Key takeaways
- The real difference is ownership, not protocol. With LoRaWAN you build and own the network with no per-device fees. With NB-IoT you rent a carrier's network and pay per device, per month, forever.
- India's LoRaWAN band is 865–867 MHz, delicensed by the Department of Telecommunications under GSR 564(E). Hardware configured for EU868 or US915 is not appropriate here.
- Economics cross over with scale and time. A gateway is a fixed cost amortised across every device behind it; an NB-IoT subscription is linear in devices and linear in time, so it never gets cheaper. Model the full deployment life, not year one.
- NB-IoT wins on building penetration. For devices in basements, lift shafts and utility cupboards, paying the carrier to own the coverage problem is a fair trade.
- Budget for a second gateway. Nearly every deployment discovers a location the first gateway cannot reliably hear. Higher spreading factors extend range but cost battery and airtime.
Two philosophies, not just two protocols
LoRaWAN and NB-IoT both solve the same problem: get small amounts of data from many battery-powered devices over long distances. They solve it in fundamentally different ways, and the difference is less technical than it first appears. It is about ownership.
With LoRaWAN you build and own the network. You buy gateways, you site them, you run them, and nobody sends you a bill per device. With NB-IoT you rent the network. A carrier has already built it, it works, and you pay per device per month for as long as the device lives.
Most of the decision follows from that single distinction.
LoRaWAN in the Indian context
India allocates the 865 to 867 MHz ISM band for licence-free use, and this is where LoRaWAN operates domestically. Get this right when sourcing hardware: modules configured for EU868 or US915 are not appropriate here, and this catches teams out surprisingly often.
The strengths that matter in practice:
- No recurring fees. Once the gateway is up, adding a device costs you the device. This changes the economics completely at scale.
- Exceptional battery life. Ten years is a realistic target for a sensor reporting a few times a day.
- Genuine long range. Ten to fifteen kilometres in open rural terrain with line of sight, which is transformative for agricultural deployments.
- You own the infrastructure. Your data does not traverse a carrier's network unless you choose to send it there. For some clients this is a compliance requirement rather than a preference.
The weaknesses are the mirror image. You are now responsible for coverage. If a sensor sits in a spot your gateway cannot hear, that is your problem to solve, with your money. Range figures quoted in datasheets assume line of sight, and a hill, a building or a dense stand of trees will take a large bite out of them. Building penetration is mediocre, so a device in a basement may simply not work.
NB-IoT in the Indian context
NB-IoT runs in licensed cellular spectrum on infrastructure the carriers have already built. What you are buying is reliability and reach that you did not have to construct.
- Building penetration. This is NB-IoT's standout advantage. It was designed to reach devices in basements, lift shafts and utility cupboards, and it does.
- Quality of service. Licensed spectrum means no unlicensed neighbours interfering with you. Delivery is far more predictable.
- Coverage you did not build. Wherever the carrier has coverage, your device works. No gateway siting, no site surveys, no roof access negotiations.
- Simpler devices. No network to design or operate. The module talks to the carrier and you get data.
The costs are real and they are permanent. Every device carries a subscription, month after month, for its entire operational life. At ten devices this is a rounding error. At ten thousand devices it is the dominant line item in your operating budget, and it never goes away.
The economics, plainly
LORAWAN
Roughly $5–15 per device, $500–2000 per gateway, and no monthly fees. High upfront cost, near-zero marginal cost per device thereafter.
NB-IoT
Roughly $8–20 per device, no gateway to buy, and $2–5 per device per month. Low upfront cost, permanent recurring cost.
The crossover is what matters. A LoRaWAN gateway is a fixed cost that gets amortised across every device it serves. The more devices behind a gateway, the cheaper each one becomes. NB-IoT's cost is linear in devices and linear in time, so it never gets cheaper.
Run the numbers over the deployment's actual lifetime, not its first year. A project with a thousand devices and a ten-year life makes the answer obvious in a way that a first-year cost comparison actively conceals.
How we actually choose
Our deployments across Kerala have converged on a hybrid pattern, and the logic behind it generalises well.
