Which ATM route repairs are worth doing yourself

What to take away

A machine that will not dispense is earning nothing while it still holds your cash, occupies a site and tests a host's patience. This guide covers what breaks, which repairs an operator can finish at the cabinet, how to choose between fixing and swapping, and the security work that runs alongside.

Faults, sorted by what they cost

Seven fault classes cover most of what stops a machine. The table sorts them by what they cost you, not by what the part costs.

Two of the seven are not hardware faults at all. Cash-outs and consumables are planning failures, and they are the most common reason a machine is not working on any given day.

Four classes, card reader, note path, printer and power, produce the repairs an operator can usually finish on site. The rest either need a technician or need a change to how the route is planned.

Repair, swap, or replace

Three options, and the right one depends on the site rather than on the fault.

Repair on site. Suits simple, well-understood faults where the part is in the vehicle and the venue can tolerate you working there. It keeps the machine in place, and it ties you up.

Swap the unit. Take a working machine out, bring the faulty one back, and repair it at leisure. That is usually right at a busy site, because it turns a long outage into a short one. It requires a spare, a capital decision covered in the inventory and sourcing guide.

Replace the machine. When the fault pattern repeats, when parts are ending, or when the hardware can no longer be kept current with what your processor requires. Ask that last question before you buy any older unit rather than after.

The decision rule is simple to state and hard to follow: pick the option that gives the shortest outage where outages cost most, and the cheapest option everywhere else.

Most operators do the reverse without noticing. The fault at the busy site arrives at an inconvenient moment, so it gets the easiest response that afternoon.

The repairs worth doing yourself

A repair earns a place on your list when the part is cheap, the job is short, and nothing about it needs your processor, a calibration or a seal. Seven jobs fit that description on most machines.

Note path jams and pick rollers. Open the fascia and clear the note path. On most front-access machines, including the NCR SelfServ and Diebold Nixdorf Opteva series, the pick rollers and belts come out once the cassettes are removed. A roller and belt kit typically costs $60 to $200 in most markets as of 2026, and it cures double picks, skewed notes and repeat jams. Clean the note path sensors with a lint-free swab while you are in there.

Card reader module. An EMV card reader module typically costs $200 to $600, and the physical swap takes 15 to 30 minutes. The catch is key injection. Your processor has to re-inject encryption keys before the machine earns again, and opening the reader housing can trip a tamper sensor that then needs clearing.

Receipt printer and paper. Many terminals use a thermal mechanism of the Epson TM-T88 type. A replacement printer unit typically costs $150 to $400, and thermal rolls typically cost $2 to $5 each in case quantities. Clean the print head first, and replace the unit when receipts come out blank or streaked.

UPS battery. A 12 volt sealed lead acid battery typically costs $30 to $90 and takes 10 to 15 minutes to swap. It cures random reboots, intermittent comms and odd faults that appear on one machine only. Recycle the old battery.

Encrypting PIN pad. A PCI-approved encrypting PIN pad typically costs $300 to $800. The swap itself is mechanical. The key injection is not, and your processor does it.

Cabinet lock and keys. Locks and key sets typically cost $20 to $80. A stiff lock at a site is a security item rather than a maintenance nuisance, so replace it instead of spraying it.

Cassette shutter and latch parts. Springs and latches for the cash cassette typically cost $30 to $100. A cassette that will not latch cleanly causes retracts and partial dispenses, which look like dispenser faults.

A field service visit typically costs $150 to $400 plus parts in most markets as of 2026, and contract terms vary. That is the number to beat against the part price. On a $60 roller kit the answer is obvious. On a $600 reader that needs key injection, the arithmetic is closer.

Leave the safe internals, dispenser calibration, firmware and key injection to a technician. Anything that breaks a tamper seal, or that your processor's agreement reserves to its own engineers, goes back to them.

Which spares to hold

Not the cheapest and not the most frequent. Hold the parts whose failure stops all withdrawals and whose lead time is long, then the parts that fail often enough to be predictable.

