Choosing GPS hardware for a tracking business

Specifications are not the deciding factor

Datasheets converge. Almost any current tracker will report a position, a speed and an ignition state, and the differences that appear in a comparison table are rarely the ones that cost you money.

What separates hardware in practice is how it behaves at the edges: what it does when the network drops, how it reports power loss, whether it buffers while out of coverage, and whether its firmware is stable across the units in one shipment.

Questions worth asking before you buy

• Is this exact model and firmware decoded in production by the platform you intend to use — not the manufacturer, the model.
• Can the server endpoint be changed remotely? If not, every future platform decision requires physical access.
• What happens out of coverage: does it buffer and replay, or silently drop?
• How does it report external power loss, and how quickly? This is the difference between a theft alert and a report.
• Will the vendor supply a packet capture, or a single unit to test, before a volume order?

Standardise on fewer families than you think

Every additional device family multiplies support burden: another decoder to depend on, another firmware release cycle, another set of quirks your support team has to learn.

Operators who standardise on one or two families and stay there consistently spend less time on hardware problems than those who buy opportunistically on price.

A cheaper tracker that needs a new decoder is not cheaper. Price the integration alongside the unit.

Power and battery behaviour

For hardwired trackers, the question that matters is what happens when external power is cut — whether the device reports the loss, how fast, and whether it keeps reporting on internal backup. That behaviour is the difference between a theft alert and an after-the-fact report.

For battery-powered asset trackers, the trade is reporting frequency against months of life, and vendors quote life at intervals nobody uses in production. Ask what the figure assumes, then assume worse.

SIM and connectivity

Hardware choice and connectivity choice are usually made separately and then discovered to be coupled. A tracker locked to a carrier or a bundled SIM constrains where you can deploy and what you pay per device per month, sometimes for longer than the hardware lasts.

Also worth confirming: which network generations the modem supports. 2G shutdowns have already stranded fleets of otherwise functional trackers, and a cheap unit today is worth nothing if its network is scheduled for retirement in the region you sell into.

Installation is a cost you are quoting for

Hardwired trackers need someone under a dashboard. That labour is usually a larger line on the customer invoice than the unit, and it varies enormously by device — a tracker with a short harness, awkward connector or fussy antenna placement turns a twenty-minute job into an hour.

It also determines who can install. If a device needs the ignition wire identified correctly to report ignition at all, you are relying on installer skill for a feature you are selling. OBD-port devices avoid most of this and buy it back in other ways: they are removable, which matters when you are tracking an asset someone may not want tracked.

Price the install alongside the unit, and ask the vendor how long a competent installer takes.

Failure rate and what happens after it

Some proportion of units will fail, arrive dead, or be pulled from a vehicle and returned. What matters commercially is not the rate so much as who absorbs it.

Ask what the warranty period is, who pays freight in each direction, whether the vendor advance-ships a replacement or waits to receive the faulty one, and what evidence they require. A vendor who wants the unit back before shipping a replacement is asking your customer to go untracked for the round trip, which will be your problem rather than theirs.

This is worth handling as a real process rather than ad hoc — returns accumulate, and an operation with no record of which IMEIs went back is an operation that pays for the same failure twice.

The RMA terms matter more than the failure rate. A 2% failure rate with advance replacement costs less than 1% with three-week round trips.

Regulatory approvals are per-market

A device certified for one region is not thereby certified for another, and carriers can and do refuse to activate uncertified hardware on their networks.

Ask which approvals the exact model holds for the markets you sell into, and get it in writing before a volume order. This is the kind of question that costs nothing to ask and strands a shipment when skipped — and unlike a decoding gap, no amount of platform work fixes it.

A pre-order checklist

Before a volume order, have written answers to:

• Exact model and firmware version being shipped — not the family.
• Confirmation that this firmware is decoded in production by your platform.
• Whether the server endpoint is remotely reconfigurable.
• Out-of-coverage behaviour: buffer and replay, or drop.
• External power loss reporting, and how quickly.
• Network generations supported, and their retirement timeline in your markets.
• Whether the vendor will supply one unit and a packet capture first.

Accessories change what you are selling

Trackers are often quoted with add-ons — relay outputs, panic buttons, temperature probes, driver-ID readers. Each one changes the product more than the line item suggests.

Relay outputs are the clearest case. A relay wired to interrupt a starter circuit turns a tracking product into something that can prevent a vehicle from starting. That is a genuine capability with genuine demand in lending and rental, and it carries obligations that a map does not: who is authorised to trigger it, what happens if it fires while the vehicle is moving, what the audit trail looks like afterwards, and what your jurisdiction requires by way of notice to the driver.

Sensors are milder but not free. A temperature probe means someone will eventually rely on that reading for a cold-chain claim, which makes calibration and gaps in the record your problem.

Decide deliberately which of these you sell. Each is a support surface, a liability question and a reason a customer calls — not merely a part number.

What a pilot has to prove

Buying a handful before a container is obvious advice. Less obvious is that a pilot which only proves "positions arrive" has proven almost nothing — that works on nearly every device on the market.

A pilot worth running answers the questions that cost money later:

• Does it report ignition correctly when wired by someone who is not you?
• What happens across a full weekend parked, and does it drain a vehicle battery?
• What arrives after driving through a coverage gap — buffered history, or silence?
• Does a remote configuration change actually take effect, and can you tell that it did?
• How long does an installer genuinely need, timed rather than estimated?

Run it for at least two weeks on real vehicles doing real work. A device on a desk demonstrates that the radio works, which was never the risk.

Verify support properly

Treat a manufacturer logo on a compatibility page as marketing until proven otherwise. The question to ask is whether a decoder exists for that family and is running in production today.

Nexara publishes its supported list by manufacturer with the protocols behind each, and deliberately omits families it cannot decode — including ones that appear elsewhere in the platform as configuration. A support claim that fails on first contact costs more than an honest gap.