TL;DR
- The purchase price of a laptop is somewhere between a third and two thirds of what it costs you, and the rest is spread across people's weeks where nobody counts it.
- If you buy laptops in one country and hand them over in person, your real cost is close to the sticker price. This article is for everybody else.
- Six costs sit behind the purchase: delivery, configuration, support, replacement, recovery and disposal. Only the first is ever quoted.
- The write-off line is the one companies omit and it is frequently the largest single addition.
- Divide by the real life you get, not the life your policy claims. That one correction moves the figure more than any other.
- Build the number before you need it. Every vendor conversation and every budget argument is easier with a figure of your own.
The Business Case That Compared Two Different Things
A company evaluating a device-as-a-service arrangement built a comparison. On one side, the provider's monthly figure times the fleet times three years. On the other, the purchase price of the same laptops.
Buying won comfortably, and the finance partner approved continuing as they were. Eighteen months later somebody rebuilt the comparison after an unrelated audit, this time including what the company actually spent on getting devices to people, replacing broken ones, chasing returns and clearing a storeroom. The gap narrowed to almost nothing, and on a couple of reasonable assumptions it reversed.
Nobody had been dishonest. The first comparison used the number that was easy to obtain against the number the vendor provided. The problem is that a vendor quote is an all-in figure and a purchase price is not, so comparing them directly is comparing a complete cost against a partial one, and the partial one always wins. The decision may still have been right. It was not an informed one.
Best tools for Device Management
So the useful exercise is building your own all-in figure, and it is a morning's work on data you already have.
When You Don't Need to Do This
When the manual way is genuinely fine
One country, devices handed over in person, no recovery problem, and nobody asking you to justify hardware spend. Your purchase price is close to your real cost and the extra precision will not change a decision. Spend the morning on something else.
When friction starts appearing
Somebody asks what hardware costs per head and the honest answer is that you would have to work it out. That question usually arrives from finance during planning, and the first time it arrives is the cheapest time to build the figure, because nothing is riding on the answer yet.
When it becomes a liability
When you are about to sign a multi-year agreement, or decline one, without a comparator. A three-year commitment evaluated against an incomplete baseline is a decision made on a number you know to be wrong, and both directions of error are expensive.
The edge case that forces it
A board or investor question about cost per employee, which arrives with no notice and expects a figure. Also any cost-reduction exercise, where hardware is an obvious target and an undefended number gets cut arbitrarily.
Five Questions People Ask First
"Is this just total cost of ownership?" It is a narrow, practical version of it, scoped to a figure you can actually produce in a morning and defend in a meeting. Full TCO exercises tend to sprawl into allocated overheads and never finish, which is why most companies do not have the number.
"What is the single biggest omission?" Devices that were never returned, valued at what they were worth when the person left. It is the largest addition in most companies and the most reluctantly included, because producing it means stating the recovery rate out loud.
"Should we include people's time?" Yes, at a realistic internal rate, for the time that is genuinely spent: configuring, shipping, chasing, handling tickets. Do not allocate management overhead or office costs, which is where these exercises go to die.
"Over how many years?" The real life you get, measured from your own fleet, not the life your policy specifies. If the policy says three and the fleet averages four and a half, use four and a half, which will reduce your annual figure by roughly a third and is the honest number.
"Will this make buying look worse?" It will make buying look more expensive than the sticker price, which it is, and it may still be the cheaper option. The point is not to reach a conclusion, it is to compare like with like. Several of the six costs are also reducible without changing anything about how you finance hardware.
The Six Costs Behind the Purchase Price
Delivery
Getting the device to the person. Shipping, any customs charges, and the time somebody spends arranging it. For a distributed company this is the largest of the six after the write-offs, and it is invisible because it arrives as courier invoices rather than as hardware spend. Include failed deliveries and redeliveries, which are more common than anybody tracks and cost close to the original shipment each time.
Configuration
Preparing the machine. Smaller than it used to be, because modern enrolment means most of it happens on first boot, which is worth reflecting honestly rather than inflating. Count the time actually spent, which for many teams is twenty minutes rather than two hours. The exception is any market where you cannot use automated enrolment and somebody builds machines by hand, which is a much larger per-device figure and worth recording separately so the contrast is visible.
