Editorial Article
A 43.1 Per Cent Cut in Search Time That Appeared on No Invoice: Who Pays for a Sensor That Only Helps the Driver
San Francisco instrumented a kerb and then paid somebody to find out whether it worked. That second decision is rarer than the first, and it is the reason SFpark is still the only evaluation in this category worth arguing with. Reported search time fell from 11 minutes 36 seconds to 6 minutes 36 seconds — 43.1 per cent — against 12.5 per cent in areas left alone. Cruising greenhouse-gas emissions fell 30.0 per cent, from 7.0 to 4.9 metric tons a day, against 7.41 per cent in the control. Time spent at target occupancy rose 31 per cent against 6. The share of blocks sitting too full fell 16 per cent while the control areas rose 51.
Four clean results, measured by the city rather than by the supplier, from SFMTA’s June 2014 pilot evaluation. And not one of them appears on the income statement of the operator who bought the hardware.
That is the quiet problem inside the parking instrumentation market, and it explains a great deal that unit economics cannot. The global category reaches USD 6,020 million by 2033 from USD 2,871.7 million in 2025, at 9.69 per cent — USD 3,150 million of it in 2026 — across 12.27 million monitored bays rising to 24.8 million at 9.19 per cent. A category that produces that much measurable public benefit and grows at nine per cent is not a category with a demand problem. It is a category with a collection problem.
Four results, and the four people who receive them
Read the SFpark numbers as a list of beneficiaries rather than a list of outcomes, and the market’s structure falls out of them.
- Five minutes of the driver’s time. 11:36 to 6:36 per search, against a 12.5 per cent improvement in the control. The beneficiary is the person circling. The operator’s revenue is a function of the tariff and the dwell, not of the driver’s search time, and a faster search is as likely to shorten a stay as to lengthen one.
- 2.1 metric tons a day of avoided cruising emissions, 7.0 down to 4.9, which is 766.5 metric tons a year if the pilot rate held. The beneficiary is the city and everyone breathing in it. The operator is not paid for the tonne avoided.
- Occupancy at target, up 31 per cent against 6. The beneficiary is the pricing policy, which is to say the municipality that set the target. This is the one result an operator can monetise, and only if the operator is the one who set the tariff.
- Blocks too full, down 16 per cent against a 51 per cent rise in the control. The beneficiary is the kerb as a whole, including the loading bay, the bus stop and the disabled space nobody was watching.
An operator buys a system that produces all four of those and books the fourth one only partially, and the third only if the contract lets it set the price. Everything else leaves the building. That is not a failure of salesmanship. It is a structural gap between where the value lands and where the invoice can be sent.
What a monitored bay is worth to the person circling
Put a number on the driver-side benefit and the gap stops being rhetorical. Five minutes saved per search, at a round USD 15 an hour, is USD 1.25 of driver time. At a deliberately conservative three searches per bay per day, that is USD 1,368.75 per bay per year. Both inputs are assumptions — ours, stated as such — but the direction survives any sane variation in them.
Against that, the entire global instrumentation industry collects USD 242.74 per monitored bay per year in 2033, and USD 234.04 in 2025. That is the whole stack divided by the installed base: sensor, camera, counter, sign, software, installation and service contract. On those assumptions, the benefit a monitored bay delivers to drivers is 5.64 times the revenue the industry collects for producing it.
A market where the buyer captures less than a fifth of the value created does not allocate itself efficiently, and it does not grow at the speed of its own usefulness. It grows at the speed of whatever mechanism happens to convert the public benefit into a private line — a contract, a mandate, a booking funnel or an enforcement power. Where one of those exists, instrumentation is bought without argument. Where none exists, the same hardware is a capital item competing against a resurfacing programme, and it loses.
Why the operator cannot simply charge for it
The obvious response — charge more for a better search — does not survive contact with a tariff schedule. Parking is priced per hour at the bay, and the driver’s search happens before the bay is occupied. There is no moment at which a parking operator can meter five minutes of somebody’s life, because the transaction it can bill has not started yet. The one lever available is the one SFpark pulled: set the price so the block sits at target occupancy rather than at saturation, and collect the difference on turnover. That requires the operator to hold pricing power, which in leased municipal assets it usually does and in fee-managed assets it usually does not.
