TL;DR
Approval limits exist so that somebody senior sees the purchases worth seeing, and so that nobody waits three days for a keyboard. Most companies set them once, at founding, and never look again. By the time headcount has tripled, the bands are wrong in both directions and people have quietly learned to work around them.
- Three bands is usually right. Two is too blunt, five is theatre.
- Set the bottom band high enough that nobody approves a mouse.
- The band that matters is the middle one: a laptop for one person.
- Name a deputy for every approver, or August breaks the process.
- Review the bands when headcount changes by half, not annually.
- If people route around the limits, the limits are wrong, not the people.
How many approval limits you actually need
Three. Almost always three.
Two bands cannot separate routine spend from a real decision, so either everything needs approval or almost nothing does. Five bands produce a matrix nobody remembers, and people default to asking their manager regardless, which is what you had before.
Best tools for HR Operations
| Band | Who approves | What it covers |
|---|---|---|
| Routine | Nobody | Accessories, cables, a mouse, a headset |
| Standard | The person’s manager | One machine, one monitor, a replacement |
| Material | Finance or a budget holder | Batches, new markets, anything unusual |
The interesting band is the middle one, because that is where a single laptop sits, and a single laptop is both routine enough to happen weekly and large enough that somebody should know.
Where to put the lower line
High enough that nobody is approving consumables. If a manager is signing off a keyboard, the process is costing more in attention than the keyboard costs in money.
Teams consistently set this too low, then watch compliance decay, then blame the people rather than the threshold. The threshold is the thing you control.
Who should sit in each band
Disclosure: RemoAsset, named below, is owned by the same people who publish HROpsLab. It is described with its limitations stated, on the same terms as every other supplier. This site takes no vendor money and sells no placements.
The common mistake is routing by seniority rather than by who has the context.
A manager knows whether their new designer genuinely needs the heavier machine. A finance lead does not, and asking them to decide produces either a rubber stamp or an argument about something they cannot assess.
So push the specification decision down to whoever understands the work, and keep the budget decision where the budget sits. Those are different questions and they get conflated constantly.
Deputies are not optional
Every approver needs a named alternate, written down, with the same authority. Without one, the process has a predictable annual failure every time somebody takes two weeks off.
What happens instead is that people order anyway and paper it afterwards, which teaches everyone that the process is advisory. One holiday can undo a year of discipline.
The turnaround commitment
Set a service standard on yourselves: approvals answered same day, or the request escalates automatically to the next band.
That sounds generous to the requester and it is actually protection for the approver, because it converts a vague expectation into a bounded one. It also removes the main reason people bypass the system, which is that waiting is worse than asking forgiveness.
When to move the lines
Not annually. Annual reviews happen when nothing has changed and get skipped when everything has.
Review when headcount changes by half, when you enter a new market, or when you notice people routing around the bands. Those three events cover almost every case where the numbers have drifted out of usefulness.
The signal to watch for is not complaints. It is silence plus receipts. When expense claims for equipment start appearing, the approval process has become slower than a personal credit card, and people have made a rational choice.
What to look at during a review
- How many requests fell in each band. If ninety per cent were routine, the bottom line is too low.
- How many were rejected. A rejection rate near zero means the band is ceremonial.
- How long each band took. The slow band is where the bypassing starts.
- What arrived without approval. Count it honestly rather than treating it as noise.
Four numbers, thirty minutes, once or twice a year. That is the whole exercise.
The decision table
| Situation | Scale | Setup | Primary pain | Recommended starting point |
|---|---|---|---|---|
| Everything goes to the founder | Under 30 | No bands | Founder is a bottleneck | Add a routine band immediately |
| Managers approve everything | 30 to 100 | Two bands | No visibility on big spend | Add a material band above |
| Expense claims appearing | Any | Bands exist | Process slower than a card | Set a same day standard |
| Approvals stall in August | Any | One approver per band | Single point of failure | Name deputies in writing |
| Nothing ever rejected | Any | Mature process | Bands are ceremonial | Raise the lines until they bite |
What to do when somebody spends without approval
It will happen, and how you respond the first time sets the pattern for everybody watching.
The instinct is to make an example. That works once and then costs you honesty, because the next person who bypasses the process has a strong incentive to hide it rather than mention it.