Support
Hardware-related tickets, their handling time, and the elapsed days somebody could not work. The second part is usually larger than the first and is almost always left out. If you cannot separate hardware tickets cleanly, sample a month and extrapolate rather than abandoning the line, since an approximate figure here beats a zero by a wide margin.
Replacement
Machines that fail within their expected life and have to be replaced early. Count both the hardware and the expedited shipping, since a failure replacement is nearly always urgent and urgent shipping costs more. Where warranty covered the repair, the hardware line is zero and the shipping and downtime lines are not, which is a distinction worth preserving because it shows what better regional cover would actually save.
Recovery
Getting devices back. Courier costs, packaging, and the chasing time, which is the part people underestimate because it is spread across several short interactions rather than one task. Ask whoever does the chasing how many messages a typical return takes. The answer is usually three or four, across two weeks, and multiplying that by your annual leaver count produces a figure that surprises people.
Disposal
Wiping, certified destruction where needed, and recycling. Frequently close to zero in cash terms and occasionally negative where resale works, and it belongs in the model either way so the figure is complete. A zero here is only honest if disposal is actually happening. If there is a storeroom nobody has dealt with, estimate what clearing it will cost and carry that, because the obligation exists whether or not an invoice has arrived.
The Seventh Line, Which Is Not a Cost
Resale or residual value, which reduces the total and is the one line that moves in your favour. Include it honestly: what you actually recovered last year, not what a vendor quoted for machines in better condition than yours.
Two cautions. Value the devices at the age you actually sell them rather than the age your policy implies, which is the same gap that distorts everything else here. And only count value from devices you actually recovered, since a laptop with a notional resale value sitting in somebody's flat is worth nothing to you.
A third caution worth adding, because it catches people out at the point of sale. Activation locks and devices still enrolled in management are frequently worth a fraction of their released equivalent, since a refurbisher who cannot reset a machine can only break it for parts. Releasing devices from management before they go to resale takes a few minutes per batch and it is the difference between a resale line and a recycling line. Build the model on what a cleared machine fetches, then make sure the devices are actually cleared before they leave.
Putting It Together
| Line | Where the number comes from | Typical blind spot |
|---|---|---|
| Purchase price | Finance or purchase records | Using list price rather than what you paid |
| Delivery | Courier invoices plus arranging time | Customs charges booked to a different cost code |
| Configuration | Time per device at an internal rate | Overstating it, since enrolment automated most of this |
| Support | Hardware tickets, handling time, days lost | Omitting the days the employee could not work |
| Replacement | Early failures, hardware plus expedited shipping | Counting the device and not the urgency premium |
| Recovery | Courier, packaging, chasing time | Chasing time, spread thin and never logged |
| Disposal | Vendor invoices, or an honest estimate | Assuming zero because nobody has dealt with the cupboard |
| Less resale | What you actually recovered last year | Using a quote for machines better than yours |
| Divided by | Real average life in years, from your own fleet | Using the policy figure instead of the actual |
Work down the table, produce a figure per device per year, and write the assumptions next to it. The assumptions matter as much as the number, because the first thing anybody does with a figure like this is ask how you got it.
A note on precision. This model is accurate to within perhaps ten or fifteen per cent, and that is entirely sufficient for every decision it informs. The temptation is to refine it further, chasing exact courier allocations and precise support minutes, and the refinement adds days of work without changing a single conclusion. Get the eight lines to a defensible approximation, write down how confident you are in each, and stop. A rough figure that exists beats a precise one that is still being built when the decision is made.
