We won't get too deep into the reasons for these rate reductions, but basically they boil down to either that 1) things were bad back then, 2) things are great right now, 3) it's a little of both, or 4) none of the above.

Nvidia, for instance, definitely falls into the first group, reducing its claims rate from 5.8% back then to only 0.33% now. For a printed circuit board manufacturer in the computer industry, 0.33% is normal and 5.8% is a clear sign of manufacturing problems.

For an automotive supplier such as Valspar Corp., however, an 0.69% claims rate is more or less normal for its industry and 0.27% is really, really good. Same story for Medtronic Inc. in the pacemaker industry. They're both having really good years, in terms of claims costs.

A Question of Timing

For most companies, it's a little of both. But then there are all the companies that fall into the fourth category. Perhaps because of the unfortunate timing of a past claims payment or the fortunate arrival of a reimbursement from a supplier, the claims rate was artificially too high or too low for reasons that had nothing to do with repair costs. One company laid off some warranty clerks, and then congratulated itself when claims fell. But then it found out why: the claims processing was all backed up because of the labor shortage.

However, that's why we're looking at one-, three-, six-, and nine-year trends. They can't always be lucky (or unlucky). A company can disconnect its phones and cut its claims rate to zero, but not for long. They can cut their warranties to 90 days or 90 minutes. In the long run, that's going to hurt more than it helps.

The bigger danger in relying on the claims rate, however, is two-fold. First, there is a lag time between when the product revenue is earned and when the repair cost is incurred. So we might be comparing one year's sales to another year's repair cost. Also, in a time of generally rising sales volumes, such as 2012, that lag time artificially depresses the recent percentages and makes the improvement look better.

Those factors are not present with accrual data. In every case, we compare current accruals to current sales, with no lag time. And since accruals are set aside as each unit is sold, the percentage rates should stay the same as sales and accruals rise and fall together.

If there is a massive decrease in accrual rates, it could mean one of four things: 1) product quality was really bad back then, 2) product quality is really good now, 3) a little of both, or 4) none of the above.

Wrong Estimates

In this case, the none of the above contingent is a little larger, because accrual rates are even more susceptible to artificial manipulation. A company's chief financial officer is free to make the absurd claim that every product made in a given quarter is perfect, and therefore accruals are not necessary. If time proves that boast to be incorrect. it can be fixed with a change of estimate. But it's an estimate, and an estimate can only be proven wrong over time.

In Figure 3, there are no 0% accrual rates, but Lear Corp. set aside only $200,000 in claims to cover $7.3 billion in sales during the first half of 2012. That's only a 0.003% accrual rate, and that's going to look good no matter what year it's compared with.

Turns out, it's by far the biggest improvement in accrual rates, whether you're looking at one-, three-, six-, or nine-year trends. But it's not because Lear has suddenly begun making nearly flawless products. It's more of an accounting anomaly that won't be repeated.