the US to sign a service provision agreement with a transportaon network company (TNC) to oer

Direct Connect was originally conceived in 2015 as a replacement for two under-performing, low-

frequency feeder bus routes. Specically, riders were given a $3 subsidy for rides to or from bus

stops in two zones via Uber, United Taxi or the WAV provider, Care Ride. While Direct Connect

ridership was minimal during the inial six months of the pilot, low operaonal costs helped the

agency to jusfy connuing and expanding the service on a provisional basis.

With the goal of increasing ridership, PSTA expanded the Direct Connect service area to eight

zones across the county in 2017. Leading up to the expansion, Uber made usability improvements

to the in-app experience while PSTA switched wheelchair service providers, increased the per-

trip subsidy to $5, and added to the pilot’s overall budget. A greater eort towards markeng and

outreach by PSTA, Uber, and United Taxi also led to several months of consistent ridership growth,

While the pilot achieved the agency's cost-cung goals, both overall and on a per-rider basis, there

were clear shortcomings. Direct Connect's zone-based service design limited the transit routes

users were funconally excluded enrely due to the pilot’s xed-subsidy (as opposed to xed-

fare) pricing. PSTA’s ability to evaluate Direct Connect’s ecacy in providing a desirable service

alternave to those riders has been limited by a lack of agency rider surveys, eld observaons,

in their service zone, while the rider subsidy for

Connect fares comparable for WAV and non-WAV trips.

Connect as a xture of its transit operaons for the near

Important data gaps remain heading forward. In early

2019, PSTA agged an issue with Uber’s app, eventually

learning that an overly large geofence had resulted in a

signicant overstatement of the number of rides made

evaluaon and auding of its largest provider.

While PSTA is currently unable to understand how Direct Connect riders interact with scheduled

service (if they do at all), soluons seem aainable in the near future. Since October 2018, Uber

has oered a data dashboard for its late-night pilot, which allows PSTA to visualize trip origins and

desnaons. Addionally, PSTA recently helped launch an account-based fare app and entered into

a partnership with the mulmodal trip planner Transit App, both developments that oer potenal

paths to track transfers between xed-route and on-demand service.

PSTA’s overall experience developing, managing, and adapng the Direct Connect pilot provides

insight into what transit agencies can expect when working with on-demand service providers.

While operang on a larger scale, in a denser environment, or with a dierent ridership base

may have oered dierent lessons in implementaon, the Direct Connect pilot’s service design

shows what is necessary for a successful launch of a pilot program: good data and transparency

from all pares, as well as concrete plans for outreach and evaluaon. Though the program faced

challenges, PSTA is to be commended for taking the chance on a new service format and for

adjusng as they learned more about how it was working for riders and for the agency itself.

so soon aer a major funding setback for PSTA speaks to the organizaonal resilience and

dedicaon of key leadership and sta in seeking new ways to provide service.

the diversity of TNC, taxi and wheelchair-accessible service.

service providers and be proacve in lling informaon gaps that exist. Basic aspects of pilot

ulizaon, parcularly around equity implicaons of this service model, remain unknown aer

several years due to a lack of survey and TNC data.

aributable to resource constraints, seem to have stemmed from a lack of service quality goals

or subsequent plans for assessment. Pilot iteraons and expansion eorts likely would have been

beer informed had these been arculated.

PSTA is governed by a board of elected ocials from the region and relies heavily on property taxes

seasonal tourists, the majority of the agency’s riders are transit-dependent residents in the region.

Like many other transit agencies around the country, the 2008 recession led to an increase in

ridership. Forecasng higher operaonal expenses in the face of a housing slump, PSTA made a

push for a more stable source of funding that could support expanded service.

