Project Smart City: The Claim, the Pipe, and the Number Nobody Checked [2026]

Project Smart City  ·  Series Opener

Project Smart City: The Claim, the Pipe, and the Number Nobody Checked

A recurring audit of what cities are sold against what the sensors actually deliver.

Written by Scot Free

Every city has a smart-city page now. Every vendor has a smart-city slide. The word has been sold so hard for fifteen years that it means almost nothing, and that is exactly why this series keeps it. "Smart city" is the label on the invoice. What we do here is open the invoice.

The Underground's editorial position is that the model gets the credit and the pipe does the work. Nowhere is that more true, or more expensive, than when the customer is a municipality. A city buys a sensor network the way it buys a bridge: once, on a bond, with a press conference. If the arithmetic is wrong, nobody finds out for six years. This series is for the engineer who will be asked to make that arithmetic right, and for the reader who wants to know what "smart" cost.

The method is a reconciliation

If you have ever matched a vendor's invoice against a usage log, you already know how every post in this series works. There are five parts, and they map one-to-one onto an audit:

Section	The audit equivalent	The question
The claim	The invoice	What did the vendor or the city say this does?
The pipe	The contract	What sensor, what radio, what data layer, what enclosure?
What the data shows	The usage log	What do the published results, open data, audits, and measurement papers say?
Where it breaks	The exception report	Where does the claim and the record diverge, and by how much?
The Underground Take	The finding	What should the next city do differently?

Two rules of evidence carry through every post. The first is Sagan's: extraordinary claims require extraordinary evidence, and "we cut congestion 30%" is an extraordinary claim. The second is Bergstrom and West's, from Calling Bullshit: you do not need to be an expert in the domain to check how a number was produced. You need to know where the data came from, who got counted, what got left out, and whether the metric quietly became the target. That is the pipe. Anyone who has built one can audit one.

One more rule, and it is the one that separates a skeptic from a cynic: sometimes the finding is that the claim holds. When it does, this series will say so, with the same receipts.

A demonstration, from the buyer's chair

Rather than promise what the series will uncover, here is the method run once, on a claim I could drive to.

THE CLAIM

In November 2016 a Denver startup called ParkiFi launched a consumer app on top of a network of parking sensors. The public record from that year and the next reads like every smart-city pitch you have ever seen, and I am quoting it because it is on the record, not because it is unusual:

• Sensor accuracy that "exceeds 99 percent."
• A hockey-puck sensor that cost about $20 to make against an industry standard of about $250, a claimed 90% hardware cost cut achieved by using a single magnetometer instead of a stack of sensor types.
• Roughly $13.5M in venture funding by late 2016, later reported at $13.9M.
• Sensors in lots and garages across eight, then nine, then eleven states; a partnership with the City of Denver on off-street lots under the Smart City Challenge banner; a stated goal of 10,000 more downtown spaces within a year.

Product, price, scale, city partnership. That is the full smart-city invoice, and the coverage repeated every line of it.

THE PIPE

A magnetometer in a sealed puck, epoxied or cored into the pavement of a single space, with a battery, a short-range radio to a gateway, and the gateway's backhaul to a cloud analytics layer. Occupancy detection this way is sound engineering: a car is a large ferrous object, and the change in the local magnetic field is a clean signal. Nothing in the claim about the sensor was implausible.

WHAT THE DATA SHOWS

Run the arithmetic the coverage never ran. These are illustrative Fermi numbers, not ParkiFi's books, and you can swap in your own city's:

• A downtown surface-lot space earns on the order of a few thousand dollars a year. Call it $3,000.
• The revenue a sensor can add is the occupancy it fills that would otherwise sit empty, plus the enforcement it tightens. A generous read is a few percent. Call it 3%, or about $90 a space a year.
• Installed cost per space is not $20. It is the puck plus the core drill or the surface mount, plus a share of the gateway, plus the platform subscription, plus a battery service visit. Call it $150 installed and $30 a year to keep alive.

On those numbers the sensor pays back in about two and a half years. That is not a disaster. That is a fundable business, which is why it got funded. The arithmetic on paper closes.

Then the paper meets Denver. Now put one more line in the model, the one that never appears on a slide: replacement rate. A puck in a Denver parking lot lives through freeze-thaw, plows, road salt, and years of meltwater looking for a way in. If the seal fails at eighteen months instead of ten years, the per-space cost is not $150 once. It is $150 every eighteen months, plus a truck roll each time, and the payback period runs past the life of the company.