Where Each Number Hides in Your Systems
The reason most companies do not have this figure is not that it is hard to calculate. It is that the inputs live in six systems owned by four teams, and nobody has ever pulled them together.
| Cost line | System it lives in | Who owns it | How long it takes to extract |
|---|---|---|---|
| Purchase price | Accounting or purchase orders | Finance | An hour, filtered by supplier or category |
| Delivery | Courier account, expense claims | Operations or whoever holds the account | An hour, and it is usually split across codes |
| Configuration | Nowhere. It is time | IT | Ask the two people who do it for an estimate |
| Support | Service desk, filtered to hardware | IT | An hour, and filtering is the fiddly part |
| Replacement | Purchase records, cross-referenced to failures | Finance and IT together | Two hours, since it needs a join |
| Recovery | Courier account plus the chasing nobody logs | Whoever handles offboarding | Ask them. There is no report |
| Disposal | Vendor invoices, or absent entirely | IT or facilities | Ten minutes, or an honest estimate of zero |
| Resale | Credit notes, or a separate receipt | Finance | An hour, and it is frequently miscoded |
Two of the eight are not in any system. Configuration time and chasing time exist only in people's weeks, which is exactly why they are omitted, and an estimate from the person who does the work is far better than a zero. Ask them for a per-device figure and use it.
The miscoding problem is worth anticipating. Courier spend on devices is routinely booked to general postage, and resale receipts often land as miscellaneous income, so both need somebody in finance to look rather than a report to run. Budget an extra hour for that conversation and the model is a morning rather than a project.
The Four Lines You Can Reduce Without Buying Anything
Producing the figure is useful. Three of the lines are reducible immediately, which is the part that pays for the morning.
Recovery, by asking earlier. Moving the equipment request to the first week of the notice period rather than after the last day improves return rates substantially, and it costs nothing. Every device recovered is both a write-off avoided and a resale value gained, so this line moves twice. It is the highest-return change available in the entire model and it requires a calendar trigger rather than a budget.
Replacement, by holding regional spares. A spare machine in the market where somebody works turns an international expedited shipment into a domestic one. The capital cost of holding it is usually less than the premium on two urgent shipments. Fund the first spares from recovered machines rather than new purchases, which costs nothing and connects the two halves of the lifecycle in a way that makes both easier to justify.
Delivery, by buying locally. Shipping a laptop from your home country is a commercial import with the full customs process. Buying in the destination market is frequently cheaper all-in despite a higher unit price, and it keeps warranty cover local, which reduces the replacement line too. Test it in one market before changing the policy everywhere, since supply and price vary enough that the conclusion in one country does not transfer to the next.
And disposal, by wiping on arrival. Not a cash saving but a risk one. Wiping when a device comes back rather than when it eventually leaves closes the data obligation immediately, which removes the urgency from disposal and lets you batch it properly. Batching is where the disposal line actually improves, since a vendor quoting on forty machines gives a better answer than one quoting on four.
What you cannot reduce without a decision is configuration, which is already mostly automated, and support, which tracks fleet age and therefore belongs to the refresh conversation rather than this one. Both are worth measuring anyway, because they are the two lines that will move on their own if the fleet drifts older, and noticing that early is the main reason to rebuild the model annually rather than once.
Why the Figure Varies So Much Between Companies
Two companies with identical fleets can land on figures a long way apart, and the differences are structural rather than a sign that one of them calculated wrongly. Five variables account for most of the spread.
How far devices travel. A single-country fleet pays a courier a modest amount per device. A fleet spread across eight markets pays international shipping, sometimes duty, and occasionally a second shipment when the first is held. This is the largest source of variation between otherwise similar companies.
Recovery rate. The gap between recovering 90 per cent of devices and 60 per cent is enormous once write-offs and forgone resale are both counted, and it is almost entirely a function of process rather than of geography.
Real device life. A fleet running four and a half years has a materially lower annual hardware cost and a materially higher support cost than one running three. The two partly offset, which is why the right cycle is a crossover rather than a maximum.
Standardisation. A fleet with three models has interchangeable spares, shorter support runbooks and better resale batching. One with eleven has none of those, and the cost appears in support and resale rather than in purchasing.
Whether anybody does disposal. A company that has never dealt with its storeroom has a disposal cost of zero in the model and a growing liability in the corner, which is an accounting artefact rather than an efficiency.
The practical use of knowing this is in comparison. If somebody quotes you an industry figure for cost per device, the first question is which of these five they resemble, and the answer is usually that you cannot tell. Your own number, built from your own records, is worth more than any benchmark precisely because these five vary so widely.
