Warranty Costs Are Booked Before a Single Product Has Broken
When a product sells with a guarantee, the expected cost of honouring it is recognised immediately, before any customer complains. The estimate is management judgement and it moves reported profit.
The Matching Logic
Sell a machine with a three year guarantee and you have incurred a future obligation at the moment of sale. Waiting until repairs occur would put the revenue in one year and the cost in the next three.
Accounting requires the expected cost to be recognised when the revenue is, as a provision: a liability of uncertain timing or amount, measured as the best estimate of what settling it will cost.
The obligation is created by the sale, not by the failure. Recognising it later would let a company book profitable sales today and absorb their real cost in a period nobody connects to them.
How the Estimate Is Built
The usual method is historical: failure rates by product, average cost per claim, and expected claims across the warranty period, applied to units sold.
| Input | Source | Reliability |
|---|---|---|
| Failure rate | History on similar products | Weak for new designs |
| Cost per claim | Parts and labour history | Reasonable |
| Units under warranty | Sales records | Known |
| Claim timing | Historical curve | Reasonable |
The weakness is concentrated in the first row. A genuinely new product has no failure history, so the rate is assumed from something adjacent, and if the new design fails differently the provision is wrong from the start.
The Movement Table
The disclosure worth reading is the reconciliation of the provision balance: opening balance, additions for new sales, amounts used for actual claims, revisions to prior estimates, and closing balance.
The revisions line is the honest one. It reports how wrong last year estimate turned out to be, in the company own numbers, every year.
Consistent downward revisions mean systematic over provisioning, which suppresses profit initially and releases it later. Consistent upward revisions mean under provisioning, which flatters current profit and creates a growing catch up.
The Rate Is the Comparison
Absolute provision balances mean little as sales change. The useful figure is the provision charge as a percentage of revenue, tracked over time.
A falling rate deserves an explanation. It may be genuine quality improvement, which is a real achievement and usually discussed openly. It may also be an assumption change that lifts margin with nothing physical behind it. The two are distinguishable by whether actual claims fall in the following years.
Comparing the rate against direct competitors is informative, since firms selling similar products with similar terms should provision at broadly similar rates. A persistent gap is either a quality difference or an estimation difference, and both are worth knowing.
Recalls Are the Tail Risk
Ordinary warranty provisioning assumes failures arrive at a predictable rate. A systematic defect breaks that assumption entirely, because the whole population fails at once for the same reason.
Recall costs regularly exceed years of accumulated provision, which is why they are usually disclosed separately as a specific charge rather than absorbed. The provision was never sized for correlated failure.
Extended warranties sold separately are different again. Those are effectively insurance contracts with their own revenue recognised over the coverage period, and they should not be confused with the provision for the standard guarantee.
The Bottom Line
A warranty provision books the expected cost of future repairs against the revenue that created the obligation, which is right in principle and an estimate in execution. Read the movement table, because the revisions line reports the company own accuracy each year. Track the provision as a percentage of revenue rather than in absolute terms, and treat a falling rate as a question rather than an achievement until actual claims confirm it.