POS Spare Parts and DOA Planning for a Hardware Rollout

After-sales for a hardware estate runs on two clocks, and most plans are built on only one of them. The first clock is the 30 day dead-on-arrival window, which is about the journey and the first few weeks of use. The second is the repair turnaround, 7 to 15 working days from the moment a failed unit reaches us to the moment the repaired or replacement unit is dispatched. A lane that fails on day 60 is governed by the second clock, and that is where spare parts planning either works or quietly becomes a support queue.

Two clocks, and one of them is a published number

The dead-on-arrival window is short, absolute and easy to plan for: a unit that fails within 30 days of delivery is replaced rather than repaired. The repair route is the opposite. It is a range, because a repair is a build rather than a swap: production batches carry different component revisions, so a unit built in March may not take the parts pulled for a June build, and those parts have to be procured before the unit is opened.

That procurement step is the reason the number is 7 to 15 working days rather than a single figure. A supplier who quotes one number is either quoting the fast batch or padding for the slow one, and neither is useful when the question is how long a lane stays down. A published range can be planned against, because the planner takes the long end.

Two details sit outside both windows and are worth putting in the plan rather than discovering later: the transit time in each direction, which is not part of either number, and the time it takes the store to report the fault at all. In practice a dead lane is down for the repair window plus two transits plus however many days passed before anyone raised a ticket.

What two percent actually buys

Bulk orders carry a spare parts allowance of two percent of the ordered quantity, calculated on unit count rather than invoice value, and shipped in the same container so it clears the same customs entry as the goods. On a 500 unit estate that is 10 spare units.

Ten units sounds like a lot until it is divided by a failure rate. The allowance is a fixed quantity, so the cover it provides shrinks as the failure rate rises, and the range is wide enough that the same 10 units can be two years of cover or less than five months.

Read the table as a warning about how the number is usually interpreted. Two percent is not a claim about reliability and it is not derived from one; it is a shipping allowance sized by the order. A buyer who treats it as a guarantee that two percent of the estate can fail is reading a quantity as if it were a rate.

The pipeline: how many units are actually in repair

The number that decides how many spares a region needs is not the annual failure count. It is how many units are simultaneously away being repaired, and that follows from the failure rate and the repair window together. Divide annual failures by twelve to get a monthly arrival rate, then multiply by the window in months.

The pipeline is small, and that surprises most buyers the first time they run it. Even at a ten percent annual failure rate on 500 units, only about three units are in repair at once. Ten spare units is therefore far more than the pipeline requires, which raises the obvious question: if the spares are not covering repair volume, what are they for?

They are for site coverage. The pipeline number says how many units are away, not how many sites are waiting for one to come back. Those are different problems, and the second one is the expensive one.

What a dead lane costs, in transactions

A unit that fails and has no replacement behind it takes the lane with it for the whole repair window plus transit. The cost is easiest to see in transactions, because that is the unit the store already measures.

Between 1,800 and 4,200 transactions is what a single uncovered failure costs before any transit time or reporting delay is counted. Put the store's average basket against that number and the argument for a regional spare pool stops being a matter of taste. It also explains why the question "how many spares" gets a different answer from operations than from finance: operations is pricing downtime, finance is pricing inventory, and both numbers are real.

Where the units live

Three places are possible and each has a different failure mode. The choice is not about cost alone, because the point of a spare is the time it saves.

One spare per site is the arrangement everybody wants and almost nobody buys, because the arithmetic is brutal: 50 sites means 50 units, ten times the shipping allowance and a tenth of the estate sitting in a cupboard. The workable compromise is a regional pool sized to the pipeline number with one or two units of buffer, plus an agreement about which sites are allowed to wait. That last part is a business decision, not a hardware one, and it is worth making before the first failure rather than during it.

Wear parts are a different list

A spare unit pool covers failures. It does not cover consumption, and consumption is a separate budget line. Printheads, cutter blades, drawer solenoids, power supplies and cables are consumed by use in a way that a mainboard is not, and a deployment plan that treats them as failures will run out of both patience and parts.

Two of those wear items are worth naming with the page that covers them. A receipt printer consumes a printhead and a cutter, and the duty cycle data for both sits with the thermal receipt printer range. A cash drawer consumes a solenoid and a latch, and the service route for a mechanism that has stopped latching is cash drawer repair rather than a replacement order.

The distinction matters commercially as well. Coverage is defined by category rather than by catalog, so a terminal line, a scanner, a printer and a kitchen display do not share a single term, and what is inside the term differs between them. The warranty and RMA page sets out the category terms, the exclusions and how a claim is opened. The wear parts list belongs in the deployment plan, built from those terms rather than from an assumption that anything inside the period is free.

A practical minimum for a 500 unit estate: one printhead per twenty printers, one cutter assembly per twenty, a drawer solenoid per fifty, and a stock of power supplies and interface cables sized at one percent of the device count. None of that comes out of the two percent unit allowance, and treating it as if it did is how a quarter three parts request turns into a quarter four escalation.

Dead on arrival is a thirty day window, not a quality plan

The 30 day window is generous by industry standards and it is also widely misunderstood. It covers the journey and the first weeks of use, which is where a manufacturing or configuration fault announces itself. It says nothing about month seven, when a failure is more likely to be wear, environment or handling than it is to be the factory.

The split matters because the two clocks lead to different processes. Inside 30 days the question is whether the unit is replaced, and the answer is yes. After 30 days the question is what failed, whether it is covered, and how long the repair takes, and the answer starts with a parts search. A deployment plan that only budgets for the first clock will be surprised by the second, and the surprise always arrives in the same place: the site with no spare behind it.

