended from one year to two changes the exposure profile entirely. A business that recently lengthened its warranty term, launched a new product line with no failure history, or changed a key component supplier has a forward claim rate that its own backward-looking average will understate. Ask what has changed in the product, the term, and the supply chain before you trust the historical rate as a forward estimate - the assumptions matter as much as the arithmetic, which is why this is judgement work rather than a lookup.
Numbers make it concrete. Take a consumer-electronics manufacturer. Its own claims history shows warranty claims running at 4% of revenue on a two-year warranty. Revenue has grown sharply, but the provision has barely moved because nobody re-based the assumption. Here is the gap.
| Item | Value |
|---|---|
| Current-year revenue | €60,000,000 |
| Historical warranty claim rate | 4% of revenue |
| Implied annual warranty cost | €2,400,000 |
| Warranty provision actually charged this year | €1,500,000 |
| Annual EBITDA overstatement | €900,000 |
That €0.9m is not a one-off. If the under-provisioning has run across the whole review period, each year's EBITDA was flattered by a similar amount, and the normalised earnings a buyer should underwrite are lower than the reported figure across the board. The knock-on is mechanical: on a business valued at, say, 8x EBITDA, a €0.9m recurring overstatement is worth roughly €7.2m of enterprise value - a number no buyer wants to find after signing. And separately, the cumulative under-provision on the balance sheet is a shortfall in a liability, which raises the question of whether it should also be treated as a debt-like item in the net debt bridge, not only as an earnings adjustment.
Warranty and product returns are cousins, not twins, and conflating them muddies the analysis.
Product returns - customers sending goods back within a returns window, often for a refund - are primarily a revenue question. Returns should be netted against gross revenue, and a returns provision or refund liability should be carried for goods sold but not yet past their return window. Under-provisioning here overstates net revenue rather than understating cost, but the effect on reported profit runs the same direction. The two interact in retail and e-commerce especially, where a generous returns policy is a commercial feature with a real accounting tail. Where returns are large enough to distort the top line, they sit alongside your revenue quality work, because a revenue figure that ignores expected returns is not a clean number.
| Dimension | Warranty provision | Returns provision |
|---|---|---|
| Primary statement affected | P&L cost / EBITDA | Revenue (net) |
| Trigger | Product fails or faults in warranty period | Customer returns within returns window |
| Built from | Historical claim rate x in-warranty population | Historical return rate x recent sales |
| Under-provisioning distorts | Cost, understated | Revenue, overstated |
| Where it flows in analysis | EBITDA adjustment, possibly net debt | Revenue quality, net working capital |
Keeping them in separate columns of your model prevents the classic error of testing one and assuming the other is fine.
Beyond the routine flow of claims sits genuine tail risk: a product recall. A systemic defect can force a business to proactively replace or repair a large volume of units at a cost far exceeding anything routine warranty experience would predict. Recalls do not follow the smooth statistical pattern that makes ordinary warranty provisioning tractable - they are lumpy, occasionally enormous, and sometimes regulator-mandated.
This is a contingent liability question rather than a routine-provision one, and it earns specific attention wherever there is a history of quality issues, ongoing regulatory scrutiny of product safety, or concentration in a single product line where one defect hits a large installed base. It belongs in the red flags section when the signals are present, and it is worth asking directly whether any live safety investigation or field action is running that the routine provision would never capture.
A warranty finding rarely stays in one box, and part of doing the work well is routing each consequence to the right place in the analysis.
The recurring under-provision is an EBITDA adjustment - it reduces the sustainable earnings the valuation multiple is applied to. The cumulative balance-sheet shortfall may be a debt-like item, pulled into the net debt schedule and the equity bridge. And the go-forward provisioning policy feeds the completion mechanism: whether the deal settles on locked box or completion accounts, the treatment of warranty and returns provisions needs to be pinned down so the buyer is not surprised by a provisioning true-up after the fact. A vague finding helps nobody draft any of that; a quantified one - this claim rate, this population, this shortfall - gives the deal team something real to price and paper.
Warranty is a favourite because it tests whether a candidate understands accruals and can connect a mechanical check to a valuation consequence. A strong answer runs like this:
"In any product business I'd want to know whether warranty costs are properly provisioned or just expensed as claims come in. Under accrual accounting you should recognise the expected cost of future warranty claims at the point of sale, matching it to the revenue. If a business instead just pays claims as they arrive, EBITDA gets overstated in any period where claims run below the true long-run rate - and it's a quiet overstatement, period after period, not a dramatic one-off, which is what makes it easy to miss. The test is to compare the provision balance against the business's own historical claim rate - claims as a percentage of units or revenue - applied to the products still in warranty. I'd flag three things specifically: a provision that's flat while sales have grown, a history of provision releases that might be smoothing earnings, and whether the actual claims trend supports the assumptions. If I found a real shortfall, I'd treat the recurring piece as an EBITDA adjustment and consider whether the cumulative balance-sheet gap is a debt-like item for the net debt bridge. And separately from warranty, I'd check product returns, because that's a revenue question - returns should be provisioned and netted off the top line. On a business trading at eight times EBITDA, a €900k annual overstatement is over €7m of value, so it's not a footnote."
That answer wins because it states the accounting principle, gives a concrete test, names the specific red flags, and closes on the valuation number - the arc an interviewer is listening for.
An under-provisioned warranty balance is the archetypal quiet overstatement - no fireworks, no restatement, just historical profit that was reliably a little better than reality, period after period. It hides behind the very ordinariness of the number. The work that catches it is unglamorous: pull the claims data, derive the real rate, apply it forward, and compare it to what the business actually booked. Do that, keep warranty and returns in separate columns, and stay alert to recall tail risk, and you convert a balance-sheet line most people skim into a quantified adjustment that moves price. That is precisely the difference between an analyst working off the trial balance and one who actually reads the business.
The Transaction Services Interview Programme (€119.99, one-time) includes a full provisions module covering warranty adequacy testing, the claim-rate build, distinguishing warranty from returns, recall tail risk, and how each finding flows into the EBITDA adjustment, net debt, and completion mechanism. Enrol today.
Hundreds of candidates prepared their interviews with this programme. Those who landed the role have one thing in common: they worked the cases before walking into the room.
