Nobody can tell you about your unit
How long should a product last is a question with an honest answer and a dishonest one, and the dishonest one is more common because it is more satisfying.
Nobody can predict the lifespan of a specific unit. Manufacturing varies, use varies, and a figure invented to look precise is worse than admitting a range. What the evidence does support is comparing two candidates on signals that correlate with life, and judging a product against its own category rather than against a general sense of how long things should last.
That second half is where most of the mistakes are. This sits inside the wider question of whether something is worth buying, and it is the input that decides what a price really was.
Judge it against its own category
A laptop lasting four years is unremarkable. A washing machine lasting four years is a failure. A pair of running shoes lasting four years means you did not run.
Applying one expectation across categories is how people conclude that everything is badly made now. The baseline differs by an order of magnitude depending on what the thing is made of, how hard it works, and whether it contains a consumable.
Three rough groups are worth holding in mind.
Things that wear out by use. Shoes, brake pads, filters, blades, printer drums. Life is measured in cycles rather than years, and calendar age barely matters. A five-year-old unused pair of boots is nearly new.
Things that age by time. Batteries, adhesives, seals, anything chemical. These start their life at manufacture rather than at purchase, which is why a new-in-box item that sat in a warehouse for eighteen months is not new in the way that matters.
Things that fail electronically. Power supplies, capacitors, controller boards. These fail on a curve with an early cluster and then a long flat middle, which is why a product that survives its first month usually survives several years.
Knowing which group a purchase belongs to tells you more than any number.
The indirect signals that actually predict
Manufacturers rarely state expected life, and when they do it is a warranty period rather than a forecast. Four indirect signals do better.
How long the model has been on sale without revision. A design in its fourth year with no changes is either very good or abandoned, and the reviews usually make clear which. A model revised every nine months is a design that is still being fixed.
Whether the parts that wear out can be replaced, and bought by you. This is the strongest single signal in most categories. A product whose battery, filter or belt costs twenty dollars and ten minutes has a much longer practical life than an identical one that is sealed. Check that the part exists and that you can buy it rather than only an authorised repairer.
What the negative reviews complain about, and when. Look for a pattern in the timing rather than in the volume. A cluster of failures at fourteen months is far more informative than an overall star rating, and it usually names the component.
Whether a repair industry exists. If independent repairers advertise work on this model, the thing is repairable and worth repairing. If nobody does, that is an answer too.
What the warranty does and does not tell you
A warranty period is a floor set by legal and financial teams. It is chosen so that claims within it cost less than the sales it produces, which makes it a statement about expected failure rates rather than about design life.
That still makes it informative. A company willing to stand behind something for five years has run the numbers and believes most units will survive that long. A company offering ninety days has run the same numbers.
Two refinements. Check what the warranty actually covers, because a five-year warranty on the motor and one year on everything else is a one-year warranty with marketing attached. And check who honours it, since a manufacturer warranty is honoured by the manufacturer rather than by whoever sold it, and several decline claims on units bought through unauthorised sellers.
Where the tiers differ mainly in warranty length rather than in function, that is worth reading as the manufacturer's own estimate of the difference between them, which is the subject of budget against premium.
Are things really built worse now?
The honest answer is more mixed than the complaint.
Some categories genuinely have thinned. Price pressure at the low end removes material, and a product engineered to hit a price point has less margin everywhere. Large appliances in particular carry more electronics than they did, and more electronics means more things that can fail.
Other categories have improved substantially. Manufacturing tolerances, corrosion resistance and electronic reliability are all better than they were, and several products that used to need annual servicing no longer do.
What has changed most is not design life but repairability. Glued assemblies, proprietary fasteners, unavailable parts and software locks shorten practical life without shortening the engineered one. A product that would run for fifteen years but cannot be opened has a life set by its least accessible component.
Which is why repairability is the signal worth weighting most heavily. It is the one that separates a product you will keep from one you will replace, and it is visible before you buy.
What shortens life faster than manufacturing
Three things account for more early failures than build quality does, and all three are within your control.
Use outside the design envelope. A domestic machine run at commercial volumes, a tool used on a material it was not made for, a device left in heat. Most of these are invisible in the moment and obvious in the failure.
Skipped maintenance on a consumable. Filters, belts, descaling, cleaning. The part that is meant to be replaced takes the rest of the machine with it when it is not, and the maintenance schedule is usually in a manual nobody opened.
Power and water quality. Unremarkable until it is the cause. Hard water shortens the life of anything that heats it, and poor supply quality shortens the life of electronics in ways that look random.
None of this shifts responsibility onto the buyer for a badly made product. It does mean that two identical units can have very different lives, which is part of why a single predicted number was never available in the first place.
Turning it into a number you can use
You do not need a prediction. You need a plausible range, and then the arithmetic.
Take the category baseline, adjust it for the signals above, and use the lower end. Then divide the price by that many years and compare the candidates on the result. A product costing half again as much and lasting twice as long is cheaper, and that comparison is invisible while you are looking at two prices.
That is the same calculation as cost per use, run per year rather than per use, and for anything you keep for a long time it is the more useful of the two.
One last habit. When something does fail, note how long it lasted. A personal record of what survives in your house, in your use, beats any published figure, and it is the only evidence that is genuinely about your conditions.
And treat a failure as information rather than as bad luck. What broke, at what age, and whether the part was replaceable tells you more about what to buy next than any review you could read beforehand.