How to Set Up an Employee Shuttle Program: What 90 Days Actually Buys

Emeryville's transportation management association issued its shuttle operations RFP on 5 May 2025. Table 1 of that document, headed "Details of the RFP Timelines and Expected Dates," runs down through a questions deadline, an addendum date, a proposal deadline, an interview period and a board vote, and ends on one line: Service Start Date, 1 January 2026. 241 days — for a shuttle network that already existed, already carried half a million passenger trips a year, and already had an operator on contract.

That number is the honest starting point for anyone working out how to set up an employee shuttle program. Ninety days is what gets promised, and 90 days is real, but it covers a specific and much narrower path than the promise implies. Inside the window: an executive mandate with a budget line, a baseline commute survey fielded in a valid season, an origin map built from HR records, a signed agreement with an operator who already owns vehicles and employs drivers, one pilot corridor carrying riders, and instrumentation live from the first trip. Outside it: a competitive procurement, vehicles of any kind, curb and stop permits in a city that regulates them, a multi-corridor network, and any defensible statement about mode shift.

That split, not the phase chart, is what a mobility program manager gets judged on in month four. The phase chart itself already exists in Ryde's employee transportation management guide, which walks the weeks from kickoff to a 30/60/90 review. The kickoff line has a queue in front of it: survey seasons, response thresholds, origin-data ownership, procurement calendars, municipal curb rules, and the supply of buses and drivers. It is the part the vendor clock assumes away, and it is where corporate shuttle program schedules actually break.

One deadline before anything else: if the site is in a jurisdiction that regulates commute surveys, the autumn field window opens in September and closes in November.

What the 90-day clock is actually measuring

Two 90-day clocks exist in this field as a matter of law, and both of them buy paperwork.

Washington State's Commute Trip Reduction law applies to worksites with 100 or more full-time employees who begin their shift between 6 and 9 a.m. on weekdays, in the state's nine most populous counties. King County Metro states the obligations plainly for employers who cross that threshold: "conduct a baseline survey within 90 days of becoming affected by the CTR law," then file a program report with the jurisdiction 90 days after the baseline results arrive. Six months of statutory clock, and at the end of it the employer owns a document. Not a route. Not a vehicle. A filed plan describing what it intends to do.

Regulators who have spent three decades watching employers stand up commute programs set that clock at a plan, because a plan is what fits inside it.

The gap between a plan and a service is where the schedule risk lives, and it is not evenly distributed. Some of it compresses with money and executive attention. Some of it does not compress at all.

Read the right-hand column and the pattern is clear enough. Everything in the top half is governed by decisions the employer controls. Everything in the bottom half is governed by a calendar somebody else owns — a manufacturer's order book, a city's permit queue, a procurement code, or the rate at which people change how they get to work.

The commute survey has a season, and this year's closes on 15 November

WSDOT does not let employers survey whenever they are ready. Its Commute Trip Reduction Guidebook restricts worksite surveys to two annual periods: "the first Monday in March through the last full week before Memorial Day," and "the third Monday in September through the last full week before Nov. 15." The stated reason is data quality. Surveying outside those dates or running the field period long "will bias the data," the guidebook says, listing holidays, inclement weather, vacations and road construction as the contaminants.

For a program starting today, that translates to a hard date. The autumn window opens on 21 September 2026 and shuts before 15 November. An employer who misses it waits until 1 March 2027, and every downstream milestone slides with it. Which is why the first question in a shuttle rollout plan is not "who are the vendors" but "which survey window can we still hit."

The field period is short by design: two weeks, with an optional third if response rates need help. Three weeks is the ceiling, not the target.

Then there is the response-rate question, where the folk rule and the published rule disagree. Most program managers arrive believing they need something like half the workforce. WSDOT's targets run the other way as sites get bigger — 70% at worksites under 200 employees, 60% at 200–500, 50% at 501–1,000, 30% at 1,001–5,000, and 10% above 5,000. The rationale is sampling: a bigger site needs a smaller share of responses before the sample holds. The guidebook is also explicit that these are "targets, not requirements," and that a worksite making a good-faith effort but falling short does not have to resurvey. The jurisdiction can be stricter. King County Metro's own survey guidance says "a survey response rate of at least 50% is required to be included in WSDOT's measurement database," and that an employer who falls short "will be required to survey again."