There is a second constraint, and it is organisational rather than economic. Occupancy data pays back only if somebody acts on it: repricing a permit, reallocating a section, releasing capacity to a pre-book channel, dispatching an officer to a named bay, routing a driver at the ramp. In a facility where parking is run as a property overhead — a cost line attached to an asset, staffed by a manager who also runs security — that person does not exist. The system is installed because the design standard specified it, and it produces a dashboard nobody opens. Those sites report the lowest satisfaction of any buyer group in our interview programme, and they do not renew.
Where the benefit does get invoiced
The contexts where the conversion already works are narrow, and they are all defined by an obligation rather than by an optimisation.
Accessible bays. Accessible parking spaces carry statutory design, location and access-aisle requirements under the Americans with Disabilities Act in the United States and the Equality Act 2010 in Britain. Where the obligation is to keep a specific space available, monitoring is not a business case — it is compliance, and compliance is the least cyclical demand line in the model.
EV bays. A charging bay has to be defended against unauthorised dwell or the charger’s utilisation collapses. Regulatory pressure is turning the recharging bay from an amenity into an asset with a reporting duty attached, under Directive (EU) 2024/1275 and Regulation (EU) 2023/1804 in the European Union, which is the mechanism our EV charging at parking assets report treats at length.
Airport pre-book. A pre-booked space is sold before it is occupied, so inventory certainty is the product. Airports carry 26 per cent of this market’s 2026 value on a small fraction of global floor area, and it is not because airports are keener on technology. It is because they sell the bay in advance and lose money every time it is given away. The airport parking and pre-booked reservation platforms report sets out the same argument from the booking side.
Enforcement. A penalty has to rest on a defensible record of where a vehicle was and for how long. That record is worth more than the sensor that produced it, which is why the recognition half of this market grows at 12.25 per cent against 7.67 per cent for bay-level detection and takes 44.25 per cent of the whole USD 2,870 million increment from a base of under a third.
The complication worth pricing in
One development would change the size of this market rather than its composition. Occupancy data is becoming a by-product of vehicles rather than of buildings. Connected cars already report where they parked; fleet telematics knows to the minute when a vehicle arrived and left. If that data becomes accurate, timely and licensable, the bay sensor’s remit narrows to the cases where a vehicle cannot be trusted to report on itself — enforcement, accessible bays, EV blocking, permits and pre-book.
That would not delete instrumentation. It would cap it, and move the value towards whoever holds the vehicle’s identity and the right to its data. Any operator signing a seven-year bay-sensing contract today is making an assumption about the floor plate being the only place occupancy can be observed. The evaluation suggests the sensor earns its keep wherever an obligation exists. It does not establish that the obligation will always be discharged from above the bay.
The company set behind the model
The market splits into two competitive sets that meet the same customers weekly and rarely bid for the same order. At the barrier and in revenue control: Metropolis, SP+, SKIDATA and Q-Park. In the bay: Amano, Nedap, Genetec, Robert Bosch and Continental, with Indigo operating across both. Ten firms, and the report states the count rather than implying it.
Capital has already chosen a side. Metropolis’s acquisition of SP+ at an enterprise value of approximately USD 1.5 billion — USD 54.00 a share in cash, on the company’s own closing release of 16 May 2024 — bought operated inventory and the transactions running through it, not a sensor estate. That is the correct purchase if the marginal economics belong to the transaction, and this article is an argument that they do.
What the reader should take away
A sensor that produces a measurable public good and no billable line will be bought late, bought thinly, and bought only where a rule forces the decision. The evidence for the good is not in dispute — a 43.1 per cent cut in search time and a 30 per cent cut in cruising emissions came from a city that went and measured them. What is in dispute is who pays. Until an operator can invoice the five minutes, the million-bay forecast is a forecast about obligations, not about desire.
The full model, the four revenue lines and the per-bay series are in the Global Parking Occupancy Sensing & Space-Guidance Market Report 2026–2033. The market-structure argument in this piece is set out at greater length in The Sensor That Isn’t Installed: Where the Money in Parking Sensing Actually Goes.