LoRaWAN for rural and semi-urban sites. Agricultural monitoring, water level sensing, remote environmental stations. These are the conditions LoRaWAN was built for: sparse, spread out, line of sight available, no cellular economics that make sense at that density. We site a gateway on the tallest available structure and cover a large area for one fixed cost.
NB-IoT for dense urban deployments. Devices inside buildings, in basements, in places where we cannot control the site and cannot guarantee a gateway will hear them. Here we are paying the carrier for a coverage problem we do not want to own, and that is a fair trade.
The decision is rarely ideological. It is a question of who should be responsible for coverage at this particular site, and what that responsibility costs over ten years.
Coverage planning: the part people underestimate
If you choose LoRaWAN, you have taken on responsibility for coverage, and it is worth being clear about what that involves before you commit.
The range figures in datasheets assume line of sight. Reality intervenes. A hill between the sensor and the gateway takes a large bite out of your link budget. So does a dense stand of trees, particularly wet foliage, which absorbs 868 MHz signals more effectively than most people expect. Monsoon conditions in Kerala measurably degrade links that worked fine in dry weather, and a network commissioned in March can develop dead spots in July.
Gateway height is the single most effective lever you have. Elevation buys you line of sight, and line of sight buys you range. A gateway on a rooftop or a mast will comfortably outperform one at ground level, often by a factor that makes the difference between one gateway and three.
Plan for this properly. Do a site survey with a real device rather than a coverage map. Walk the boundary of your intended deployment area and confirm the link actually holds where you need it. Then place your sensors, not the other way round.
Budget for a second gateway. Almost every deployment we have run discovered at least one location the first gateway could not reliably hear. Gateways are cheap relative to the cost of sending an engineer back to a site repeatedly. Assume you will need more than one.
Spreading factor, and the trade you cannot escape
LoRaWAN lets a device trade data rate against range through the spreading factor. A higher spreading factor spreads the signal over more time, which makes it detectable at lower power and therefore reaches further. It is the mechanism behind LoRaWAN's remarkable range.
It is not free. A higher spreading factor means the device transmits for longer, which costs more energy per message and occupies the channel for longer. Because the 865 to 867 MHz band carries duty cycle obligations, airtime is a genuinely finite shared resource, and a handful of distant nodes shouting at maximum spreading factor can consume a disproportionate share of your network's capacity.
The practical consequences are worth stating plainly. Devices at the edge of coverage cost more battery and more airtime than devices near the gateway. Adding a gateway does not just fix dead spots; it lets distant nodes drop to a lower spreading factor, which improves both their battery life and your overall network capacity. And a network that works comfortably with fifty nodes can hit a wall at five hundred if too many of them are far away.
Adaptive Data Rate, where the network tells each device the lowest spreading factor it can get away with, exists precisely to manage this, and it should be enabled unless you have a specific reason not to.
A short decision guide
- How many devices, and for how long? Large fleets with long lives push hard towards LoRaWAN, because subscription costs compound.
- Where physically are the devices? Deep indoors or underground pushes towards NB-IoT. Open fields push towards LoRaWAN.
- Do you control the sites? If you cannot put a gateway where it needs to go, LoRaWAN's economics collapse.
- What is the battery target? Multi-year life on a small cell favours LoRaWAN.
- Does the data have residency or compliance constraints? Owning the network can be the deciding factor regardless of cost.
If the answers point in different directions, that is not a failure of the framework. It is a signal that you have two different deployment environments, and the correct architecture is probably to use both.
Sources and further reading
Primary references for the standards, regulations and figures cited above:
- Delicensing of the 865-867 MHz band — GSR 564(E), Department of Telecommunications, Government of India — The primary regulatory instrument delicensing India's 865–867 MHz band for low-power short-range devices.
- LoRa Alliance — LoRaWAN specification and regional parameters — The standards body defining LoRaWAN, including the IN865 regional parameters and Adaptive Data Rate.
- 3GPP — NB-IoT (Narrowband IoT) — The standards body defining NB-IoT in licensed cellular spectrum.
- National Frequency Allocation Plan (NFAP), WPC Wing, Government of India — The authoritative Indian spectrum allocation reference.