Low response is still a real failure mode, and Cambridge has the receipts. Its 2024 Transportation Demand Management Program Report, drawn from 24,118 employee responses across 345 workplaces, records four projects whose response rate was too low for the city to tell whether they had met their commitments. Four properties spent a year running a program and could not prove anything about it, because the instrument failed.

One more design note from the WSDOT rules, and it is the one most employers get wrong: survey everyone assigned to the worksite. Full-time remote workers. People who arrive at 2 p.m. Segment afterwards in the analysis. A survey that pre-filters to "likely shuttle riders" produces a denominator nobody downstream will accept.

Size the route against in-person days, not the weekly average

Cambridge published the most useful planning number in this field. It is a comparison rather than a single figure.

Across all workdays in 2024, single-occupancy vehicle share at reporting properties was 30%. Now filter to in-person workdays only, which is the only population a shuttle can serve: driving alone rises to 44%. The 2023 edition of the same report shows where the rest of the week goes — remote work 31% of all workdays, transit 21%, biking and walking 7% each, and on in-person days transit 30% with biking and walking at 10% each. "Half of commutes to in-person work were made by car," the 2023 report notes, counting that year's 46% driving alone plus 4% in carpools.

Fourteen points of difference, in both years. A planner who sizes a shuttle against the 30% figure has quietly removed about a third of the demand from the model before drawing a single route.

Work it through on a 1,200-person site with a three-day hybrid policy. A heavy midweek day holds about 720 people, and both Cambridge shares have to be applied to that same 720. Size it on the all-workdays share and you plan for 30% of 720, or 216 drive-alone commuters. Size it on the in-person share and you plan for 44% of 720, or 317. The 101-rider difference is capacity the weekly average never asks you to buy, concentrated into the same two hours, on the day the program gets judged on. Our midweek attendance peak analysis works the breakeven for this pattern in more detail.

Genentech's campus in South San Francisco runs on exactly this shape. The company employs roughly 7,800 people, the San Mateo Daily Journal reported in March 2026, with about 4,000 on site on a typical day and Tuesdays, Wednesdays and Thursdays carrying the heaviest load. A shuttle sized to the 7,800 is the wrong shuttle; the midweek 4,000 is the population it has to carry.

Survey data and route data are also different animals, and conflating them is a governance problem rather than a routing one. A survey measures stated behaviour and stated preference. Routes get built on home locations, which live in the HRIS, belong to HR, and carry privacy obligations that take longer to clear than anyone budgets. Start that conversation in week one. The approval to use address data is frequently the longest-lead item inside the 90 days, and it is invisible on every vendor timeline.

What two published shuttle procurements actually took

Emeryville's numbers are worth reading in full, because the association wrote them into its own RFP and then held to them.

RFP issued 5 May 2025. Proposals due 27 June, with each required to "constitute a firm offer for 120 days from the date of submittal." Review, interviews and reference checks ran 1 July to 19 August. Board approval, 16 September. Service start, 1 January 2026.

That is 241 days end to end, and 107 of them fall after the board approved the service agreement. Mobilisation — hiring, vehicle staging, systems, training, route dry-runs — took more than three months on a network the incoming operator could study in advance.

Texas Southern University ran an unrelated procurement on a different governance model in a different state, and landed one day apart. Its shuttle RFP posted 17 February 2025; the scope names service starting 17 October 2025 and running to 15 May 2026. That is 242 days. The evaluation matrix is worth stealing from directly: shuttle bus specification sheet 35 points, driver qualifications and experience 15, purchase price 50.

Two independent public buyers converging within a day of each other is a market fact, not a coincidence. Eight months is what a competitive shuttle procurement costs in calendar time.

Only one compression move buys months rather than weeks: do not run a competitive procurement inside the 90 days. Buy fleet-hours from an operator already under a master services agreement, or ride an existing regional contract, and the 241-day path collapses to the mobilisation portion. You give up three things by doing that — price tension, a documented selection record, and the negotiating position a competitive process hands you at renewal. That is a defensible trade for a pilot corridor and an indefensible one for a five-year network contract. Run the pilot on paper you already have, then run the full procurement against evidence the pilot produced.