Related research
- Parking Access & Revenue Control Systems — the lane as the unit of sale, and why the gate itself stopped being the product.
- ALPR & Vehicle Identity for Access — the recognition half, and the statutory price of a misread plate.
- On-Street Parking Enforcement & Digital Permitting — where the record of occupancy becomes an invoice by design.
- Parking Concession Monetisation — what the revenue a sensor helps produce is worth once it is discounted.
- Parking Management & Concession Operations — why the same bay earns 32.0 per cent on a lease and 8.8 per cent on a fee.
Sources and method
The SFpark results are from SFMTA’s SFpark Pilot Project Evaluation, published June 2014, and are quoted as reported: reported search time 11 minutes 36 seconds to 6 minutes 36 seconds (−43.1 per cent) against −12.5 per cent in control areas; cruising greenhouse-gas emissions 7.0 to 4.9 metric tons a day (−30.0 per cent) against −7.41 per cent in control areas; time at target occupancy +31 per cent against +6 per cent; blocks too full −16 per cent against a 51 per cent increase in control areas.
Market values, revenue-line splits, installed-bay counts and the per-bay series are Pheonix Research model estimates, base 2025 and forecast to 2033, stated in USD. The per-bay figures are the total market divided by the monitored-bay count. The driver-time valuation in this article — USD 15 an hour and three searches per bay per day — is a Pheonix assumption, labelled as one, and is not a measured input from the evaluation. Currency, unit and methodology conventions follow the report.
Regulatory references are to public instruments: the Americans with Disabilities Act accessible-parking requirements as set out by the US Department of Justice; the Equality Act 2010; Directive (EU) 2024/1275; Regulation (EU) 2023/1804; the General Data Protection Regulation. Company identification uses Wikidata entities: Metropolis Q131191821, SP+ Q7598289, SKIDATA Q1168035, Q-Park Q1127798, Amano Q11284890, Nedap Q1479312, Genetec Q30267239, Robert Bosch Q234021, Continental Q163241, Indigo Q112822073. The SP+ transaction detail is from Metropolis’s closing release of 16 May 2024.
The analysis is built 70 per cent on the secondary record — regulation and its annexes, standards, municipal and airport procurement documents, operator disclosures and patent filings — and 30 per cent on primary research conducted by Pheonix: 31 semi-structured interviews across seven markets, nine conference-floor interviews, one closed roundtable with five operators and one toll-authority finance director, and 47 written survey responses, 2024–2026. No competitor research is cited.
Frequently Asked Questions
How much time did SFpark actually save drivers?
Reported search time fell from 11 minutes 36 seconds to 6 minutes 36 seconds, a 43.1 per cent reduction, against a 12.5 per cent improvement in control areas. The same evaluation found cruising greenhouse-gas emissions down 30.0 per cent from 7.0 to 4.9 metric tons a day, time at target occupancy up 31 per cent against 6 per cent, and the share of blocks sitting too full down 16 per cent against a 51 per cent increase in the controls.
Why doesn't the parking operator capture that benefit?
Because the driver's search happens before the billable transaction starts. Parking is priced per hour at the bay, so there is no moment at which an operator can charge for the five minutes saved. The only lever available is setting the tariff so a block sits at target occupancy rather than saturation and collecting the difference on turnover, which requires pricing power many fee-managed operators do not hold.
What is a monitored bay worth compared with what it collects?
On Pheonix assumptions of USD 15 an hour for driver time and three searches per bay per day, the driver-side benefit is USD 1,368.75 a year per bay. The entire instrumentation industry collects USD 242.74 per monitored bay in 2033 and USD 234.04 in 2025. That puts the benefit at about 5.64 times the revenue, which is why the technology is bought where an obligation exists rather than where the benefit does.
Where does bay-level monitoring get bought regardless?
Wherever an obligation rather than a calculation decides the purchase: accessible bays under the ADA and the Equality Act 2010, EV bays that must be protected against unauthorised dwell, airport pre-book inventory where the space is sold before it is occupied, and enforcement regimes where a penalty has to rest on a defensible record of where a vehicle was and for how long.