What the coverage period provides is the backstop for the second clock. Terminals run to 12 months, selected terminal lines and scanners to 24 months, printers to 12, kitchen displays, mini PCs and payment kiosks carry 24 month service coverage, and electronic shelf labels run to three years. The details of what each means are on the ordering terms and the warranty page; the planning consequence is that the repair route, not the replacement window, is what an estate lives inside for most of its life.

The paperwork that decides the outcome

Most slow repairs are not slow in the workshop. They are slow in the first three days, while the two sides work out what the unit is and what is wrong with it. Four items remove almost all of that delay.

• The serial number, from the device label rather than the invoice. It ties the unit to a production batch, which is what makes the parts search possible.
• The failure symptom in the operator's words, plus a photograph. A symptom that says dead is not actionable; one that says screen lights, no receipt, red light flashing is.
• The classification. DOA if inside 30 days of delivery, repair otherwise. Getting this wrong at the start restarts the clock.
• The address the unit returns to, confirmed in writing. A repair that ships to the wrong site costs a second transit and a second week.

One commercial term belongs alongside them and is worth confirming on the quotation rather than assuming: who pays freight in each direction. It is a term of the sale, not a technical detail, and it is the line that decides whether a repair is worth doing for a low value device at all.

A worked plan for 500 units across 50 sites

Assemble the numbers in one place and the shape of the plan becomes obvious. The estate is 500 units at 50 sites, ten units per site. The shipping allowance is 10 spare units.

The review point is the part most plans omit and the part that makes the rest work. The failure rate in this table is an assumption until the estate has run for a quarter, and after that it is a measurement. A plan reviewed at 90 days can be corrected while there is still budget; a plan reviewed at renewal is a post-mortem.

This is the same arithmetic a pilot is meant to produce, which is why the two belong together: the pilot generates a failure rate, and the spares plan consumes one. Designing the pilot covers how to get a number worth using, and comparing quotations covers the line where the spare allowance is stated, since two percent by unit count and two percent by value are not the same promise.

Five lines for the spare parts clause

1. The allowance and its basis. Two percent of the ordered quantity, by unit count rather than by invoice value, shipped with the goods.
2. What the units are. The same model and configuration as the order, not a cheaper or older build, and stated as such.
3. Where they ship. One address or several, and whether the allowance can be split across regions.
4. The wear parts list and its budget. Separate from the unit allowance, with quantities per hundred devices and the first order placed with the rollout rather than after it.
5. The review date. The point at which actual failure data replaces the assumed rate, and who is responsible for reordering at that point.

The clause that causes the most trouble is the second one. A spare pool made of a different build is not a spare pool, because the unit that comes out of the cupboard has to be the unit the store already knows how to use. It is worth writing down, even to a supplier who would never do otherwise, because it is the line an integrator will be asked about when a swap goes wrong at two in the morning. Support and RMA detail, including how a claim is opened and what happens after the coverage period, sits on the support page.

Frequently asked questions

What is the difference between DOA and a normal repair?

DOA is a replacement window, not a repair process. A unit that fails within 30 days of delivery is treated as dead on arrival and replaced rather than repaired, because a failure that early says something about the unit rather than about wear. After 30 days the unit enters the standard repair route, which takes 7 to 15 working days from the moment we receive it to the moment the repaired or replacement unit is dispatched. The two clocks are different, and a deployment plan needs both.

Why does a repair take 7 to 15 working days?

Because the repair is a build, not a swap. Production batches carry different component revisions, so the parts for a unit built in March are not necessarily the parts used in June, and the matching parts have to be procured before the unit can be opened. That procurement step is what stretches the window. A single figure would be either optimistic for the batch with the longer lead or padded for the batch with the shorter one, which is why the range is published as a range.

Is 2% spare parts enough for 500 units?

It depends on the failure rate and on where the units sit, and 2% is sized by the order quantity rather than by either. Two percent of 500 is 10 spare units. At a one percent annual failure rate that is five failures a year, so 10 spares cover two years of replacements. At five percent it is 25 failures a year and the same 10 units cover under five months. The allowance is insurance with a fixed premium, not a prediction.

Where should spare units be held?

Regional rather than central if the estate is spread, because the alternative to holding a spare is holding a dead lane. A central pool adds transit time to a repair that already runs 7 to 15 working days. A spare at every site is not affordable: one unit per site on a 50 site estate of 500 units is 50 units, which is ten percent of the estate and ten times the shipping allowance. The workable shape is a small regional pool plus a swap agreement, not a spare per store.

Which parts wear out and which are warranty?

Wear parts and failures are different categories and the distinction decides who pays. Printheads, cutter blades, drawer solenoids, power supplies and cables are consumed by use, and how each is treated in and out of warranty depends on the product category. The category-by-category terms, including exclusions and how a claim is opened, are on the warranty and RMA page. Build the wear parts list from that, not from an assumption that everything inside the coverage period is free.

How many units will be in repair at the same time?

Multiply the failure rate by the repair window. On 500 units at a two percent annual failure rate you have 10 failures a year, about 0.83 a month, and with a 15 working day window roughly 0.6 units in the pipeline at any moment. At five percent it is about 1.5 units. That pipeline figure is what the spare pool has to cover, and it is usually much smaller than buyers expect, which is why the real problem is site coverage rather than repair volume.