A wheelchair accessibility rule for the city’s for-hire fleet, with a deadline and nothing yet to run it. Left behind the program that made it work, and an industry that adopted it.
- Walked into
- A finalized rule, a deadline, and no program behind it: no way to run it, measure it, or bring the industry along.
- Built
- The program’s design, from vehicle inspection to the trip-level logic that decided compliance, and the change management that got the industry to adopt it.
- Kept running
- Accessibility built into how the for-hire industry operates, and a program the agency can run and enforce.
For years, the app-based companies that had come to dominate New York’s for-hire trips only carried passengers who could transfer into a car and fold their wheelchair, which shut out everyone who couldn’t. The TLC set out to change that with a rule requiring the whole for-hire industry, from neighborhood car services to Uber and Lyft, to provide equivalent wheelchair accessible service.
The rule on paper
By the time it reached me, the policy was decided. What didn’t exist was the program: how vehicles would qualify, how compliance would be decided, and how companies would learn what was expected of them. There was a rule, a deadline, and nothing yet to run it.
This was a data-driven program. Whether a company was complying wasn’t something anyone could see on the street. It was a calculation. So designing that logic wasn’t support work for the program. It was the program, and building it, with the industry and every part of the agency it touched, was my job.
Every system it touched
A rule like this looks like one requirement. In practice it touches everything. Tracking compliance meant following vehicles and trips through systems that had never been built to know anything about accessibility. It also meant working across nearly every part of the agency, including legal, enforcement, licensing, policy, and IT, plus the industry groups who would have to change how they operated.
I mapped how the change would move through each system and wrote detailed product specs for what needed to be built.
Just as important was bringing the industry along. A rule like this changes which vehicles companies put on the road, how they dispatch, and what they report. I drafted communications for each type of licensee, since a small car service and a national app company needed to hear very different things, and worked with industry groups so that adoption was part of the design rather than something enforced after the fact.
Deciding when no one would
The rule said what had to happen. It didn’t answer many of the questions that determined whether a given company was compliant, and there wasn’t much appetite inside the agency to make those calls.
Which vehicles could be converted to be accessible was one example. There were no hard rules. The taxi industry had established standards for accessible vehicles, so I based the requirements on those, put the decision forward, and watched the response. Where no objections came, the decision held.
On the inspection floor
Every accessible vehicle had to pass through the agency’s inspection facility, and that inspection decided a lot downstream. Whether a vehicle counted as accessible determined its eligibility and how its trips were monitored for compliance. If the inspection data was wrong, everything built on it would be wrong too.
So I worked with the outside vendor who managed the inspection software, then went to the facility to confirm the whole thing worked end to end. I stood with the inspectors as they confirmed each vehicle met the standard, and watched those results flow into the systems that would track eligibility and compliance.
Following every trip
Qualifying vehicles was only the first half. The program itself lived in the trip data. Whether a company was complying could only be known by merging vehicle records with trip records and applying the rule’s logic to every trip. Without that logic there was no program, just a rule.
The rule gave companies more than one way to comply, and each needed its own logic. For bases held to a share of trips, we had to confirm that the required proportion was actually being dispatched in accessible vehicles. For bases on the exception path, the question was whether passengers who needed an accessible ride could actually get one. And the exceptions let bases farm out accessible trips to other bases, so a ride requested through one company could be fulfilled by another. We had to work out how to follow those trips across companies so each one counted toward the right base’s obligation.
Getting this right mattered in both directions. The rule existed to open the fleet to New Yorkers who’d been shut out of it, so the tracking had to hold companies accountable. But bad data would also mean summonses for businesses that were actually complying, in an industry already under strain, and extra work for the agency’s own staff.
It worked. In the first six months, the number of accessible for-hire vehicles on the road grew more than tenfold, and monthly trips in them rose from under 7,000 to nearly 140,000.
What travels
A policy isn’t finished when it’s adopted. That’s when the program has to be designed: how it will actually run, how anyone will know it’s working, and how the people it governs will come to adopt it. Much of that work is making decisions the policy left open. When nobody wants to own a call, waiting for consensus stalls the work.
It’s often better to make a reasonable decision, anchored in whatever precedent exists, and put it where the people affected will see it. A concrete proposal draws out objections that an open question never will.