There is no shortcut in the form of a template, and anyone offering one is selling something. What exists instead is better: four public documents an employer can open right now without handing over an email address. ETMA's RFP carries a complete employer-funded shuttle scope, per-route service specifications, cost forms, a fleet list, cleaning schedules and a workforce-retention declaration under California Labor Code §1072(a). Texas Southern's carries the weighted evaluation matrix. The FTA's Best Practices Procurement and Lessons Learned Manual, signposted from National RTAP's procurement guide, carries sample forms in Appendix B along with an independent cost estimate form. Cambridge's PTDM sample plan is a model for the plan document itself. For turning any of that into a scored shortlist, the shuttle software evaluation rubric published here already lays out eight dimensions and how to weight them.

The curb is a permit, and permits have queues

Nothing surprises first-time buyers more than the curb. Stopping a bus at a public bus stop requires permission, a fee, and a wait, and the wait is the part that moves a launch date.

Seattle is the reference case here, because the city wrote its rules down and posted the prices. Its employer shuttle program began as a shared-stop pilot in April 2017, designed over roughly the preceding three years by SDOT, King County Metro, Seattle Children's Hospital and Microsoft under a memorandum of agreement. Six months was the original pilot duration. It was extended into June 2019, evaluated in October 2018, and only in January 2023 did the City Council vote it into a permanent program of the transportation department. Call it five years and nine months from first shuttle at a shared stop to permanent status, on top of three years of prior development.

What an employer pays today is published. SDOT's 2026 schedule:

• $664 per shuttle vehicle permit, annually
• $5,000 per shared transit stop permit, annually
• $2,500 reduced stop rate for specific institutions running shuttles under an official transportation management plan
• $367 per hour of City staff time reviewing a shuttle zone or shared stop request
• Roughly $750 in one-time signage and paint costs to create a new curbside shuttle load zone

Those figures are indexed to inflation and rise every year. Permits last one calendar year, so the renewal is an annual budget line rather than a project cost.

Sequencing is where schedules break. Before an employer can even start a shared-stop application, SDOT requires shuttle vehicle permits for the entire fleet. Volume is capped as well: an employer may apply for a maximum of 20 new stops per calendar year — denied applications count against the total — may hold no more than 50 shared stops citywide at any time, and all employers combined are limited to 250 stops, under 10% of the city's transit stops. RapidRide stations and bus layover zones are excluded outright. Director's Rule 06-2023, the operative rule for the permanent program, requires a notice on an SDOT-approved form to sit at each conditionally approved stop for a minimum of ten business days before the permit issues, and any change to shuttle service to reach SDOT within five business days.

Nelson\Nygaard's evaluation of the pilot produced one engineering finding that transfers to any employer designing stops in any city. The consultants measured bus stop failure rates — the share of arrivals where every loading area is occupied and an arriving bus must wait. Every shared stop on a two-lane roadway came in below King County Metro's 5% acceptable threshold. Three of the four shared stops on single-lane roadways came in above the 2.5% threshold that applies there, and were rated unsatisfactory. Neither Metro nor the employer operators reported operational problems at any shared stop. Public sentiment, gathered through emails to SDOT, was mixed.

Design rule, then: an in-lane shared stop on a single-travel-lane street is the configuration that fails. If the candidate stop list has three of those on it, the permit conversation will go badly, and it will go badly slowly.

Vehicles and drivers are the line nobody can project-manage faster

Everything above compresses, at a price. Vehicle supply does not.

Metro Magazine's April 2026 review of the transit bus market put new bus lead times "in the two-year-plus range, up from the traditional 12 to 18 months that typified lead times when there were more suppliers." Deliveries did recover modestly — 4,690 buses in 2025 against 4,447 in 2024, a 5% increase — but the order-to-delivery clock is measured in years. Any employee shuttle program plan that includes buying vehicles has a launch date in 2028, whatever the slide says.