In 2011, PSTA and the Metropolitan Planning Organizaon (MPO) Forward Pinellas put forward a

plan for the rst light rail line in the Tampa Bay area. The rail would be complemented by addional

bus service that, together, would form the backbone of a high-frequency transit network, and a

shi away from PSTA's exisng hub-and-spoke model. These combined proposals became known

Pilot parcipaon was low for a variety of reasons. Rider surveys conducted during a system

redesign study earlier in 2015 indicated that many circulator riders believed they would be able to

likely never learned about the pilot in the rst place or failed to use it correctly, due either to limited

inial outreach on the part of PSTA and Uber or to the program’s complexity—a new service model

requiring coupon codes and conngent on narrow geographic requirements. Other lower-income

Transit magazine. PSTA sta endorsed the pilot’s

area, increasing the overall budget, and focusing more

the PSTA board voted to formally extend and expand

PSTA also expanded the approved pick-up and drop-o zones at the center of each designated

intersecon. PSTA and Uber observed that the 400-foot restricon led some passengers to be

to 800 feet would improve drop-o safety while remaining small enough to minimize the likelihood

of using the program for other purposes besides connecng to transit. Trips coordinated through

the taxi service—without the use of an app or geo-fence—connued to operate without a strict

county, while the four original stops in East Lake and Pinellas Park were consolidated into two

stops and 20 dierent routes.

According to Ly’s representaves, the company was uncomfortable with the possibility that

the pilot might be abruptly disconnued if it were to exhaust its budget, which would result in

technically implement the Direct

Connect program’s geo-fence

structure in me for the launch.

in taxis and buses, and physical

engaged Uber and United Taxi to

PSTA and Uber also sent sta to

Ancipang increased usage, PSTA

replaced the promoonal day

presentaon of a recent Uber or

United Taxi receipt to bus drivers.

three rides per day to nearly ten, connuing to grow over the summer and eventually seling at

Transport during this me, despite the company reporng customer inquiries. Uber rides made up

most of the usage, though United Taxi grew steadily over Phase II and maintained comparable fares

from at least August 2017 through March 2018.

be the preferred service for riders who did not already use app-based, transportaon services.

Phase II aracted roughly 40 rst-me users per month through June, with growth slowing slightly

for the remainder of the summer and picking up again in the fall. Between February 2017 and

March 2018, at least 468 individuals took nearly 12,750 rides — or slightly less than PSTA's average

daily systemwide ridership (Figure 4).

While Phase II drew riders from around the county, most rides originated from several clusters. Of

the top three performing zones, Zone 5 in Pinellas Park and Zone 1 in East Lake corresponded to

county in terms of median age, household size and transit dependency. They were dierent in terms

While there was lile dierence in zero-vehicle households, transit commuters represented 3% of

60% of them took fewer than ve rides during that me. By comparison, the top ten most frequent

users had each taken at least 40 rides, as many as 75 mes for two riders, and alone represented

40% of all rides. Trips through United Taxi followed a similar trend through the last point of analysis

in September 2017, with 55 of the 78 disnct users in Phase II taking fewer than ve rides.

Of the

468 individuals who had used Direct Connect through Uber by March 2018, 277 had taken fewer

than ve rides, while the top ten individuals made up 30% of overall use, each taking upwards of

200 trips over thirteen months. For these individuals, Direct Connect had become part of a weekly

roune. For most others, it remained mostly a one-o or infrequent amenity (Figure 6).

several partner agencies selected to beta test a feature allowing users to plan and book rides either

on Uber or Ly on the same plaorm as schedule service.

Despite the forward momentum for Direct Connect, several challenges came to light in the lead-up

to its formal connuaon. In December 2018, PSTA and a vising scholar agged what appeared

to be an issue with the geofence around certain stops. Aer several aempts and several months

later, Uber noed the agency that during the transion between oces, the geofence boundaries

outside the intended range. While ridership had grown slightly, gures were overstated anywhere

from 60 to 300% for any given month aer March 2018, represenng tens of thousands of rides

using a $5 subsidy. While Uber xed the glitch and never invoiced the agency for the extraneous

The Pinellas Suncoast Transit Authority has one of the lowest operang budgets for a transit agency

serving a metropolitan area of its size and, since the failure of its 2014 referendum for a consistent

funding source, has needed to cut or recongure a dozen of its scheduled routes. It is despite

that adequate support—not because of it—that PSTA has been able to establish a name for itself

One PSTA sta member was the point person responsible for developing and managing the pilot

day to day, coordinang among various project stakeholders, and overseeing the research and

planning necessary to make the project possible. While over me the responsibility for the pilot

was distributed more widely throughout the agency, the value of a few dedicated individuals in

launching complex, highly visible projects cannot be overstated.