Your IoT Pilot Works. Why Isn’t It in Production Yet?

Most IoT pilots don’t stall because the technology fails. They stall because a few decisions were made too late, or never made at all: what the problem is worth, where coverage really exists, what the deployment costs over ten years, who owns the business case, and how much to spend on hardware.

We see this pattern across the thousands of production deployments running on The Things Stack, which connects more than 6 million LoRaWAN devices worldwide. The deployments that scale make these five decisions early and on purpose.

The pilot worked. Now what?

The sensors send data, the dashboard fills up and the demo goes well. Then the pilot is extended, extended again, and no one decides to roll it out. The industry calls this pilot purgatory, and it has been well documented for years. A 2017 Cisco survey found that 60% of IoT initiatives stalled at proof of concept.

At The Things Conference 2025, Pete Bernard, CEO of the EDGE AI Foundation, described where the market is heading: “What we are seeing now is a transition from pilots to infrastructure.”

That transition doesn’t happen by itself. Here are the five decisions that separate the pilots that stall from the deployments that scale, drawn from what our team sees in the field.

1. Know what the problem is worth before the pilot starts

The deployments that scale fastest come from companies that already serve a specific group of customers. They know exactly what a problem is worth before they install a single sensor, so the pilot only has to answer one question: does the technology work?

Thermostats in hospitals, hotels and care facilities are a good example. Adoption there grew very fast. It is one problem that repeats across many sites, and customers rely on one provider for many locations, so the provider gets the scale effect. Retail and building management services follow the same pattern.

Compare that with automatic refilling of consumables in building maintenance. The idea is clear and the sensors work, but the value per site is small. The business case is very thin, and in many cases it isn’t there yet.

The economics of the process matter too. LoRaWAN usually automates an existing process, and the value of that process decides whether the solution can scale. In oil and gas, equipment is expensive and customers are less price sensitive, so extensive monitoring pays off if it reliably solves an issue. Cold chain monitoring and metering are different: the solution has to be extremely cheap to work as a business case. Know which kind of process you are in before you design the pilot.

On The Things Stack: multi-tenancy lets one service provider run every customer and every site on a single network server, so the hundredth site is as easy to manage as the first.

2. Check your coverage assumptions

Tracking is one of the most common pilots to stall. Tracking with other technologies can be expensive, and LoRaWAN is attractive because it is cheap. Teams then assume there will be coverage wherever their assets go. In practice, LoRaWAN coverage is not available everywhere, so a tracking project often has to build its own network. That cost was never in the pilot plan.

The deployments that scale check where their devices will be, and what coverage exists there, before the pilot starts. If they need their own gateways, that goes into the plan and the business case from day one.

On The Things Stack: you can run your own private LoRaWAN network wherever you need coverage, and connect it to public and partner networks through Packet Broker.

3. Price the whole 5 to 10 year lifetime

System integrators are good at estimating costs over one to two years. A LoRaWAN deployment can run for a decade. Over that horizon the numbers become abstract and hard to estimate, so they often get left out. This is the part of operations pilot teams underestimate most.

The deployments that scale make an estimate for the full lifetime, even a rough one. It covers labour, the availability of field service, battery costs, and the ongoing work of updating and maintaining software.

At The Things Conference 2025, Kai Hackbarth of Bosch made the same point: real progress comes from IoT systems designed to evolve over time and stay operational for years. Buyers look for this too. Fab Controls, one of our customers, applies two checks when choosing a platform: can the costs be forecast, and will the vendor stay committed for five years?

On The Things Stack: gateway fleet management and firmware updates over the air reduce the field visits that drive lifetime cost, and The Things Industries has built one product, the LoRaWAN Network Server, since 2015.

4. Bring the commercial side in early

When a pilot works and still doesn’t scale, the missing piece is usually the business case. The technical team runs the pilot alone, proves the technology, and only then goes to management for budget. By that point the commercial side has no stake in the result.

The deployments that scale bring management or the commercial side in from the start. That creates accountability early and, just as important, a shared upside. The pilot becomes a joint project, and the budget conversation starts on day one instead of after the demo.

On The Things Stack: organizations and role-based access let technical, commercial and operations teams work in the same environment, each with the view they need.

5. Spend on hardware quality to cut recurring costs

Stalled pilots focus on the one-off price per device. They choose the cheapest hardware, then reserve budget for failures and replacements in the years ahead. Nobody can predict that cost, so the budget ends up either too small or too large.

The deployments that scale pay more for higher quality hardware at the start. That cost is known now, with some certainty, and it fits into the business case. Better hardware lowers the uncertain cost of failures and replacements over the life of the deployment.

On The Things Stack: the LoRaWAN Device Repository lists over 1,500 device profiles with decoders ready to use, so you can choose proven hardware and onboard it in minutes.

One thing to do on day one

Don’t worry too much about one-off costs. Be very careful with recurring costs and responsibilities.

On the first day of your pilot, write down every recurring cost the deployment will create over its lifetime, and name who is responsible for each one. Most of the five decisions above follow from that list.

————————

Frequently asked questions

What is pilot purgatory in IoT?
Pilot purgatory is when an IoT pilot works technically but never moves to a full rollout. The pilot gets extended again and again because no one has made the commercial and operational decisions needed to scale it.

Why do most IoT pilots fail to scale?
Usually because the business case is missing, not the technology. Other common causes are overestimating network coverage, underestimating the cost of running a deployment for 5 to 10 years, and leaving the commercial side out until the pilot is over.

How long does a LoRaWAN deployment last?
A LoRaWAN deployment can run for a decade. That is why production deployments estimate labour, field service, battery and software maintenance costs over the full lifetime, not just the pilot period.

How do you move an IoT pilot to production?
Decide early what the problem is worth, check coverage where your devices will be, price the full lifetime, involve the commercial side from day one, and invest in hardware quality to reduce recurring costs.

Ready to move your pilot to production?

If your pilot works but isn’t moving forward, talk to our team. We’ll help you work through these five decisions and plan the move from pilot to production on The Things Stack.