TL;DR

IT cost optimization almost always starts with software, because the data is sitting in a billing portal and unused seats are easy to prove. Hardware gets skipped for the opposite reason: nobody can see it. That is unfortunate, because the hardware savings are usually more durable. A cancelled licence comes back next year. A fixed refresh process does not.

  • Software savings are easy to find and easy to lose. Hardware savings stick.
  • The biggest hardware saving is usually devices you already own.
  • Over-specifying costs more than anyone measures, because it is invisible.
  • Extending refresh by a year moves cost, it does not remove it.
  • Count recovery rate. Every unreturned device is a full replacement.
  • Measure cost per working person, not spend by category.

Why IT cost optimization starts with software

Because it works, and quickly.

Licence data is in a portal. Usage data is usually available. Finding two hundred seats nobody has logged into in ninety days takes an afternoon and produces a number you can take to a finance meeting the same week.

Hardware offers none of that. The data is in an asset register that drifts, the usage data does not exist, and the devices are in people’s homes.

The durability difference

Software savings decay. Seats get re-added, usage grows, the vendor raises prices at renewal, and the same exercise is needed again next year.

Hardware savings tend to persist, because most of them are process changes rather than one-off cuts. A recovery process that works keeps working. A specification standard keeps applying to every machine bought after it.

That is the argument for spending some of the optimisation effort here even though it is slower to start.

The largest saving is usually devices you own

Before buying anything, the question worth asking is how many usable machines the company already has and cannot see.

Three pools, in descending order of size for most companies.

  • Unreturned devices. Machines with former employees. Each one is a full replacement cost that was already incurred.
  • Stock that is not visible. Devices recovered and stored somewhere nobody logs, so they are never reissued.
  • Devices with people who do not need them. Usually accumulated through role changes.

The first pool is normally the biggest and the most ignored, because recovery is treated as a compliance task rather than a cost one.

Action Where the saving comes from How durable
Improve recovery rate Avoided replacement purchases Permanent, if the process holds
Log and reissue stock Machines already paid for Permanent
Recover at role change Surplus specialist kit Permanent
Standardise specification Stop over-buying Permanent, applies forward
Extend refresh cycle Deferred spend Temporary, and it returns

The last row is the one most often chosen and the weakest. Extending a refresh cycle by a year defers cost into the next year, where it arrives alongside the following year’s cost. It is a legitimate thing to do deliberately and a poor thing to do as a savings programme.

Over-specifying is invisible and expensive

Most companies buy more machine than most roles need, and nobody notices because no single purchase is unreasonable.

The mechanism is usually procedural rather than technical. There is one standard machine, it has to be adequate for the most demanding role, so everybody gets that machine.

Three tiers tied to actual workload costs meaningfully less across a fleet and takes very little administration once defined. The resistance is rarely about cost, it is about fairness, which is a real objection and best answered by tying tiers to work rather than to seniority.

The counter-argument worth taking seriously

Under-specifying is more expensive than over-specifying, because a slow machine costs time every day and produces replacement requests ahead of schedule.

So the tiers should be set from observed need rather than from a cost target. Our note on total cost of ownership on devices covers how that calculation usually comes out.

If you want a structured way to think about the whole lifecycle rather than the purchase alone, ISO 55001 on asset management systems describes what an asset management system is meant to cover, and specification sits inside it rather than being a procurement decision in isolation.

Measure cost per working person

Spend by category tells you what you bought. It does not tell you whether it was efficient.

A more useful figure is annual device cost per person actually equipped, including purchase, shipping, support, recovery and disposal. That single number is comparable across years and across regions in a way that a purchasing total is not.

It also exposes the geography problem. If the cost per person in one market is three times another, the cause is usually supply rather than product, and the fix is a supply chain change rather than a buying policy.

Include the things people leave out

Shipping, customs, recovery and disposal are real costs and are frequently booked elsewhere, which makes hardware look cheaper than it is and makes the case for improving recovery look weaker than it is.

the EPA guidance on electronics stewardship covers the disposal side and why routing matters. Disposal is the smallest line and the one most likely to be done badly if it is invisible in the numbers.

The decision table

Start where your worst data is, not where the biggest number is.