How to Choose: Five Questions Before You Talk to Any Vendor
Do you have the figure at all? If not, get it before any conversation. A vendor with an all-in monthly number and a buyer with a purchase price is not a negotiation, it is a presentation. Half a day of preparation changes the shape of every meeting that follows.
What is your real average device life? This single input moves the annual figure more than any other, and most companies quote their policy rather than their fleet. Pull the age distribution rather than asking anybody, since the policy figure is what people will tell you in good faith.
What is your recovery rate? It drives the write-off line, it will be an input to any lease contract, and knowing it lets you negotiate the assumption rather than inherit one. Calculate it as devices returned over devices issued to people who have since left, across the last twelve months.
Which lines would a vendor actually remove? Score any proposal against your six lines specifically. A service you already do well is worth nothing to you however prominently it is presented. Write the scores down during the demo rather than afterwards, because a good presentation makes everything sound necessary in retrospect.
Who owns the figure afterwards? A number produced once for a business case and never refreshed becomes wrong quietly, and the next person to use it will not know how old it is. Put a date on it and rebuild it annually. Store it alongside the assumptions in the same place each year so the series is comparable, which is what makes the trend readable later.
Where the Numbers Come From
Figures below were read from each vendor's own pricing page, most recently 9 October 2026, and they illustrate which parts of the model are knowable in advance and which are not.
The register tools that help you build it
Snipe-IT holds purchase dates, costs and custom fields, which is everything the model needs, and its self-hosted edition is free and open source. Hosted tiers are published at $39.99 monthly or $399.99 annually for Basic, and $99.99 or $999.99 for Small Business.
AssetTiger publishes pricing by asset count at $20 a month for 500 assets, $40 for 2,500 and $75 for 10,000, reducing to $18, $37 and $69 on annual billing.
Both will give you the purchase, age and assignment data. Neither knows what your shipping or support cost, which you take from courier invoices and your service desk.
The platforms that absorb some of the lines
Workwize, Deel IT, Firstbase, GroWrk and allwhere each take on delivery, replacement and recovery to varying depths. None publishes a price, so their effect on your model can only be established by quote.
RemoAsset
Disclosure: RemoAsset is owned by the same people who publish HROpsLab. It appears here because it competes in this category and is assessed against the same criteria as everything else on this page, with its limitations stated in the same detail.
Relevant to this model because it addresses three of the six lines, meaning delivery, recovery and the write-offs that follow from poor recovery, and because the asset record is created at delivery, which is what makes the model buildable in the first place.
Where it struggles: it publishes no price and requires a demo, so it cannot be modelled before a sales conversation. It is weaker for hardware it did not supply. And it is not a certified disposal vendor, so the disposal line stays with somebody else.
Cost Per Employee Is Not Cost Per Device
The title of this article says per employee and the model above produces a figure per device, and the gap between those two is worth being precise about because it is where several arguments go wrong.
Not everybody has one device. Some roles have two, some have a phone as well, and some share. A cost per device multiplied by headcount overstates in companies with shared or partial allocation and understates in those issuing phones alongside laptops.
Spares are a cost with no employee attached. Regional stock, loan machines and the buffer that makes replacement fast are all real spend sitting against nobody. They belong in the total and they inflate the per-device figure, which is correct, since the buffer exists to serve the fleet.
Unissued stock distorts the denominator. A company that bought forty machines in anticipation of hiring has forty devices and no corresponding employees. Count them in the fleet for asset purposes and exclude them when expressing cost per head, or the figure moves for reasons unrelated to how you operate.
And leavers' devices sit in limbo. A machine recovered in March and reissued in June belongs to nobody for a quarter. Over a year this is a meaningful slice of a fleet and it is one of the reasons the per-device and per-head numbers diverge.
So produce both and label them clearly. Cost per device answers questions about the fleet and about vendor comparisons. Cost per employee answers the question finance actually asked, and it is the first number divided by the ratio of devices to people, which is a figure most companies have never calculated and which is rarely exactly one.
