Project Smart City: Nine IoT Programs. Six Died Before the Field. [2026]
The Pipe · Project Smart City 03
Project Smart City: Nine Programs, Six Never Reached the Field
Networks & Connectivity edition. An inventory of every connected-device program I ran, and what actually killed them.
Written by Scot Free
Between roughly 2017 and 2019 I ran nine connected-device programs inside a large satellite television and wireless carrier's narrowband IoT initiative, reporting to the founder. I brought several of the partners in myself through cold outreach. All nine are dead. Most were parted out to salvage companies years ago.
That is not an unusual outcome. IoT Analytics counted more than 620 IoT platforms at the peak of that cycle, and the overwhelming majority of them are gone. The manifesto that opened this site is partly about why.
What is unusual is having the full list, with the cause of death for each, from the person who ran them. Most post-mortems of that era are written by analysts who were never in the room. This one is an inventory.
The headline finding is the part I did not expect when I wrote it out:
Three programs died on physics. Six died before a device was ever deployed in the field.
The inventory
Program names are genericized. The failures are described as they happened; no partner is named as the source of one, because in most cases the partner was not the reason.
| # | Program | The claim | The pipe | Where it broke | Class |
|---|---|---|---|---|---|
| 1 | Commercial weather station | Professional-grade site weather, reported over cellular IoT instead of a local gateway | NB-IoT | Firmware. The device could not be made to report reliably in the field. | Physics |
| 2 | Asset locator | Track high-value movable assets on a multi-year battery | NB-IoT | Battery life. The duty cycle the use case required and the duty cycle the radio allowed were different numbers. | Physics |
| 3 | Residential weather station | Consumer version of #1, sold through an existing subscriber channel | NB-IoT | Died at the cost and timing proposal. The revised numbers did not clear. | Commercial |
| 4 | Developer kit | Seed the ecosystem; let third parties build on the network | NB-IoT | Cut after three months. Cutting-room floor during a scope pass. | Scope |
| 5 | Air-quality monitor | Municipal air-quality sensing at neighborhood resolution | NB-IoT | The City of Denver delivered a master service agreement. It died in legal review. | Governance |
| 6 | Multi-sensor node | One device, several environmental channels, long-range unlicensed radio | LoRa | Stability. It would not stay up. | Physics |
| 7 | Soil-moisture sensor | Irrigation telemetry for agriculture | LoRa | Failure to launch. Never got far enough to fail technically. | Scope |
| 8 | Low-power broadcast | Linear television over the new broadcast standard | ATSC 3.0 | The product requirements document stalled. It never came out the other side. | Governance |
| 9 | High-power broadcast | Chapter two of #8, at scale | ATSC 3.0 | Same story, same gate. | Governance |
Author's firsthand record. Program descriptions are generalized to the device class.
What the data shows
Sort the nine by class and the shape is clear:
- Physics: three. Firmware that would not report, a battery budget that could not meet the duty cycle, a node that would not stay up. Real engineering failures, in the field, with hardware in hand.
- Governance: three. A signed municipal agreement that died in legal review, and two broadcast programs that stalled inside the requirements process.
- Scope and commercial: three. A consumer variant that didn't clear on cost and timing, a developer kit cut during a scope pass, and an agriculture sensor that never got moving.
Six of nine never reached a deployment. They were killed by a legal department, a requirements document, a budget gate, and a reprioritization — none of which has anything to do with radio propagation, battery chemistry, or firmware.
That matters because of how this era gets narrated. The standard story is that the first IoT wave failed because the technology wasn't ready: the silicon was too power-hungry, the networks too thin, the data too dirty. All of that was true, and I've argued it myself.
But in my own inventory, the technology was the minority cause of death. Two thirds of these programs were killed by the organization around them before the physics ever got a vote.
That is the same finding as Toronto Quayside, at a hundredth of the scale. A $1.3 billion project and a nine-program portfolio inside one carrier broke the same way, for the same reasons, in the same years.
The tenth row
There is one more entry, and it belongs in a different column because it was a different kind of thing.
During the same period I was developing partnerships to place network gateways in partners' existing street footprint. One of them was Bigbelly, the smart solar trash compactors you have walked past in a hundred downtowns, which I brought in through cold outreach.
That one worked. Not because the technology was better — it was the same radio, the same era, the same company. It worked because of what it was:
- The footprint already existed. The units were already installed, already powered, already permitted, already maintained by somebody with a route.
- Nobody had to approve a new object in the public right of way. The governance gate that killed program #5 was never opened, because nothing new was being placed.
- It was infrastructure, not a device program. There was no product requirements document to stall in, no consumer price point to clear, no use case that had to be invented after the hardware.
Nine device programs looking for a use case: zero survivors. One infrastructure placement on an existing footprint: the arithmetic worked.
Where it breaks, generally
If you are standing up connected infrastructure now, the inventory suggests three gates that deserve more attention than they usually get, and they are all upstream of engineering.
- Find the legal gate before you find the customer. Program #5 had a municipality that wanted the thing and had executed an agreement to get it. It still died. If your deployment requires a contract that has to survive two legal departments, that review is your critical path, not your pilot. Run it first, in parallel with the technical work, not after the demo succeeds.
- A requirements document that will not close is an answer. Programs #8 and #9 did not get rejected. They stalled, which is slower and more expensive than a rejection and feels like progress the entire time. If a PRD has not closed after a reasonable interval, the organization has already decided something; it just hasn't said so. Treat the stall as the decision and reallocate.
- Prefer existing footprint to new objects. Every new thing you put in the physical world draws a permit, an owner, a maintenance route, and a liability question. Everything you attach to something already out there inherits all four. That is the entire difference between the nine and the tenth.
The Underground Take
Writing the list out was the uncomfortable part. You remember these as a run of bad luck and a technology that wasn't ready, and then you sort them into columns and two thirds never got to the part where technology could fail.
The honest reading of this inventory is not that the organization was bad at its job. Legal reviews exist for reasons. Scope passes exist for reasons. A consumer price point that doesn't clear is real information, delivered early and cheaply. Every one of those gates was doing what it was built to do.
The failure was upstream of all of them: nine device programs went looking for a use case after the hardware existed. The one that survived started from infrastructure that was already in the ground, already permitted, already someone's responsibility. The pipe was never the problem. The thing we kept trying to attach to it was.
Next in Project Smart City: [TBD].
Sources
- Program inventory, causes of death, and the Bigbelly partnership: the author's firsthand record as the program manager, 2017–2019. Program descriptions generalized to device class; no partner is identified as the cause of a failure; no commercial terms are disclosed.
- IoT Analytics, IoT platform market research — the 620+ platform count at the peak of the cycle, as cited in The Intelligent Edge.
- Bigbelly: smart solar compaction units, public street deployments. Named with the partner's awareness; the relationship is described only as far as the public footprint and the author's own outreach.
A note on this piece: it is a single practitioner's inventory of nine programs inside one organization, not a dataset. It is offered as a counterweight to post-mortems written by people who were not in the room, and it should be read as one sample. If your own inventory sorted differently, that is worth knowing and the inbox is open.
Corrections and additions: Scott@IoTunderground.com.
