TL;DR
Most technology refresh plans use one rule: replace at three years. It is simple, it is easy to budget, and it is wrong in both directions. It replaces machines that are working perfectly and keeps machines that have been failing for months, because age is a weak predictor of whether a particular laptop is still doing its job.
- Age alone replaces working machines and keeps failing ones.
- Combine age with failure history, battery health and performance complaints.
- Smooth the spend. A fleet bought in a burst fails in a burst.
- Count recovered stock as supply before ordering anything.
- Extending the cycle is a deferral. Say so when you do it.
- Replace early where failure is expensive, late where it is not.
Why a technology refresh plan based on age underperforms
Age is used because it is the only field everybody has. It is not used because it predicts anything well.
Two machines bought on the same day diverge enormously depending on workload, travel, charging habits and luck. One is fine at four years. The other has been slow and unreliable since month eighteen.
A pure age rule replaces both at the same moment, which wastes money on the first and leaves somebody struggling on the second for two years.
The cost nobody books
The time lost to a machine that is slow but not broken is invisible, because it never becomes a ticket. People adapt, work around it and complain occasionally.
That is the largest cost in this area and the hardest to evidence, which is why refresh arguments tend to be made on asset age instead. It is the available number rather than the right one.
Four inputs instead of one
All four are available from tooling most teams already run.
| Input | What it indicates | Where it comes from |
|---|---|---|
| Age | Baseline, warranty status | Asset register |
| Failure and repair history | Devices that keep breaking | Ticket system |
| Battery health | The most common real failure | Management agent |
| Performance complaints | Machines people are fighting | Tickets, or asking |
Failure history deserves a note too. A device that has already been repaired once is disproportionately likely to need attention again, partly because the underlying cause is often environmental rather than random, and partly because a repaired machine has been opened.
Battery health deserves particular attention because it is the failure mode that most affects daily experience and the one most easily measured remotely. A laptop that cannot hold charge for a meeting is functionally broken regardless of its age.
Our note on device health monitoring covers collecting those signals.
A simple scoring approach
Rank devices by a combination rather than sorting by purchase date. Anything with two or more negative indicators moves up the queue regardless of age, and anything with none moves down regardless of age.
That alone converts a refresh plan from a calendar into a prioritisation, and it tends to replace fewer machines while producing fewer complaints.
Smooth the wave
Companies that equipped a large intake at once face a replacement wave at the same moment several years later. The spike is both a cash problem and a supply problem, and the supply half is worse.
Ordering two hundred machines in one quarter means lead times stretch, stock runs short and everything else queues behind it.
Smoothing deliberately over two or three quarters costs very little. Start with the devices scoring worst on the inputs above rather than the oldest, which means the earliest replacements are also the most justified.
Plan the next wave out of existence
A smoothed refresh also prevents the next wave, because the fleet stops being bought in bursts. That is a quiet long-term benefit worth mentioning when the smoothing is proposed, since it is otherwise hard to see why spreading cost is better than deferring it.
Our piece on device total cost of ownership covers how the whole-life figure changes when replacement is steady rather than lumpy.
Count what you already have
Before any refresh order, establish how many usable machines exist outside active assignment.
Recovered devices from leavers, spare stock, machines returned after role changes. For many companies that pool covers a meaningful share of a refresh cycle and is invisible because nobody counts it.
The qualifications are real. Stock only counts if it is logged, wiped, of an acceptable specification and physically somewhere it can be deployed from. A cupboard of unlogged five year old laptops is not supply.
NIST Special Publication 800-88 on media sanitization covers what wiping should involve before a device is reissued, and it matters more for reissue than disposal because the machine stays inside the company and the next user should not inherit anything.
Where the stock sits
For distributed teams, recovered devices are scattered or uncollected, which is why they rarely get reused. RemoAsset handles recovery and regional storage, which is what makes reissue a real option rather than a theoretical one.
The decision table
The right rule depends on the cost of a failure in each population.
| Situation | Scale | Setup | Primary pain | Recommended starting point |
|---|---|---|---|---|
| Office-based, spares available | Any | Loan pool exists | Low, failure is recoverable | Run devices longer, replace on failure |
| Remote, no spares nearby | Any | Shipping takes days | Failure costs days of work | Replace earlier, keep regional stock |
| Fleet bought in one burst | Series B+ | Uniform age | Refresh wave approaching | Smooth across three quarters |
| Heavy workloads | Engineering teams | Demanding use | Machines age faster | Shorter cycle for that population only |
| Budget freeze | Any | No capital available | Cannot replace anything | Reissue stock, defer knowingly, document risk |
Rows one and two together make the real point. The right refresh interval is not a property of the hardware, it is a property of how expensive a failure is where that person sits.
