IoT development: cost, pitfalls, and what to build first

I have seen enough IoT projects to know the pattern. The demo works. The dashboard looks good. Then the first real devices go into the field and everything goes quiet. Batteries drain in three days. Devices drop offline whenever the network dips. A firmware update bricks a batch of units two time zones away. None of this shows up in the pitch deck.

That gap between the demo and the field is where IoT budgets go to die, and it is also where a competent team earns its fee. Let me walk through what IoT development actually involves, where the money goes, and how to scope a first build that survives contact with the real world.

Where IoT projects actually stall

The cloud side of IoT is the easy part now. AWS IoT Core, MQTT brokers, time-series databases. None of it is exotic, and a competent team can stand it up in weeks. The hard parts are boring and physical.

Connectivity comes first. A device that behaves on your office Wi-Fi will misbehave on a rural cellular network, and a device that works on 4G in one country may not roam cleanly across borders. Power is second. If a sensor runs on a coin cell, every extra wake-up shaves weeks off its life, and most teams only notice when the fleet starts dying in the field. Firmware updates are third. You will ship a bug in firmware, and you need a way to patch it over the air without turning thousands of units into paperweights.

Fleet management belongs in the same bucket. Twenty devices are a hobby. Five thousand devices are an operations problem. You need remote config, health monitoring, and a clear picture of which units are offline and why. That layer takes real engineering, and clients rarely budget for it up front.

Security gets waved off more than it should. A device that talks to your backend is a door into your network, and the default posture for most cheap modules is wide open. Encrypting traffic, rotating keys, and locking down the provisioning flow are not optional once real customer data is involved.

Data volume surprises people too. A modest fleet of a few hundred sensors can push out millions of readings a month, and a backend that handled the pilot fine will sag under full deployment if nobody thought about retention and storage early.

What a first build actually costs

Costs swing a lot depending on how mature the hardware is. Building firmware for an existing board is far cheaper than designing a custom PCB and getting it certified. Here is a rough split based on projects we have delivered.

These are ranges, not quotes. The real driver is hardware risk. Every custom board has to pass certification, survive supply chain hiccups, and be testable in production. If you can start from an off-the-shelf module or a reference design, do it. The savings are bigger than they look because they also cut the timeline.

How to scope a first build that ships

Start with one device and one use case, in one region. Do not build a platform on day one. Pick the single most valuable telemetry stream or control action, get it working end to end, then expand. Most teams over-build the dashboard and under-build the firmware, and it is always the firmware that kills the timeline.

Time zones matter more than people expect when hardware is involved. A device in Brazil, firmware engineers in China, and a client in South Africa means the team has to be used to working across hours. We run overlapping sessions and keep a dedicated group chat per project, in English and Portuguese, so a field issue does not wait twelve hours for a reply. That rhythm is part of why we have delivered projects for clients in Brazil, South Africa, Singapore, and the US.

The other thing I tell clients is that we are not picky about industry. We have built for agriculture, property, and logistics, and we will take on any legal vertical. We have shipped financial systems, property systems, and payment rails before, so the IoT layer plugs into software that already runs a business.

If your IoT idea is still on a whiteboard, the cheapest thing you can do is spend an hour talking it through with someone who has shipped one. That conversation usually surfaces the two or three things that would otherwise stall the project six months in.

Talk to us about your project