The Decision Table
| Situation | Scale | Setup | Primary Pain | Recommended Starting Point |
|---|---|---|---|---|
| One country, in-person handover | Any | None | Purchase price is close to true cost | Use the purchase price. Do not over-engineer this |
| Evaluating a lease or DaaS arrangement | Any | Full six-line model | Comparing an all-in quote to a sticker price | Build the model first. The quote is meaningless without it |
| Finance asking cost per head | Any | Six-line model plus assumptions | No figure exists | Produce it with the assumptions written beside it |
| Recovery rate below 80 per cent | Any | Write-off line, honestly valued | The largest omission in most models | Include it, then fix recovery, which moves two lines |
| Distributed fleet, high shipping spend | 100 plus | Delivery and replacement lines | Courier cost booked away from hardware | Pull the courier invoices. Consider buying locally |
| Cost-reduction exercise underway | Any | Defensible figure with a date | An undefended number gets cut arbitrarily | Have the model ready before the conversation |
| Model built two years ago | Any | Rebuild annually | Quietly wrong, and nobody knows how old it is | Date it and refresh it each planning cycle |
Presenting It So It Survives Contact With Finance
A figure built carefully and presented badly gets argued with rather than used. Five things that make it hold up.
Lead with the method, not the number. Open by saying which eight lines you included and where each came from. A figure whose derivation is visible invites scrutiny of the assumptions, which is a productive conversation. A figure presented alone invites scrutiny of the figure, which is not.
Show the purchase price separately. Everybody in the room recognises it and will look for it. Presenting the all-in figure without the familiar number inside it reads as inflation, even when it is not.
State the assumptions as a short list with values. Real average device life, recovery rate, internal hourly rate, resale realised. Four lines. These are where a reasonable person might disagree, and surfacing them turns disagreement into an adjustment rather than a rejection.
Give a range where you estimated. Configuration time and chasing time are estimates, so present them as a band rather than a point. A model that admits which parts are soft is more credible than one that implies every line came from a system.
And say what you would do about it. A cost figure with no accompanying actions is a complaint. The three reducible lines give you a proposal, which converts the exercise from reporting into a plan and is the reason anybody will fund the next step.
One thing to avoid: do not build the model to support a decision you have already made. A comparator constructed to win an argument will win it, and the arrangement will then underperform against a number that was never real, which is the opening example running in the other direction.
What Getting This Wrong Costs
The direct cost is the decision itself, and it runs in both directions. A company that underestimates its real cost declines arrangements that would have helped. One that inflates it to justify a purchase signs something that underperforms against a number that was never real.
The second cost is the negotiating position. Going into a vendor conversation without your own figure means accepting their framing of the comparison entirely, and their framing will be their all-in number against your purchase price. That is not a trap anybody is setting; it is simply the only comparison available if you have not done the work.
The third is the one that recurs. Hardware is a visible line in any cost-reduction exercise and an undefended line gets cut by a percentage rather than by analysis. The team then runs an older fleet, support load rises, and the saving reverses over the following two years in costs that appear somewhere else entirely and are never connected back.
The fourth cost is organisational and it is the reason this keeps happening. Hardware spend is distributed across finance, IT, operations and whoever holds the courier account, so no single person sees the total and nobody is accountable for it. Each team can truthfully report that their own line is under control while the aggregate drifts upwards unexamined for years. That is not a failure of any of them; it is what happens when a cost has four owners and no total, and producing the figure is mostly an act of making something visible rather than of analysis.
So the question worth having an answer to, before anybody asks, is what a device costs you per person per year, and which three of those lines you could reduce this quarter.
Using the Figure for Something Other Than a Vendor Comparison
The model gets built for a procurement decision and it is more useful afterwards, for four things nobody anticipates.
Budgeting a new market. Opening in a country has a hardware cost that is knowable in advance once you have per-device figures broken out by shipping destination. Most expansion plans cost salaries and office space and treat equipment as a rounding error, which it is not when every device is an international shipment.
Justifying recovery work. Device recovery is a thankless programme that never gets resourced, because its benefit is expressed as a percentage rather than as money. Multiply your unrecovered devices by their residual value and it becomes a number, and numbers get funded.
Setting an equipment stipend sensibly. Any company considering giving people money to buy their own equipment needs to know what providing it costs, including the parts they would stop paying and the parts they would not. The model gives you that directly, and the answer is often that a stipend saves less than expected because delivery and configuration disappear while support and the lack of any recovery route do not.