If you extend the cycle, say what it is
Extending refresh is a legitimate decision and a poor savings story, and the difference matters internally.
Pushing replacement from three years to four moves cost into the following year, where it lands alongside that year’s normal volume. Done once, knowingly, as a cash timing decision, that is fine. Presented as a structural saving, it produces an uncomfortable conversation twelve months later.
Where a deferral is taken, record what it means operationally: more failures, longer average age, more support load, and a larger order next year.
What to monitor during a deferral
Failure rate and battery health across the deferred population, monthly. If either moves sharply, the deferral is costing more than it saved and the decision can be revisited with evidence rather than argument.
ISO 55001 on asset management systems describes lifecycle decisions as part of asset management rather than procurement, which is the right framing for a conversation that is really about risk rather than price. Our note on IT cost optimization covers where this sits among the other options.
What happens to the machines you replace
A refresh plan that stops at the purchase is half a plan. Every replacement produces a device that needs collecting, wiping and then either reissuing, selling or recycling, and that work arrives in the same burst as the deliveries.
Teams routinely budget for the new machines and not for the handling of the old ones, which is how refresh cycles end with a room of unprocessed laptops and an asset register full of rows nobody closed.
Decide the destination before the order is placed. Older machines going to recycling need a certified route and documented sanitisation. Newer machines coming back from a specification upgrade are often perfectly good for a different role, and that reissue is the cheapest device the company will acquire all year.
RemoAsset recovers and stores devices regionally so refresh can draw on stock you already own. More at PeopleOpsHQ.
Frequently asked questions
Should a technology refresh plan be based on device age?
Age is a reasonable baseline and a weak predictor on its own. Two machines bought the same day diverge enormously depending on workload, travel and charging habits, so a pure age rule replaces one that is working fine and keeps another that has been struggling for two years. Combine age with repair history, battery health and performance complaints, and rank devices rather than sorting them by purchase date.
What is the most useful signal for replacement?
Battery health, for laptops. It is the failure mode that most affects daily experience, it is easily measured remotely through the management agent, and a machine that cannot hold charge through a meeting is functionally broken whatever its age says. Repair history is the second, because devices that have failed once are disproportionately likely to fail again.
How do we handle a refresh wave?
Smooth it across two or three quarters rather than absorbing it in one. A fleet bought in a burst fails in a burst, and the supply problem is worse than the cash problem, since ordering two hundred machines at once stretches lead times and queues everything else behind it. Start with the devices scoring worst on health and failure history rather than the oldest, so the earliest replacements are the most justified.
Should recovered devices be used for refresh?
Yes, and they are usually not counted at all. Machines recovered from leavers, spare stock and devices returned after role changes often cover a meaningful share of a refresh cycle. The qualifications matter though: stock only counts if it is logged, wiped, of an acceptable specification and somewhere it can actually be deployed from. An unlogged cupboard of old laptops is not supply.
Is extending the refresh cycle a saving?
It is a deferral, and being honest about that internally is worth more than the number it produces. Moving replacement from three years to four pushes cost into the following year where it lands on top of that year’s normal volume. Taken once as a deliberate cash timing decision it is perfectly reasonable. Presented as a structural saving it produces a difficult conversation twelve months later.
What should we monitor during a deferral?
Failure rate and battery health across the deferred population, monthly. If either moves sharply, the deferral is costing more in support load and lost time than it saved, and the decision can be revisited with evidence rather than by argument. Record what the deferral means operationally at the point it is taken, so nobody is surprised by the larger order the following year.
Should everyone be on the same refresh cycle?
No, and the right interval is a property of the cost of failure rather than of the hardware. Somebody office-based with a loan pool nearby can run a device longer, because a failure costs an hour. Somebody remote in a market where replacement shipping takes days should be replaced earlier, because the same failure costs days of work. Heavy workloads justify a shorter cycle for that population alone.
RemoAsset recovers and stores devices regionally so refresh can draw on stock you already own.