Drivers are the second constraint, and the procurement documents show how seriously the market treats them. Texas Southern scored driver qualifications and experience as its own 15-point category, held apart from the 35 points on vehicle specification and the 50 on price. ETMA went further and required bidders to file a workforce-retention declaration under California Labor Code §1072(a) — the statute that obliges a successor contractor to retain the incumbent's qualified employees for a transition period. That provision exists because the driver roster, not the vehicle, is the scarce asset in a shuttle handover. A new operator that cannot hire the outgoing operator's drivers cannot run the outgoing operator's service.

San Francisco's transit agency supplies the sharpest illustration of what driver scheduling actually costs. SFMTA runs an employee shuttle for its own staff, on three routes as of 6 June 2026: two morning routes running 3:30 a.m. to 6:30 a.m. every 20 minutes from downtown garages out to the operating divisions and yards, including a 4:55 a.m. trip timed to the first BART service, and an afternoon route running until 11:30 p.m. A transit agency, in a city with dense transit, running buses so its own bus operators can get to work. Shift boundaries — not average commute distance — are what shuttle design answers to.

The conclusion for a 90-day window follows directly. Buy hours, not assets. A separate build-versus-buy analysis works the five-year cost consequences of that choice, which run in the opposite direction from the speed consequences. An operator with vehicles in the yard and drivers on payroll can start a corridor inside the quarter; an order for vehicles cannot.

Instrument the pilot on day one, because cost per rider is a productivity number

Commute.org, the public agency that runs shuttles for 40-plus employers, property managers and cities in San Mateo County, publishes something almost nobody else does: per-route unit economics for an entire network.

Its FY2025 annual report spans 19 routes carrying 317,492 boardings, up 12.1% year over year, with a system-wide estimated cost of $11.08 per passenger and 250,212 single-occupant vehicle trips removed. Across the year fourteen routes gained ridership and five lost it. The per-route spread is enormous. South San Francisco's Oyster Point Caltrain route carried 22,196 riders at 14.5 boardings per service hour and $6.63 per passenger. The Hillsdale Caltrain Commuter carried 4,390 riders at 2.7 boardings per service hour and $35.28 per passenger.

Multiply each route's cost per passenger by its boardings per service hour and something falls out of the table.

Oyster Point: $6.63 × 14.5 ≈ $96. Hillsdale: $35.28 × 2.7 ≈ $95. Burlingame Point: $17.17 × 5.6 ≈ $96.

The arithmetic is derived from the published figures rather than stated in the report, and the reading is hard to escape: a contracted shuttle hour costs about the same whether three people board it or fourteen. Cost per rider is a productivity measure wearing a price tag. Boardings per service hour is the only variable underneath it that moves.

Which tells a program manager exactly what to instrument before the first trip. Boardings by trip and by stop, not by day. Service hours delivered against service hours contracted. The pairing of those two is the whole diagnostic.

Public funders already demand this. San Mateo County's shuttle call for projects requires grantees to share automatic vehicle location and automatic passenger counter data with the transit district, file quarterly ridership by stop and by trip, join the district's annual survey, and pass a network-concurrence review confirming the proposed route does not duplicate existing fixed-route service. Seattle attaches a similar condition to its shuttle permits: the employer agrees to share route data "including but not limited to routing pathways, service levels, delay and crash incidents, and passenger utilization," on a regular basis. An employer collecting none of that has no defence when the review comes and no ability to fix a route mid-cycle. The guide to shuttle service level agreements sets out the data-ownership and audit clauses to write into the contract at signature.

Day 90 answers a smaller question than the annual review

Program reviews in mature networks happen at the funding boundary, not at the three-month mark.

Commute.org's FY2025 report describes what a real corrective action looks like. Seventeen applications went into the countywide shuttle call for projects and all seventeen were approved — but the agency then merged two underperforming ferry routes, Oyster Point Ferry and Utah-Grand Ferry, into a single one-vehicle route under the Oyster Point name, and rebuilt the Hillsdale Caltrain Commuter route during the application process. Consolidation at the boundary, not cancellation mid-year. Cambridge runs on a comparable rhythm: a property files its first report about a year after the building is occupied, on monitoring windows fixed by the date the certificate of occupancy issued.