A better sequence is to pay the invoice, then ask one question: what made the process slower than ordering directly? The answer is nearly always a specific friction. The approver was away, the request form asked for a cost centre nobody knew, or an urgent replacement could not wait two days.
Each of those is fixable, and fixing it prevents the next ten breaches rather than punishing the first.
When it is not a process problem
Sometimes the answer is that somebody wanted a particular machine and knew they would not get it. That is a different conversation and it belongs with their manager rather than with finance.
The way to tell the difference is whether the purchase matched the policy specification. A standard laptop bought outside the process is a speed problem. A premium machine bought outside the process is a decision somebody made for themselves.
Record both kinds in the same log and review them together. After two quarters the pattern is usually unambiguous, and it tells you whether to fix the process or have four specific conversations.
What approval limits do not solve
They control the decision to spend. They do nothing about what happens next, and teams often expect more from them than they can deliver.
They do not tell you whether the right machine was bought. They do not track whether it arrived. They do not notice when it never came back from somebody who left.
Disposal is the clearest example. A device approved, bought and issued still has to be collected and sanitised at the end, and the standard most contracts name for that is NIST Special Publication 800-88 on media sanitization, which no approval band touches.
Those are lifecycle problems and they need a different instrument. Approval limits sit at one moment in a chain that runs from request to disposal, and a tidy front end with no visibility afterwards is a common and expensive shape.
Where buying, delivery, retrieval and disposal run through one supplier, the record from approval through to wipe certificate stays joined up, which is the practical case for handling the lifecycle as one chain. The limit worth stating plainly: a supplier does not set your approval policy, and it will not stop somebody buying the wrong specification. It only means you can see what happened afterwards.
For the accounting side, equipment above a capitalisation threshold is treated as an asset rather than an expense under IAS 16 on property, plant and equipment, so your approval bands and your capitalisation threshold should at least know about each other. They are usually set by different people who have never spoken.
For who ultimately signs, see our guide to who signs off equipment spend.
See how the record stays joined from approval to disposal at laptop lifecycle management, or read our guide to who signs off equipment spend.
Frequently asked questions
How many approval limits bands should we have?
Three. Two cannot separate routine spend from a real decision, so either everything needs sign off or almost nothing does. Five produce a matrix nobody remembers, and people default to asking their manager anyway, which is where you started. Routine with no approval, standard at manager level, and material at finance covers almost every company under a thousand people.
Where should the lowest threshold sit?
High enough that nobody is approving consumables. If a manager is signing off a keyboard, the process costs more in attention than the item costs in money. Teams reliably set this line too low, then watch compliance decay and blame the people. The threshold is the part you control, so move it until the routine band genuinely covers routine things.
Who should approve a single laptop?
The person’s manager, because they know whether the heavier machine is genuinely needed for the work. Finance cannot assess that and will either rubber stamp it or argue about something they have no basis to judge. Keep the specification decision with whoever understands the job and the budget decision where the budget sits, because those are different questions that get conflated constantly.
What if an approver is on holiday?
Name a deputy in writing with the same authority, for every approver, before it happens. Without one the process fails predictably every August, and what people do instead is order anyway and paper it afterwards. That single fortnight teaches everyone the process is advisory, which undoes a year of discipline for no benefit.
How do we know the bands need changing?
Watch for expense claims for equipment. That is the clearest signal, because it means somebody decided a personal credit card was faster than your process, which is a rational choice rather than misbehaviour. Also review when headcount changes by half or when you enter a new market. Annual reviews happen when nothing changed and get skipped when everything did.
What should we measure at a review?
Four things, in about thirty minutes. How many requests fell in each band, because ninety per cent in the routine band means the bottom line is too low. How many were rejected, because a rate near zero means the band is ceremonial. How long each band took, because the slow one is where bypassing starts. And what arrived with no approval at all.
Do approval limits control equipment cost?
They control the decision to spend and nothing after it. They will not tell you whether the right machine was bought, whether it arrived, or whether it came back when somebody left. Those are lifecycle questions needing different instruments. A tidy approval process with no visibility afterwards is a common shape and an expensive one, because the losses happen downstream of the control.
