Card-based fare collection system
The card carries stored value or an entitlement, and the validator deducts locally. That's why it works on a minibus climbing an unpaved road with no signal. Kigali's Tap&Go runs this way, and the Rwanda deployment is the strongest proof that offline-capable automated fare collection suits flat-fare, single-operator-zone networks.
The cost shows up as operations. You issue media and replace it when it breaks. You also secure the keys and synchronize every device so that yesterday's hotlist reaches today's bus. AC Mobility, formerly AC Group, reports 3 million Tap&Go cards issued and over 1,000 registered agents supporting them. That agent network is the product as much as the card is.
Fare flexibility is the other limit. Rwanda's regulator wanted distance-based fares and couldn't get there, because the stored-value model constrained what the system could calculate at the point of tap. If you're planning zonal pricing or caps, inventory that constraint now.
Account based ticketing
Here the fare is computed centrally, after the tap, from a journey record. The media is only a token, which means a closed-loop card or a bank card can point at the same account, and so can a QR code on a phone or a registered identifier. That's what makes multi-operator interoperability tractable without reissuing anything.
Account based ticketing earns its complexity on fare rules. Daily caps and transfer windows live in software you can change without touching a single device in the field, as do concession entitlements and post-trip corrections. Maputo's FAMBA system was structured as a 10-year public-private partnership with revenue centralized under the metropolitan transport agency, which is the institutional arrangement this architecture requires.
The dependencies are real and they're mostly invisible at tender stage. Backend payment infrastructure was cited more often than any other barrier to open-loop payments in a 2023 survey of US transit agencies. Identity mapping and settlement controls have to work before the first tap, as does a top-up route for passengers without bank accounts, because account based ticketing fails silently when the account layer is wrong.
Design for exceptions
Normal operation is the easy case. Spend most of your specification's pages on what happens when things break, because that's where revenue leaks and passengers get stranded. Write the expected behavior for each failure before a vendor writes it for you.
-
Connectivity loss. The validator accepts taps against a locally cached deny list and queues transactions. Define the maximum offline window and the cap on cumulative offline spend per token.
-
Delayed uploads. Journey records arriving six hours late still have to join the correct fare calculation. Specify how the back office reopens a settled day, and who signs off.
-
Repeated taps. A second tap within a defined window is a duplicate. Pick the window and make it configurable per mode.
-
Insufficient funds. Decide whether the passenger travels and owes, or is refused. Littlepay reports that a managed deny list keeps fare losses at 0.2%, which is the benchmark to hold a vendor to.
-
Incomplete journeys. A missing tap-off needs a default fare that doesn't punish the passenger for your device failure, plus a correction path when the record arrives late.
-
Device failure and refunds. Name the replacement time and the balance recovery method, plus who funds the refund.
Visa's acceptance specification requires the contactless card read under 500 milliseconds, and your exception logic has to run inside that budget. Anything that needs a round trip to the back office before the gate opens will miss it. That's the single hardest constraint in automated fare collection engineering, and it's why offline decision rules exist at all.
Allocate revenue transparently
Shared journeys are where multi-operator schemes break politically. One operator carries the first leg and another carries the transfer. A daily cap means the second leg earned nothing extra, and now two companies disagree about a number neither can audit. Account based ticketing makes the calculation possible. Apportionment rules have to be configurable and published to participants.
Concessions need a funding source named in the rules, because a discounted student fare is a subsidy someone pays. Incomplete records need a default allocation rather than a dispute queue that grows forever.
Settlement timing drives operator cash flow and therefore buy-in. Patrick Buchana, co-founder of AC Group, explained what the Kigali operators were actually trying to fix: "I found out that bus companies were losing a big portion of their revenue to cash handling. This was mainly revenue lost due to leakage between the driver and the conductor who collected the money." Daily settlement replaced daily cash, and that's the trade operators accepted. Build the audit trail so any participant can reconstruct a single journey's split, because automated fare collection without that capability becomes a trust problem the technology can't solve.