Warranty timing
A Socratic walk-through of warranty timing — reasoned out one step at a time, not lectured.
The question we started with
THE QUESTION #Why do appliances seem to fail just after the warranty runs out?
The washing machine dies at twenty-six months. The warranty was twenty-four. Everyone has this story, and the explanation everyone reaches for is that the manufacturer built it to happen — a part chosen, or a counter set, to give out just past the date the company stops paying.
It is a satisfying theory and it deserves to be tested rather than repeated. So: what evidence would distinguish a machine deliberately timed to fail from a machine that simply wore out, in a world where warranties were placed where they were placed for other reasons? If the two are hard to tell apart from the sofa, that itself is worth knowing.
Reasoning it through
REASONING #Start with the shape of failure over an object's life, because it is not flat. Reliability engineers draw it as a bathtub. Early on the failure rate is relatively high and falling: units with a bad solder joint, a mis-seated seal, a defective bearing die in the first weeks. Then comes a long stretch where the rate is low and roughly constant — failures here are essentially random, a power surge or a swallowed hairgrip. Then the rate climbs again as bearings, pumps, seals and heating elements reach the end of their fatigue lives. That final climb is the wear-out phase, and it is not a plot, it is what materials do.
Now ask where a manufacturer would put a warranty. Warranty is a financial instrument: the company is insuring you, and it prices the insurance from its own reliability data, which it has in enormous quantity. It wants the covered window to sit where claims are cheap and predictable — comfortably past the early-failure hump, comfortably before the wear-out climb. So the warranty is drawn from the failure distribution.
That inverts the causal arrow entirely. The warranty ends before failures rise because it was set to end before failures rise. It is a readout of the failure curve, not a cause of it. A manufacturer with a genuinely more durable product can afford a longer warranty and usually advertises one — which is why warranty length is a reasonable, if noisy, signal of expected life.
Then a second thing is happening, in you rather than in the machine. Consider all the failures a machine can have. Warranty periods are short relative to expected lifespan, so nearly every failure that will ever occur occurs after the warranty. Failures at month twenty-six, month forty, and month ninety are all "after the warranty" — but only the first feels like it. "Just after" is an elastic phrase with no defined edge, and it stretches to fit whatever happened.
And the two kinds of event are not equally memorable. A fault inside the warranty is fixed free and forgotten within a week; a fault outside it costs you money, or the machine, and generates a grievance you retell for years. So the remembered sample is not the sample of failures — it is the sample of expensive failures, and every one of those is post-warranty by construction. The machine that ran eleven quiet years produced no story at all.
Does this mean nothing is engineered? No — and here the honest answer is more interesting than either extreme.
The analogy
THE ANALOGY #Think of a shop that offers to replace any umbrella that breaks within thirty days, having watched thousands of its own umbrellas and knowing that the ones with faulty ribs snap almost immediately while the rest last a couple of years. The thirty days is chosen to catch the faulty ribs cheaply. Customers whose umbrella dies in month fourteen are certain the guarantee expired just before it broke — but the guarantee was drawn around a pattern that already existed.
An umbrella has one failure mode and an appliance has dozens, wearing at different rates, so a real machine has no single expiry to aim at — and the analogy leaves out the shop's genuine ability to make the ribs cheaper, which is where the real design decisions live.
Clarifying the model
THE MODEL #Three refinements, and one of them is a concession.
First, the concession. Manufacturers absolutely make durability decisions, and they make them with failure data in hand. Choosing a plastic gear over a metal one, a sealed non-replaceable bearing, or a cheaper capacitor is a decision about expected life taken with full knowledge of what it costs. That is a design-to-cost choice with a foreseeable durability consequence — meaningfully different from a mechanism timed to fail, but not innocent either, and it produces much of what people are actually complaining about.
Second, planned obsolescence is real, just usually not in this form. The clearest documented case of deliberate life-shortening is the Phoebus cartel of the late 1920s, in which major lamp makers agreed to cap incandescent bulb life at around 1,000 hours. In modern goods the effective mechanisms are different in kind: spare parts withdrawn a few years after a model is discontinued, batteries and drums glued rather than fastened, serialised components that refuse to work when swapped, and software support ending long before the hardware does. These reliably end a product's usable life on a schedule the maker controls. They are much easier to demonstrate than a timed component, because you can simply try to buy the part.
Third, what would actually settle the original claim is a failure-time distribution with a spike shortly after the warranty date, and that is not what reliability data generally shows — failures spread out along a wear-out curve rather than clustering at a line. Absence of a spike is decent evidence against a timer. It is not evidence that the curve was not deliberately pushed leftwards by component choices, which is the harder and better question.
A picture of it
THE PICTURE #How to readThe line is the bathtub curve — the failure rate falls steeply through the first year as defective units weed themselves out, flattens into a long low plateau of random failures, then climbs from around the middle of the machine's life as parts wear out. The bars mark a typical two-year warranty. Notice that the warranty covers the high early stretch and stops well before the climb, which is the point: it was placed there because the curve looks like this, not the other way round. The vertical scale is schematic — the shape is the claim, not the numbers — and the exact position of the climb differs by product and by build quality.
What became clearer
WHAT CLEARED #Two ordinary mechanisms account for the feeling without any conspiracy. The failure rate of a manufactured object is bathtub-shaped, and warranties are priced from that curve, so they are drawn to end before the wear-out climb begins — the warranty tracks the failures rather than causing them. And because warranties are short relative to lifespans, essentially every failure that costs you money is post-warranty, while the free ones inside the window are forgotten, so the remembered sample is guaranteed to look damning. What is genuinely worth being angry about is not a timer but the ordinary economics of cheaper components, unavailable spare parts, and software support that ends while the hardware still works.
Where to go next
ONWARD #- How extended warranties are priced, and why they are usually a poor purchase given the same curve.
- What right-to-repair rules on parts availability and documentation actually change about effective lifespan.
Key terms
TERMS #| Term | What it means |
|---|---|
| Bathtub curve | the characteristic shape of failure rate against age: high and falling, then low and flat, then rising. |
| Infant mortality | the early phase in which manufacturing defects surface and the failure rate falls. |
| Wear-out phase | the late phase in which fatigue and erosion make failure increasingly likely. |
| Planned obsolescence | deliberately limiting a product's usable life, whether by design, parts policy, or ending software support. |
Every term the collection defines is gathered in the glossary.