Situation Scale Setup Primary pain Recommended starting point
No asset register Under 50 staff Spreadsheet Cannot see the fleet Count first, optimise second
Poor recovery rate Any Remote workforce Buying replacements for owned kit Fix collection, it pays back fastest
One specification for all 50 to 200 Standard build Over-buying quietly Three tiers by workload
Cost varies wildly by country Series B+ Several suppliers Supply chain, not product Compare cost per person by market
Under pressure to cut this quarter Any Budget freeze Need a number now Reissue stock, defer refresh knowingly

Row five is the honest answer to a short deadline. Reissuing recovered stock produces a real in-quarter saving. Deferring refresh produces one too, as long as everybody understands it is a deferral and it is recorded as such.

The order that works

Recovery first, because it has the fastest payback and every unreturned device is a replacement you have already paid for twice.

Then visibility of existing stock, because reissuing a machine you own beats buying one at any price.

Then specification, which applies forward and compounds with every purchase after it.

Then refresh timing, deliberately and with the deferral acknowledged rather than presented as a saving.

What makes recovery work

Not policy. Mechanism. A collection the company arranges and pays for, booked during the notice period, with packaging sent out. Our piece on IT offboarding covers the sequence.

For teams without local presence, recovery is the step that stalls. RemoAsset handles collection across markets and holds recovered stock so it can be reissued rather than written off, which is where most of the durable saving in this list actually comes from.

Why the order matters more than the list

Each step makes the next one cheaper, which is the argument for not doing them in parallel.

Recovery produces stock. Stock visibility makes reissue possible. Reissue reduces purchasing volume, which is the point at which a specification change starts mattering. Doing specification first, with no recovery working, optimises the cost of machines you are buying unnecessarily.

What to report while it is happening

Resist the urge to report a single savings figure. Report recovery rate, the count of devices reissued rather than bought, and cost per equipped person. Those three move at different speeds and together they show whether the change is structural or a one-off.

A single headline number invites the obvious follow-up question about whether it repeats next year, and for most of this list the honest answer is yes, which is worth being able to demonstrate.

RemoAsset recovers and stores devices so owned hardware gets reissued rather than replaced. More on device economics at PeopleOpsHQ.

Frequently asked questions

Why does IT cost optimization usually skip hardware?

Because the data is harder. Licence usage sits in a billing portal and unused seats can be proved in an afternoon, whereas hardware lives in an asset register that drifts, has no usage data at all, and is physically in people’s homes. The irony is that hardware savings tend to be more durable, since most of them are process changes rather than one-off cuts, while software savings decay as seats get re-added and prices rise at renewal.

What is the biggest hardware saving available?

Usually the devices you already own and cannot see. Three pools matter: machines still with former employees, recovered stock sitting somewhere nobody logs so it never gets reissued, and devices with people who no longer need them after a role change. The first is normally the largest and the most ignored, because recovery gets treated as a compliance task rather than a cost one.

Is extending the refresh cycle a real saving?

It is a deferral, and it is worth being honest about that internally. Pushing refresh out by a year moves the cost into the following year, where it arrives on top of that year’s normal replacement volume. Doing it deliberately as a cash timing decision is perfectly reasonable. Presenting it as a structural saving is not, and it tends to produce an uncomfortable conversation twelve months later.

How much does over-specifying devices cost?

More than anyone measures, because no individual purchase looks unreasonable. The usual mechanism is procedural: there is one standard machine, it has to be adequate for the most demanding role, so everybody receives it. Three tiers tied to actual workload costs meaningfully less across a fleet. Set the tiers from observed need rather than a cost target, since under-specifying costs more through lost time and early replacement requests.

What is the right metric for hardware efficiency?

Annual device cost per person actually equipped, including purchase, shipping, support, recovery and disposal. Spend by category tells you what you bought but not whether it was efficient, and it is not comparable across years or regions. The per-person figure also exposes geography problems, since a market costing three times another usually has a supply chain issue rather than a product one.

Which costs get left out of hardware calculations?

Shipping, customs, recovery and disposal, mostly because they are booked to different budgets. Leaving them out makes hardware look cheaper than it is and, more damagingly, makes the business case for improving recovery look weaker than it is. Disposal in particular is the smallest line and the one most likely to be done badly when it is invisible in the numbers.

What order should we tackle hardware costs in?

Recovery first, since it has the fastest payback and every unreturned device is a replacement already paid for twice. Then visibility of existing stock, because reissuing a machine you own beats buying one at any price. Then specification, which applies forward and compounds with every subsequent purchase. Then refresh timing, deliberately and labelled as a deferral rather than a saving.