Spotting a problem early. Rebuilt annually, the figure moves, and the direction tells you something. A rising per-device cost with a stable fleet usually means age, since support and replacement climb together. A rising cost with a growing fleet usually means geography, since you have started shipping further.
That last use is the best argument for dating the model and rebuilding it each cycle. A single figure is an input to one decision. Three years of the same figure, calculated the same way, is the only early warning you will get that something in the fleet is drifting.
When You're Ready to Move Beyond the Purchase Price
Most companies use the purchase price because it is the number that exists. It is on an invoice, nobody disputes it, and for a long time it is close enough to the truth that using anything else would be false precision.
What makes it stop being adequate is distribution. The moment devices travel, the costs around the hardware grow faster than the hardware itself, and they grow in places that do not look like hardware spend: courier accounts, support time, an unrecovered laptop written off quietly at the end of a quarter. The purchase price stops being an approximation of the truth and becomes a fraction of it, and the fraction gets smaller as you spread out.
The sequence is a morning. Pull purchase records, courier invoices and hardware tickets for the last twelve months. Add the write-offs at honest residual value. Subtract what you actually recovered from resale. Divide by the real average life taken from your own fleet rather than your policy. Write the assumptions next to the figure, date it, and rebuild it each planning cycle. Then pick the three lines you can reduce without buying anything, which for most distributed companies are recovery, replacement and delivery, and work on those before anybody tries to sell you something that addresses them for a fee.
Frequently Asked Questions
What does a laptop actually cost per employee per year?
Rather more than the purchase price divided by the refresh cycle, because six further costs sit behind it: delivery, configuration, support, replacement, recovery and disposal, offset by whatever you recover on resale. For a distributed company the purchase price is commonly between a third and two thirds of the real figure, with the exact proportion depending mostly on how much you ship internationally and how reliably devices come back. The only useful figure is one built from your own records.
What is the most commonly omitted cost?
Devices issued to people who left and never came back, valued at what they were worth at the time. It is usually the largest single addition to the model and the one companies are most reluctant to include, because producing it requires stating the recovery rate explicitly. It is also the line that improves fastest, since better recovery both removes a write-off and adds a resale value.
How many years should we spread the purchase price over?
The real average life your fleet achieves, taken from your own age distribution, rather than the life your policy specifies. Companies commonly have a three-year policy and a fleet averaging four or more, and using the policy figure overstates the annual cost by around a third. If the two differ, that gap is itself worth investigating, since it usually means refresh budgets are covering only the machines that became intolerable.
Should we include staff time in the model?
Yes, for time genuinely spent on configuring, shipping, chasing returns and handling hardware tickets, costed at a realistic internal rate. Do not allocate management overhead, office costs or general IT salary, because that is where these exercises sprawl and stop being finishable. The one piece of time most often missed is the working days an employee loses while waiting for a replacement machine, which frequently exceeds the handling time itself.
How do we compare this against a device-as-a-service quote?
Build your own six-line figure first, then compare like with like, because a vendor quote is an all-in number and a purchase price is not. Score the proposal against your six lines individually and ask which ones it genuinely removes, since a service you already perform well is worth nothing to you regardless of how prominently it appears. Expect the comparison to be much closer than a purchase-price comparison suggests, and expect it to remain a judgement rather than becoming obvious.
Which costs can we reduce without changing anything we buy?
Three, immediately. Recovery improves by asking for equipment in the first week of a notice period rather than after the last day, which raises return rates and therefore moves both the write-off and resale lines. Replacement improves by holding a small number of spare machines in the markets where you hire most, converting urgent international shipments into domestic ones. Delivery improves by buying in the destination market rather than shipping from your home country.
How often should the figure be rebuilt?
Once a year, at the start of the planning cycle, with the date and the assumptions recorded beside it. A model built for one business case and never refreshed becomes quietly wrong, and the real danger is that somebody uses it two years later without knowing how old it is or what it assumed. Rebuilding is much faster than the first attempt, because the data sources are already identified.
HROpsLab takes no vendor money and publishes no paid placements, which is why this page gives you a method rather than a benchmark figure to quote.