When we started Oria, the obvious shortcut was to skip hardware entirely. Buy a generic industrial 4G gateway, add a NMEA interface, flash our firmware, ship it. On paper it saves eighteen months and a lot of capital. In practice, every prototype we tested that route with came back from sea trials with the same three problems: water where it shouldn't be, a NMEA bus that talked but didn't speak, and a modem that decided to sulk twenty miles offshore. So we built our own. Here is the honest reasoning behind that decision, and what it changes for you as an owner.

The marine environment eats generic hardware

Industrial IoT boxes are usually rated for warehouses, cold chain, or roadside cabinets. That means IP54 or IP65 at best, plastic that yellows under UV, and connectors designed for a technician in a hi-vis vest, not a skipper leaning into an engine bay with wet hands.

A boat is a specific kind of hostile. You have salt fog, condensation cycles when the engine cools down, vibration from a diesel at 2,800 rpm, and voltage that can spike ugly when the alternator kicks in or the windlass draws its full load. A generic box survives the first season. It rarely survives the second.

We specified the Oria Box for that reality:

  • IP67 housing, so a wave over the coachroof is a non-event.
  • Wide-input DC supply with proper transient protection, because 12 V on a boat is rarely a clean 12 V.
  • Operating range that covers a closed engine compartment in August and a mooring in Brittany in February.
  • Connectors that lock, are keyed, and can be mated without a torch in your teeth.

None of this is glamorous. It is the difference between a product that logs your season and one that becomes a warranty ticket.

NMEA is not actually plug and play

On the brochure, NMEA 2000 is a standard. In the engine room, it is a standard the way "European plug" is a standard: mostly, but with enough regional and vendor-specific behaviour to ruin your afternoon. Different engine manufacturers expose different PGNs. Some transmit fuel rate cleanly, others only give you instantaneous consumption you have to integrate yourself. Some multiplex GPS at 1 Hz, others at 10 Hz. NMEA 0183 legacy talkers still exist on plenty of hulls and need their own listener with the right baud rate and sentence filters.

A third-party box gives you a serial stream. That is not the same as understanding your boat. To turn raw frames into useful maintenance intelligence, you need the parser, the calibration layer, and the semantic model to live in the same firmware you control. Otherwise you spend forever writing glue code on a server for edge cases that a well-designed gateway would handle at source.

Owning the hardware meant we could:

  • Build a NMEA stack that auto-detects the buses present and adapts sampling rates.
  • Buffer locally when the 4G link drops, then reconcile without gaps in the voyage replay.
  • Push firmware updates that improve support for a new engine ECU without asking you to swap anything.

Cellular connectivity at sea is not the same problem as cellular connectivity on land. Coverage is patchy near coasts and non-existent offshore. Handover between towers is aggressive when you are on a moving platform. Roaming across European operators is a legal and technical minefield if you get the SIM strategy wrong.

A generic box ships with whatever modem the OEM sourced last quarter and a SIM slot. That is not a solution, it is a starting point. We wanted the antenna path, modem firmware, SIM provisioning, and reconnection logic under one roof. The Oria Box uses a multi-operator SIM included by default, chosen so it hops networks in the ports and coastal zones where your boat actually lives. When the signal drops, the box does not panic. It stores, it retries with backoff, it sends deltas rather than blobs when bandwidth is thin.

That matters for the practical features you actually use : anti-theft alerts that need to fire in the first seconds of a movement outside a geofence, voyage data that has to reconcile after a week in a dead zone, engine hours that must not double-count when the link recovers. This is exactly the kind of edge case we ran into managing the fleet at the Paris 2024 Olympic Marina, where 300 boats had to be tracked and maintained in parallel.

Firmware you can actually trust

The uncomfortable truth about buying somebody else's IoT box is that you also inherit their firmware, their update cadence, and their security posture. Half of the "industrial" gateways on the market ship with default credentials still active, unsigned firmware, and a web interface that hasn't been audited since it left Shenzhen. For a device that sits on your boat's electrical network and reports your position to a server, that is not acceptable.

By building our own, we control:

  • Secure boot and signed firmware, so a box in the field cannot be flashed with anything we haven't signed.
  • Encrypted transport for every telemetry frame, with keys provisioned per unit at manufacture.
  • An update pipeline that lets us patch a vulnerability in days rather than begging a supplier to prioritise us.
  • Data residency in France and the EU, which matters both for GDPR and for the professional customers whose contracts require it.

None of this is possible when you are the fourth customer on someone else's roadmap. Institutions like SNSM, IFREMER, and the professional operators we work with, including the fleet at Team Yachting, do not deploy hardware they cannot audit. Neither should you.

Made in France, because the supply chain matters

The Oria Box is designed in France and assembled in France. This is a deliberate choice, not a marketing line. It gives us three things that matter for a product that has to last a decade on a boat:

  1. Component traceability. When a batch of modems or connectors shows a weakness, we know which units are affected and can act. That is impossible with an OEM white-label where you buy what shipped that month.
  2. Repairability. We keep spares. We service units. A five-year-old Oria Box is not e-waste, it is a warranty case with a known BOM.
  3. Iteration speed. When our engineers spot something on a test bench in the morning, the assembly line can validate a fix in the afternoon. That loop simply does not exist with an overseas supplier at MOQ 10,000.

The other benefit, less visible but real, is that we can say honestly to a customer that no part of their data pipeline, from the sensor to the dashboard, sits in a jurisdiction we did not choose.

What owning the stack lets us do for you

The point of all this is not the hardware itself. Nobody buys an IoT box because they love IoT boxes. They buy it because they want fewer surprises: fewer breakdowns, fewer thefts, fewer arguments about who did the last oil change, fewer hours reconstructing a season from paper logs.

Owning the box and the platform lets us close that loop end to end. When you look at a voyage replay in the Oria app, the timestamps line up because the same team wrote the firmware clock discipline and the server ingestion. When a geofence fires, the latency is what we say it is, because we measured every hop. When your engine hours trigger a maintenance reminder, the number is right because we calibrated the counter against the ECU, not against a generic pulse counter.

If you are thinking about how connected equipment fits alongside the rest of your electronics, from your VHF setup to your chartplotter, the same principle applies : the boxes that survive are the ones designed for boats first, and everything else second.

Building our own hardware cost us time we could have spent on features. It also means that when a customer calls in November because a box in a winterised hull is behaving oddly, we can open the schematic, the firmware, and the server logs on the same screen and give them a real answer. That, more than anything on a spec sheet, is why we did it.