On mode shift, the published arcs are unambiguous and every one of them is measured in years. Genentech's drive-alone rate fell from 79% to 63% between 2006 and 2012, the Metropolitan Transportation Commission recorded — 16 points in six years. Stanford's own transportation data hub tracks drive-alone commuting among students, faculty and staff from 67% in 2003 to 38% in 2024. Twenty-nine points across 21 years, a shade under a point and a half a year. The hub also records a change in mode-split methodology in 2019, which puts a seam in the middle of that arc. Cambridge measured SOV share at its program properties falling from 37% in 2019 to 32% in 2023, worth roughly 16,000 fewer vehicle trips a week, then to 30% in 2024. The city notes the pandemic remote-work shift confounds the causal reading.

Progress is also not monotonic, which nobody warns you about. TCRP Report 95's long series for downtown Bellevue, Washington shows drive-alone share moving 79% in 1984, 80% in 1990, 76% in 1996, 68% in 2000, 68% in 2003 — and then back up to 71% in 2005. Two decades of program effort, and the series ends on a two-year reversal.

There is a measurement trap sitting underneath all of this. Genentech's 2025 report to the South San Francisco Planning Commission recorded 1,718 average single-occupancy morning-peak trips against a permitted cap of 5,216, with the company's buses used by 60% of commuters not driving alone, as the San Mateo Daily Journal reported. The same reporting flagged that the drive-alone definition changed between 2024 and 2025 to include motorcycles, taxis and rideshare. A definition change of that size makes a year-over-year comparison meaningless unless somebody restates the baseline. Write the mode definitions into the program charter in week one, and freeze them.

So what does day 90 legitimately produce? Load by trip against contracted capacity. Arrival timing against the actual shift boundary rather than the posted start time. No-show and cancellation patterns by stop. Whether the driver roster can hold the schedule across a full month including sickness. Those are operational questions with 90-day answers. Ridership curves at that point are still noise, for reasons the week-6 pilot analysis sets out in detail — and Cambridge's compliance data suggests the more common failure is upstream anyway. Of the nine properties that met none of their SOV commitments in 2024, five had not fully implemented their required measures, and four were still offering free or subsidised parking beside the alternative they were being judged on.

Timeline questions program managers ask

How long does it take to set up an employee shuttle program?

Ninety days is achievable for one pilot corridor if the operator already owns vehicles and employs drivers, the home-address data is available, and no municipal curb permit is needed. Add a competitive procurement and the published evidence says eight months: 241 days for Emeryville's shuttle RFP and 242 for Texas Southern University's, both measured from RFP issue to first day of service.

Do you need a permit to run an employee shuttle?

It depends entirely on where the vehicle stops. Private property, including your own campus loop, generally needs nothing beyond standard commercial vehicle compliance. Using public curb space or a public bus stop is a different matter: Seattle charges $664 per shuttle vehicle and $5,000 per shared transit stop annually, requires vehicle permits for the whole fleet before a stop application can even start, and caps each employer at 20 new stop applications a year. Check the municipal rule before the route map is drawn, because in a regulated city the permitted stop locations determine the route rather than the other way round.

How do you decide shuttle routes and stops?

Home locations from the HRIS drive the corridor, the shift calendar drives the timing, and the survey validates willingness rather than geography. Two rules do most of the work. Size against in-person days rather than the weekly average — Cambridge measured 44% drive-alone on in-person workdays in 2024 against 30% across all workdays, a gap that will cost you about a third of your capacity if you use the wrong denominator. And check the physical stop: SDOT's pilot evaluation found every shared stop on a two-lane road performed acceptably, while three of four on single-lane roads failed the applicable threshold.

How much does an employee shuttle cost per rider?

Commute.org's FY2025 network of 19 routes averaged $11.08 per passenger, with individual routes running from $6.63 to $35.28. The spread tracks boardings per service hour almost exactly, because a contracted vehicle hour costs roughly the same regardless of who is on board. Treat any per-rider quote as a statement about expected occupancy, not about price.

How long should a shuttle pilot run before you judge it?

