Warranty exposure: connect failure cohorts and repair capacity
Warranty activity can increase because more units are under coverage, because products are older, because first failures rise or because repairs do not solve the problem. A single claims total hides those differences. Connect sales cohorts, failure age and repair workload before concluding that product reliability improved or the service obligation is nearly finished.
Establish the covered population
The FTC’s warranty guide discusses the promises and terms that define consumer-product warranty coverage in the US. It does not provide a universal failure rate or prescribe the operating model below. Use the actual product’s coverage period, exclusions and remedy terms before deciding which units belong in the analysis.
Distinguish a unit that fails for the first time from a repeat service visit, a rejected eligibility request and a paid service contract. This task concerns ordinary product-service exposure. It is separate from an official recall campaign and from classifying warranty costs under an accounting standard.
A hypothetical age-matched comparison
Assume a fictional appliance maker offers a two-year warranty. It sells 10,000 units in January and another 10,000 in July. At year-end, the January cohort has been observed for twelve months and has 600 first-failed units. The July cohort has six months of observation and 200 first-failed units.
| Hypothetical cohort | Units sold | Observed age | First-failed units | Cumulative incidence |
|---|---|---|---|---|
| January | 10,000 | 12 months | 600 | 6% |
| July | 10,000 | 6 months | 200 | 2% |
The July figure does not prove a two-thirds improvement in reliability. It covers only half as much age. The appropriate comparison needs matched observation ages or a defensible age-specific failure model. The remaining twelve-month outcome for July is not known at this cutoff.
Also confirm that the numerator counts unique failed units and that the denominator reflects the same cohort. Counting service tickets instead of first-failed products can inflate apparent incidence without any change in the number of products affected.
Separate repair workload from failure incidence
Assume the two cohorts’ 800 first-failed units generate 160 repeat repair visits in addition to their initial visits. Total observed service work is 960 visits, or 1.2 visits per first-failed unit. The 160 repeats increase workload; they do not create 160 additional unique failed units.
Inspect what causes repetition. Parts availability, diagnosis, workmanship and customer access can all affect service completion. A repair marked closed administratively may not establish that the underlying fault was resolved. Conversely, multiple visits can be planned steps of one remedy, so repeat visits alone do not prove an ineffective repair.
Rebuild the service queue
For a separate hypothetical month in the service operation, not restricted to the two cohorts above, assume opening backlog of 90 jobs, 80 new first-repair requests and 20 repeat jobs. Four technicians each complete one compatible job per working day across 20 days, providing capacity of 80 jobs. Closing backlog is 90 + 80 + 20 − 80 = 110.
This queue grows even though the technicians work at the assumed full rate. Adding staff helps only if parts, scheduling and the required repair skills also allow completion. Use actual completed jobs, not theoretical appointments, when reconciling the observed backlog.
Review when repeat work grows faster than first failures, when unresolved jobs age, or when a young sales cohort is used to claim improved lifetime reliability. Each trigger points to a different evidence request: repair effectiveness, service capacity or matched-age failure data.
Define a bounded output
Create a cohort table with sale date, coverage, units at risk, first failures by age, repeated service and unresolved work. Keep component availability and completed remedies in a service panel. Public disclosures may aggregate several products and omit unit identity; preserve that limit rather than manufacturing an exact cohort model.
The analysis estimates operational service exposure under stated assumptions. It does not determine a warranty reserve, prove a safety defect, predict future claims or certify compliance with warranty law.
FAQ
Can fewer claims mean the product became more reliable?
Possibly, but lower sales, younger covered units or restricted access to service can also reduce observed claims. Match the exposure and observation age first.
Should every repeat repair count as another failed product?
No. Track unique units and service events separately. Repeat work matters for workload and customer outcomes, even when first-failure incidence is unchanged.