WHERE IT BREAKS

The pucks leaked. I did not read that anywhere. I was in the room, along with a good number of my colleagues, when the units were opened, and the enclosure did not hold. The $20-versus-$250 line that made every article was the claim. The seal was the pipe. The 90% cost cut had been taken, in part, out of the one component that decides whether an outdoor sensor is a ten-year asset or an eighteen-month consumable.

For the record, the rest of the story is also public. DISH Network acquired ParkiFi in February 2018 on undisclosed terms, saying it would use the company's "practical deployment and real-world experience connecting low-powered sensors with gateways to the cloud" to test and learn as it built a national narrowband IoT network, the first phase of which was due in March 2020. The acquisition release even named the next applications: utility monitoring and street lights. The narrowband plan was later overtaken by a different wireless strategy, and the parking product did not outlive the pivot. I will state the outcome the way I witnessed it: ParkiFi died in our corporate parking garage.

Here is the part that belongs to this series. In the same period, a two-person shop, a father and son out of St. Louis, showed us a connected streetlight with a better enclosure than the venture-funded puck. Their product had its own problem, which was a payback period measured in years for a feature set that included changing the light's color for holidays. Great engineering, wrong arithmetic. The puck was the reverse: a fundable model wrapped around an enclosure that could not survive the deployment. Two smart-city products, two failures, neither of them in the part the press release was about.

What the series will cover

Each post takes one subject and one claim. The subjects rotate:

• Transportation. Congestion, signal timing, parking, transit telemetry. The claims are the largest and the open data is the best, which makes it the easiest subject to audit and the hardest to defend.
• Energy. Streetlights, smart meters, building telemetry. Most of the savings claimed for "smart" lighting were captured by the LED swap that preceded it. We will separate the two.
• Infrastructure. The underground: water, bridges, tunnels, the NB-IoT physics from Subterranean Telemetry applied to what the sensors were supposed to watch.
• Networks and connectivity. LoRaWAN, NB-IoT, private 5G, and municipal fiber. Which radio the city bought, and whether it was the one the sensors needed.
• Open data. The one subject where the city's own portal is the usage log. When a live feed exists, we will pull it into a pipeline, the way The Lakehouse at the Edge did, and let the expectations tell the story.
• Project in Focus. One named project per post, from anywhere in the world, run through all five sections. The first one will be the most-funded smart-city project in North America and the shortest-lived.

The Underground Take

I spent eleven years in audit before I spent twenty in a finance organization, and the thing both jobs teach you is that the claim is never where the money went. The money went into the pipe, and the pipe is where it leaks — sometimes literally.

A smart city is a reconciliation problem. On one side, the invoice: accuracy, cost, scale, savings. On the other, the record: what the sensors reported, how many were replaced, what the city actually did with the data. Every post in this series lines those two columns up and reports the variance. Most of the time there will be one. Some of the time there won't, and that finding gets published with the same care.

The puck taught me the rule this series runs on. The vendor will show you the model. Ask to see the enclosure. Then ask what happens to the payback when the enclosure fails in a Denver February. If they have a number, you are talking to an engineer. If they have a slide, you are talking to a press release.

Sources

• Denverite, "ParkiFi brings 12 days of free parking to Denver," Nov. 23, 2016 ($13.5M raised; $20 vs. $250 sensor cost; single-magnetometer design; sensors in eight states). Denverite, "ParkiFi, an app to ease parking," Nov. 14, 2016 ("exceeds 99 percent"; City of Denver partnership).
• Planetizen / The Denver Post, "App Detects Open Parking Spaces with 99% Accuracy," Dec. 5, 2016.
• BusinessDen, Aug. 2017 (Series A extension; $13.4M+ raised; 10,000 downtown spaces target; nine states). Confluence Denver, May 18, 2017 (24 locations, 11 states, 13 markets).
• Next TV, "Dish Buys Parkifi," and Built In Colorado, Feb. 1, 2018 (acquisition; undisclosed terms; $13.9M raised; Tom Cullen statement on street lights and utility monitoring).
• DISH Network Form 10-K, FY2018 (March 2017 FCC notice of a 5G-capable network focused on narrowband IoT; first phase due March 2020).
• Bergstrom, C. and West, J., Calling Bullshit: The Art of Skepticism in a Data-Driven World, 2020. Sagan, C., The Demon-Haunted World, 1995.
• The Fermi estimate in "What the Data Shows" uses illustrative figures stated in the text and is not a representation of any company's financials. The enclosure observation is the author's firsthand account.

Corrections and additions: Scott@IoTunderground.com. The community is invited to extend or correct any finding in this series.