Decisions that put the rider ahead of the milestone

Holding a fare launch until the validators survived a prairie January

In December 2021 we installed new validators on 40 buses at a prairie agency of 260 buses that had signed for our account-based fare platform. The units had passed laboratory testing at minus thirty. On the first morning below minus thirty-four, with doors cycling open at every stop, 11 per cent of the validators took up to twenty minutes to read cards after a cold start, and some rebooted repeatedly as condensation formed once the buses warmed. Full launch was planned for March 2022, carrying a C$4.2 million milestone, and two more agencies were due to follow on the same hardware.

Our chief revenue and chief financial officers wanted to launch on time with a firmware fix and a field-replacement plan. The agency's revenue manager worried about something else: if validators accepted cards they could not fully check, fare evasion would rise. I recommended stopping the rollout at all three agencies until the hardware had passed a full field winter, and changing the design so that a validator unable to complete its checks would accept the tap, store it and settle it later against a list of cards that had failed payment. Our chief executive supported the delay, and the agency agreed after its board was briefed.

The milestone moved eleven months, into our 2023 revenue. We spent C$1.9 million retrofitting 2,300 validators with heated reader assemblies, coated circuit boards and a faster start-up sequence, and paid C$0.6 million to keep the agency's old fareboxes maintained in the meantime. Another agency postponed its launch by a year.

The field test in January 2023 recorded cold-start failures of 0.4 per cent at minus thirty-five, and the agency launched in April 2023. Offline acceptance has since accounted for about 0.2 per cent of taps, and fares lost through it for less than 0.1 per cent of fare revenue. Every Halvery hardware specification now requires a January field test in a prairie city before any launch milestone can be invoiced.

I would do two things differently. I would refuse to sign a milestone tied to a winter launch before the hardware had seen a winter. And I would have asked the agency's drivers sooner: several had noticed the problem on the pilot buses within a week and had no route to tell us.

Making French and English ship together or not at all

When I became chief product officer in April 2019, French versions of our rider features followed the English ones by six to ten weeks. Translation happened at the end, by contractors working from screenshots. Agencies in Quebec and New Brunswick, then 19 per cent of our revenue, had complained for years. The largest, a Quebec agency whose seven-year contract worth C$11 million was due for renewal in 2021, told us its next competition would score language parity explicitly. Its riders were reading French-language reviews of features they could not yet use.

I proposed a single rule: no rider-facing feature would ship in one language until it was ready in both. Our vice-president of engineering objected that every release would move at the pace of the slowest translation. Our sales lead for Ontario objected more sharply. He was pursuing an Ontario agency whose competition closed in early 2020 and needed an English trip-planner feature that the rule would delay by a month. Both concerns were sound. I kept the rule, because a rider's language should not decide when a finished product starts working for them.

We moved eleven French UX writers into product teams, took stop and street names from each agency's official gazetteer and made language review a release gate beside security. Our cadence slowed from a release every two weeks to one every three for two quarters. The Ontario feature arrived late and we lost that competition, a contract worth about C$7.5 million over five years; the agency's debrief listed the missing feature among its reasons.

No feature has shipped in one language first since November 2020. The French-language rating of our rider app rose from 3.1 to 4.4, French support tickets fell by 38 per cent and the Quebec agency renewed in 2021. Two more Quebec agencies chose us, in 2022 and 2024, and the three contracts together are worth C$19 million. Parity is now part of how the company defines a finished feature.

I would change the sequence. For the first two months the bottleneck simply moved from translation to review, because I changed the rule before we had hired the writers. I would also have called the Ontario agency myself to explain the delay instead of leaving it to a written bid response.

Taking Halvery's income out of the agencies' fare policy

In 2021 nineteen of our fare-platform contracts paid Halvery between 3.5 and 5.5 per cent of the fare revenue we processed, plus card-processing costs. Agencies paid little up front, and our revenue grew with ridership. The pandemic showed the other side. When ridership collapsed in 2020, so did our income from those contracts, while our costs did not. Then agencies began to change fare policy. Several wanted fare capping, under which a rider stops paying once their taps reach the price of a pass. Two cities were weighing free fares for children and reduced fares for low-income riders. Each of those decisions reduced our revenue, and in one city our account team had begun questioning capping in meetings with agency staff.

I proposed moving every renewal to a fixed annual fee per equipped vehicle and per station, with card-processing passed through at cost plus a disclosed fixed margin per transaction and service credits when availability fell short. Our chief financial officer and two directors appointed by Brackenridge Lane Partners opposed it. They expected ridership to return to its 2019 level by 2025 and estimated the upside we would forgo at C$8 million to C$10 million a year by 2026. A sales director argued that percentage pricing let us charge large agencies more without publishing a higher list price. We agreed a compromise: fixed pricing for every renewal from 2022 and every new contract, with existing contracts left untouched until they expired.

By 2025 fourteen of the nineteen contracts had converted. Those fourteen earned C$2.3 million less in 2023 than percentage terms would have produced, and about C$3.1 million less in 2025. All fourteen renewed, against a historical renewal rate of 86 per cent. Six of the agencies have since introduced fare capping or free fares for children without renegotiating a clause. Our account teams no longer hold views on fare policy, and sales compensation no longer refers to fare revenue at all.

What I misjudged was how agencies budget. A percentage fee was netted out of fares and rarely seen by councils, while a fixed subscription appears as an operating cost and, at two agencies, needed a council vote that delayed renewal by a quarter. I would now offer smaller agencies a transitional schedule and help prepare the council report.

Keeping the telephone an equal channel for specialised transit

Specialised transit carries riders who cannot use fixed-route buses: people who use wheelchairs, people who are blind, and people with cognitive or other disabilities. Trips are booked a day or more ahead, and on busy days there are more requests than vehicles. In 2023 we were rebuilding our booking product, used by eleven agencies. Our largest customer for it, with 11,000 registered riders and about 640,000 trips a year, asked for two changes. It wanted app-only booking by 2024, saving about C$1.4 million a year in its call centre, and in the meantime an hour's head start in the booking window for riders who used the app. Our own product manager for the line supported the earlier window, believing riders would follow the incentive.

Across our customers, 46 per cent of bookings were still made by telephone and 22 per cent online; the rest were standing weekly trips. Telephone users were disproportionately older riders and people with visual or cognitive disabilities. A head start for the app would hand the scarcest trips to the riders least likely to need help. I declined both requests and proposed one booking engine, used by riders in the app and by agents on the telephone, offering the same trips at the same moment, with an agent screen designed to shorten every call.

The agency was unhappy. It withdrew a C$2.1 million change order, delayed its renewal by a year and told us it would look at other suppliers. Our product manager disagreed with me openly in a design review. I asked her to lead the new design anyway, which she did very well, working through each stage with our rider accessibility panel.

By 2025 online booking across our customers had risen to 49 per cent without any incentive. Call handling time fell by 31 per cent because agents and riders used the same screens, and the share of requests that could be scheduled differed by less than half a percentage point between channels. The agency renewed in 2024 for five years, and its call-centre costs fell by C$0.6 million a year without closing the line.

I would build the channel-equity report first. Had the agency seen how bookings succeeded by channel before the disagreement began, it would have been an argument about evidence rather than about values.