On most routes that means one card reader, one thermal printer mechanism, a roller and belt kit, a PIN pad and a spare 12 volt battery, plus paper and cleaning cards. Match the part numbers to your own machines rather than to a general list.

The cost that justifies a spare is missed withdrawals during the wait, so a busy weekend site justifies stock that a quiet office lobby does not. Work it out per site rather than for the route as a whole.

The four fronts of security

The machine itself. Physical attack on the safe and the fascia, and attempts to attach anything to the card entry. The counter is a machine chosen and installed sensibly, a site with lighting and supervision, and a person who looks at the machine properly on every visit rather than glancing at it.

Your processor and sponsoring bank will have requirements of their own about the machine, its components and its software. Keep those in writing rather than in memory.

The data. What the machine handles, what your monitoring and reporting systems hold, and who can reach them. Small operations underrate this because the machine feels like the asset. The NIST small business quick-start guides are a reasonable starting point for how a business this size approaches it.

The run. The most dangerous part of the trade, and the one that gets casual after an uneventful year. Vary the timing and the sequence. Carry only what the run needs.

Tell someone your route and your expected finish. Decide in advance that anything wrong at a site means leaving rather than assessing. The routing consequences of that discipline sit in the operations and workflow guide.

The site. Lighting, sightlines, what has been stacked against the machine, and whether the position still matches what you agreed.

Position also carries obligations independent of security. The 2010 ADA Standards for Accessible Design include provisions addressing automatic teller machines, so review the 2010 ADA Standards with advice.

Treat a position that has quietly moved as a matter to raise, not to accept. Choosing sites with this in mind is part of the site selection work.

When something happens

Write the order of calls down now, on one page, and keep it where you can reach it from the road.

Emergency services first if anyone is at risk. Then your processor, your sponsoring bank, your insurer, and law enforcement, in whatever order your own register says. Then the venue.

Preserve everything, including the machine's records and anything you found. Write down what happened while it is fresh, because your account will be needed and memory is not evidence.

What you owe, to whom, and how quickly is not something a general article can tell you. It comes from your processor, your sponsoring bank, your insurer and your own regulatory position. It belongs in the register described in the licensing and compliance overview. Ask each of them now, not after.

Counting the cost properly

The bill for a fault is not the part. It is the part, the trip, your time, and every withdrawal that did not happen while the machine was out. At a busy site the last item usually exceeds all the others together.

An outage at your best placement needs a different response than the same outage at your weakest. Put the cost of downtime in the same arithmetic as the surcharge and the host share.

Work it out per machine. If a placement does 40 withdrawals a day at a $3 surcharge, a full day down forgoes about $120 of fee income before the host share.

the pricing and profit guide puts the cost there.

Building the habit

Three practices carry most of the benefit, and none of them needs a system.

It requires that the log exists and that somebody reads it, which is the discipline that separates a route that gets easier from one that gets worse. The wider view of how that scales is in the overview of how a route earns.

Common questions

Should I learn to repair the machines myself?

Learn to clear the common faults and to diagnose the rest well enough to describe them accurately. The seven jobs above are worth learning. Beyond that, a technician who does it every day will be faster and cheaper than your afternoon.

Which repairs should I not attempt?

Anything inside the safe, dispenser calibration, firmware, and any job that needs key injection or breaks a tamper seal. If your processor's agreement names its own engineers for a task, that task is theirs.

How do I know if a machine has been tampered with?

By knowing exactly what it should look like and checking it every visit, and by treating anything unfamiliar as a stop-work event rather than a puzzle to solve. If you are unsure, do not handle it, and follow your incident procedure.

Is a newer machine more secure?

Newer hardware is generally easier to keep current with what processors and networks require, which matters. It does not remove the need for a well-chosen site and a person who looks.

What is the most common cause of downtime?

For most routes it is the machine running out of cash, followed by note path jams. Both are managed by planning rather than by repair skill.