When to Replace a Machine, and When Not To

An old machine used for email and a browser can be perfectly fine. A much newer one running a heavy application all day can be the worst thing in the building. Replacing by purchase date means you throw away working equipment while leaving the actual problem on somebody's desk.

Four questions decide it properly. Does it still get security updates. Can it be repaired in a sensible time if it fails. Can it run what the business needs it to run. And is somebody sitting there waiting for it. Any one of those can force a replacement on its own, and only the first comes with a fixed date attached.

Run those four over your list once a year and you get a replacement plan that is defensible in a budget meeting, which the alternative never is. Nobody has ever won an argument for money with the sentence that the machines are getting old.

When the manufacturer stops issuing updates for the operating system, or the machine can no longer take the version that is still supported, it stops being safe to have on your network. That is not a performance judgement and it does not care how well the thing runs. A fast machine that stops receiving fixes is a fast machine with an expiry date.

Firmware matters here as much as the operating system, and it gets forgotten because nobody sees it. Manufacturers stop issuing firmware updates for a model at some point, usually earlier than they stop supporting the software running on it. Network equipment is the worst offender, because a switch or a firewall can run happily for many years while quietly having no supported version left.

The useful thing about this test is that it produces dates. Every operating system has a published support end date. So does most business software, and so does a good deal of hardware. Put those dates on the same calendar as your contract renewals and a large part of next year's replacement budget writes itself.

The awkward case is the machine that could take a newer operating system but cannot run the application the business needs on it. That is not a hardware decision any more. It is a conversation with a software vendor, and it needs starting long before the support date arrives.

Once the warranty ends, the question changes from whether it is covered to whether it can be fixed and how fast. For common models the answer is usually yes and quickly. For anything unusual, or anything old enough that the manufacturer has moved on, parts get scarce and then get expensive and then stop existing.

So ask what happens if this fails on a Tuesday morning. If the answer is that the person is off work for a week, the machine needs replacing or you need a spare. Sometimes the spare is the better buy. One identical machine sitting built and ready in a cupboard can cover a whole group of similar machines, and it costs a fraction of replacing all of them.

That is worth saying plainly because it goes against how replacement is usually sold. A shelf spare with a current image on it, swapped in within the hour while the broken one is looked at properly, solves the business problem faster than any warranty tier and often costs less over a few years.

This is the part that never makes it into a business case, because it feels soft and it is not. A person waiting for a machine is a person being paid to wait, and it happens many times a day in small pieces that nobody records.

You will not find it in a report, so go and look. Sit next to somebody using the machine you are wondering about, for twenty minutes, without helping. Watch what they do while they wait. Then look at the behaviour rather than at the clock. People with slow machines stop closing applications because reopening them is worse. They stop opening the big report. They leave everything running for days and never restart, which then causes its own faults. They start doing bits of the job in a spreadsheet on the side.

The other source is the ticket history for that serial number. A machine generating a steady trickle of small tickets is costing you the engineer time, the user time, and the interruption, over and over, and every individual ticket looks too trivial to act on.

Then ask the person directly, and believe them. Staff are usually reluctant to complain about equipment because it feels like asking for a treat. If somebody does raise it unprompted, it has normally been bad for a long time.

Sometimes the machine is fine and one component is the bottleneck. A solid state drive in place of a spinning one changes the feel of an older machine more than any other single change. More memory helps where the person keeps a lot open at once, which is most people. A new battery makes a laptop portable again. A failing screen or a broken hinge is a repair rather than a reason to buy.

The conditions for that being worth it are narrow, though. The machine has to still be getting security updates and have a good while left before it stops. The parts have to be available and fitted without opening something that will not go back together. And the labour plus parts has to be well below the replacement cost, because a half priced upgrade on a machine you replace next year has saved nothing.

It stops being worth it when the memory is soldered in, when the support date is close, when the chassis or the keyboard is already going, or when you would be doing it to more machines than you can be bothered to track. There is a real overhead in a fleet where every machine is slightly different, and it lands on whoever supports it later.

Fleets bought together age together and fail together, and then they all need replacing in the same budget year, which is the year it gets refused. The way out is to replace a share of the estate annually so the cost is level and predictable.

Getting there from a cliff takes deliberate unevenness for a couple of years. Pick what to do first by the job rather than by the date: the people who wait most, the roles where a failure stops something, the machines running software with a support date coming. Then let the cycle settle.

Handing machines down the business can work and it is often false economy. It makes sense when the machine genuinely suits the lighter role and has real supported life left. It stops making sense when the hand down is being used to avoid buying anything, because you now have your oldest equipment concentrated in the roles with the least influence to complain about it. That tends to be the warehouse, reception and whoever works part time.

Nearly every business has one. The computer attached to a cutting machine, a testing rig or a label printer, running software from a vendor who has stopped updating it or has gone. It cannot be upgraded because the software will not run on anything newer, and it cannot be replaced because the equipment costs more than the department's annual budget.

Stop treating that as a technology problem to be solved quietly and start treating it as a business risk with an owner. What happens if it dies tomorrow. How long would the equipment be unusable. Who at the vendor, if anyone, still answers.

In the meantime, there is sensible work to do. Take it off the general network so it can only reach what it genuinely needs. Take a full image of it and test that the image actually starts on something. Buy the spare hardware now, while the parts still exist on the second hand market, because they will not in a few years. And put the replacement conversation in front of whoever owns the equipment budget, in writing, with the risk described in their words rather than yours.

The week before a swap is when the surprises live. Local files nobody backed up, saved in a folder on the desktop. A licence tied to that specific machine that has to be released before it will activate elsewhere. A certificate installed in the browser for a supplier portal. Saved passwords in a browser profile nobody has ever signed into properly. A scanner or a printer driver that only exists on a disc in a drawer.

So swap while the person is there, not while they are on holiday, and keep the old machine intact and untouched for a couple of weeks afterwards. Something always turns up on the third day, and it is far cheaper to boot the old one than to explain why it has already been wiped.

Then finish the job. Wipe it properly, record the disposal with a date, and take it off the inventory. A machine that leaves the building with its storage intact is a data problem sitting in somebody's car, and it is one of the easiest ones to avoid entirely.