"IoT" gets used for everything from a smart bulb to a factory. For a business owner the useful definition is simpler: a physical thing you care about (a tank, a motor, a cold room, a field) sends numbers to a screen, and sometimes the screen sends commands back. This article explains the parts, what they cost, and where they break, without assuming you are an engineer.
The five parts of every IoT system
Every hardware-software project we have built breaks into the same five pieces. If a vendor's proposal does not clearly cover all five, something is missing.
- Sensors and actuators. The sensor measures: temperature, soil moisture, current draw, tank level, door open or closed, GPS position. The actuator acts: a relay that switches a pump, a valve, a siren.
- A controller. A small board (Arduino, ESP32, Raspberry Pi or an industrial PLC) that reads the sensors, does basic logic and talks to the network. It lives with the sensor, often in a plastic box on a pole.
- A network link. How the readings travel from the field to the internet. This is where most decisions and most failures happen; the next section is about it.
- A backend. A server that receives readings, stores them, checks rules (alert if temperature is above 8 degrees for 10 minutes) and exposes the data to apps.
- A dashboard or app. What the owner or operator actually looks at: a web page, an Android app, or simply a WhatsApp message.
The word "gateway" refers to a device that sits between the controllers and the internet, collecting from many nodes over a short-range radio and forwarding over broadband or 4G.
Most of the cost and most of the risk sit in parts one to three, the physical side. Most of the value sits in parts four and five, because that is where a reading becomes a decision. A good vendor is comfortable in both halves, or is honest about which half they are bringing a partner for.
Choosing the network link
| Link | Range and power | Suits | Watch out for |
|---|---|---|---|
| Wi-Fi | Tens of metres, mains power | Inside a shop, clinic, small factory | Router reboots, password changes, no coverage in the godown |
| 4G with a SIM in the device | Anywhere with signal, needs steady power | Vehicles, remote pumps, single sites | Recurring SIM cost, poor indoor signal, SIM paperwork |
| LoRaWAN | 2 to 10 km line of sight, runs years on battery, tiny messages | Farms, large campuses, many low-data sensors | Needs a gateway with internet; no video or heavy data |
| Wired (RS-485, Ethernet) | Inside a building, very reliable | Factory machines, weighbridges, existing PLCs | Cabling cost; not for moving things |
For a farm or a spread-out campus, LoRaWAN plus a single 4G gateway is usually the right answer: sensors in the field run on batteries for a year or more and cost little to operate. For a delivery van, a 4G tracker is the only practical choice. Inside a building, wired or Wi-Fi is cheaper and more reliable than either.
Patchy connectivity is normal in Vidarbha. A well-built device keeps readings in its own memory when the link is down and sends them when it returns; a badly built one loses them.
What the smart agriculture project looked like
One of our three published case studies is a smart agriculture system. Farmers needed crop monitoring, automated irrigation and weather prediction. We built an IoT sensor network feeding machine-learning crop analytics and a web dashboard, using Python, Arduino, LoRaWAN, machine learning and React.
Mapped onto the five parts: Arduino-based sensor nodes in the field, LoRaWAN to carry readings across the distance to a gateway, a Python backend for storage and the analytics models, and a React dashboard where monitoring, irrigation and forecasts come together. We wrote more about that setting in IoT smart farming in Maharashtra.
Beyond farms: where the same pattern applies
- Cold rooms and pharma storage. Temperature loggers with alerts on WhatsApp when a compressor fails at 2 am. One saved batch usually pays for the system.
- Diesel generators and pumps. Run hours, fuel level and fault codes, so service is scheduled by hours run rather than by guesswork.
- Machine monitoring in an MIDC unit. A current sensor on the main motor tells you when the machine is actually running, which is a cheap, honest measure of utilisation.
- Vehicle and asset tracking. GPS plus a door sensor on a delivery van; a tag on a rented crane.
- Hostels, clinics and colleges. Water tank level, energy metering per block, access control.
In every one of these, start with monitoring and add control later. A dashboard that shows the cold room temperature for three months teaches you what normal looks like before you let software switch the compressor. Skipping that step is how a bug turns into a spoiled stock.
Realistic costs
Rough guides for India, hardware and software together, excluding recurring costs:
| Scale | What it includes | Indicative cost |
|---|---|---|
| Pilot: 3 to 10 sensors, one site | Off-the-shelf sensors and controllers, cloud backend, simple dashboard and alerts | ₹1.5 to ₹4 lakh |
| Working system: 20 to 100 nodes | Industrial enclosures, gateway, ruggedised install, app, user roles, reports | ₹4 to ₹15 lakh |
| Multi-site or product-grade | Custom PCB, certification, fleet management, over-the-air updates | ₹15 lakh upwards |
Recurring costs are easy to forget: SIM data at a few hundred rupees per device per month, cloud hosting, battery replacement, and someone to climb the pole when a node stops reporting. Ask for these in the proposal. The real figure depends on the environment and how much of the hardware is standard; a temperature logger in an office is a different problem from a sensor in a wet field.
Where these projects fail
In our experience the failures are rarely in the software.
- Power. Solar panels sized for October that fail in the monsoon. Batteries swapped for a cheaper type. A farmhand unplugging the gateway to charge a phone.
- Enclosures. Moisture, dust, rats and heat. An IP65 box costs a few hundred rupees more than a plain plastic one and is the difference between six months and six years.
- Sensor drift. Soil moisture and pH sensors need calibration; cheap ones drift within weeks. Budget for replacement.
- The network assumption. "There is 4G there", said in the office, is not the same as 4G at the pump house behind the hill.
- Nobody owning the alerts. A system that sends 40 alerts a day is muted within a week. Alerts need thresholds tuned with the operator and a named person who responds.
- Locked hardware. Proprietary devices that only talk to one vendor's cloud. Insist on standard protocols (MQTT, Modbus, LoRaWAN) and on owning the data.
Questions to ask a vendor
- Which of the five parts are you supplying, and which are you assuming already exist?
- What happens to readings when the network is down?
- Who owns the data, and can I export it?
- What is the recurring cost per device per year, all in?
- Can I see a device that has been in the field for a year?
A vendor who has only demonstrated on a desk will struggle with the last one. That is worth knowing before you sign. For the wider picture of how connected devices fit into a business system, see digital transformation for mid-size businesses.
Where SailRight fits
We design and build sensor networks, controllers and the dashboards that read from them, and we integrate existing machines and PLCs into web and mobile apps. Our IoT and hardware integration service starts with a site visit, because the site decides the network and the enclosure. If you have a thing you would like to see on a screen, tell us about it.
- IoT
- Hardware integration
- LoRaWAN
- Sensors
- Smart agriculture
- Dashboards