Long enough for commute habits to form, which is longer than a quarter. Ninety days answers operational questions — load, timing, no-shows, driver roster durability. It does not answer the mode-shift question; the published arcs for that run six years at Genentech and 21 at Stanford. On the US tax treatment of what you hand riders, the 2026 commuter benefits guide covers the §132(f) position.

The two decisions that set your launch date

Both can be made this week, and neither of them needs a vendor in the room.

The first is the survey window. Find out whether your jurisdiction regulates commute surveys, and if it follows the WSDOT pattern, work out which window you can still field in. The autumn one closes before 15 November; the next opens on the first Monday in March. That single fact reorders the whole schedule, because everything downstream is built on a baseline you cannot legitimately collect at an arbitrary time of year.

The second is whether you are buying fleet-hours or a fleet. An operator with vehicles and drivers already in service can put a corridor on the road inside 90 days. An order for vehicles puts the launch two years out, and no amount of program-management energy changes that. Pick the first for the pilot, and use what the pilot produces as the evidence base for a full procurement you run properly on an eight-month calendar.

Everything else is sequencing. Get HR's approval to use home-address data in week one, because it is the longest-lead internal item and the least visible. Freeze the mode definitions before the baseline, so the year-two comparison survives. Write the AVL, passenger-count and per-trip reporting obligations into the operator agreement while the price is still open. Then commit to the number you can actually defend: one corridor, instrumented, in service by day 90, with the network decision scheduled for the following budget cycle.

If you want a second read on your own dates before you commit to them, Ryde's smart employee commuting team offers a 20-minute walkthrough.

Sources

• Commute Trip Reduction — employer programs — King County Metro. Accessed 2026-08-30.
• CTR surveys and reports — King County Metro. Accessed 2026-08-30.
• Commute Trip Reduction Guidebook M 3136 — Washington State Department of Transportation, 2023–2025 biennium edition. Accessed 2026-08-30.
• Transportation Demand Management Program Report 2024 — City of Cambridge, Massachusetts. Accessed 2026-08-30.
• Transportation Demand Management Program Report 2023 (with errata) — City of Cambridge, Massachusetts. Accessed 2026-08-30.
• PTDM Sample Plan — City of Cambridge, Massachusetts. Accessed 2026-08-30.
• Request for Proposals: Shuttle Operations and Maintenance — Emeryville Transportation Management Association. Accessed 2026-08-30.
• RFP #717-25-715 — Shuttle Bus Services — Texas Southern University. Accessed 2026-08-30.
• Employer Shuttles Program — Seattle Department of Transportation. Accessed 2026-08-30.
• Employer Shared Transit Stop Pilot Evaluation Report — SDOT / Nelson\Nygaard Consulting Associates, October 2018. Accessed 2026-08-30.
• Employer Shuttles Program Director's Rule 06-2023 — Seattle Department of Transportation, published November 2023, superseding draft Director's Rule D-2022. Accessed 2026-08-30.
• Recovery and Risk Define the Transit Bus Market in 2025 — Metro Magazine, 20 April 2026. Accessed 2026-08-30.
• Annual Report, Fiscal Year Ending June 30, 2025 — Commute.org (Peninsula Traffic Congestion Relief Alliance). Accessed 2026-08-30.
• Shuttle Call for Projects Guidelines FY 2024 and 2025 — San Mateo County Transportation Authority & C/CAG. Accessed 2026-08-30.
• MTC Honors Contributions to Bay Area Transportation — Metropolitan Transportation Commission, 23 October 2012. Accessed 2026-08-30.
• South City's Genentech saw limited development in 2025 as part of annual update — San Mateo Daily Journal, 27 March 2026. Accessed 2026-08-30.
• Transportation data hub — Sustainable Stanford. Accessed 2026-08-30.
• Transit Managers Toolkit — Procurement — National RTAP. Accessed 2026-08-30.
• TCRP Report 95, Chapter 19: Employer and Institutional TDM Strategies — Transportation Research Board, 2010. Accessed 2026-08-30.
• SFMTA Employee Shuttle — San Francisco Municipal Transportation Agency. Accessed 2026-08-30.