Hardware failures are undesirable, unpredictable, and misunderstood.

A UPS protects you from power failure, but has components that have degradation over time and MTBF ratings, and often moving parts (fans). All of these will fail, but predicting when is hard. 

Many people would assume that a UPS that is a few years old is well into its service life, and entering the attrition of wear out failure. This does not square with what we are seeing. 

Think about the common electronics failure mode chart:

The most likely times for UPS failure are

1/ Within 1 minute of it first being turned on (the highest risk)
2/ When a sub component fails after years of use either due to breakdown, internal physical short, or seizure.
3/ First turn on after shipping (transit shock which caused physical damage)
4/ 100% failure by the time the Sun absorbs the Earth

1, and 3 are self explanatory. 

(2) is more interesting. We find there are two modes on this
2-A Manufacturing defect that passes initial tests at the manufacturers, but blows shortly after installation
2-B Nominal part that has intrinsic chemical or otherwise degradation over the years

Manufacturers are testing UPS before shipping them out, probably with overloads to over stress to try to find all cases of 2-A, but they are restricted by time. We are up against the same time challenges here and understand that is unavoidable.

<< The advantage that we have is the knowledge of the UPS having been used for a length of time before it got to us. >>

This is counter intuitive - buying used is reducing one of the risk factors given the UPS has had a long time to prove its sub component. 

Customers will be asking themselves - how long will a UPS last? 

Answer: As long as a piece of string - Who knows? But do not underestimate the advantage of buying used vs new (outside of cost + environmental considerations).

We have have a test processes (which changes depending on what work has been carried out on the unit) and have become good at sniffing out problematic sounds / characteristics, and our UPS come with 12 month warranty, which is an indication that we are standing by what we are doing.

We benefit from from working with UPS over the yearss. When we get a very old one, we have a project with it - efforts are made to see if it works. In 2023 we received a 1998 era APC UPS (with the old logo, and with a nice large ball bearing fan) that still worked. The capacitors used had not degraded to failure. Conversely we have had UPS that were less than 12 months since manufacture, with their original packaging cellophane screen protectors, low or no "power on hours" in immaculate condition, that failed within 10 minutes of power on testing. Cosmetic condition and age are not criteria that are not a certain indicator of reliabilty.

I'm writing this to give people some idea of how long a UPS can last under optimal circumstances, or even none optimal, the low risk constant failure rate section of the units risk vs time chart is vast. When people have come for their second or third battery replacement (IE after 15 years of the UPS in use) they often opt to change the UPS, either on the basis "they have had a good run", its got too dusty with no way to clean out, or just wanting a change of scenery, but a notable absence of the UPS throwing a fault. 

Given all the above, I reverted to MS-Paint to create what I think is an accurate graph of failure over time for a UPS