The profile as a contract
Stated preferences turned an awkward, mid-journey negotiation into a pre-trip filter — and made the match itself do the social work.
The profile is read once, quickly, at this size. Everything it has to establish, it establishes here.
Photograph from the tripda.com picture kitHow a checkbox became a commitment
The first problem a carpooling platform has to solve is not routing — it is the five minutes of silence after a stranger climbs into your car and neither of you knows the rules. Do you talk? Is smoking allowed? Can they eat? Will there be music, and whose? These are small things, and they are nearly impossible to negotiate in the moment without one party feeling either rude or resentful. Left unresolved, they are the friction that turns a single trip into a non-repeating experience.
The answer that platforms like BlaBlaCar converged on was to move the negotiation out of the car and into the listing. A driver publishing a seat from Paris to Lyon would declare, before any match was made: talks a lot or prefers quiet, music yes or no, smoking never, one bag per passenger. These were not suggestions. Choosing them was public and searchable; a passenger who booked after reading them had implicitly agreed. The profile had become a contract — not a legal one, but a social one, and in practice that is the binding kind.
The mechanism matters because of what it replaced. Without stated preferences, passengers were buying a seat; with them, they were buying a defined environment. The variance in the product collapsed. A rider who values silence could filter for it and arrive at a match already knowing the driver shared that preference. No awkwardness, no mid-journey request, no reluctant concession. The preference field did not just inform — it sorted, and sorting is the service.
The category owns no infrastructure. It borrows this: a car park off a trunk road where two strangers agree to meet.
Photograph: Cullompton motorway service area · geograph.org.uk 3218604 · Wikimedia CommonsThe arithmetic of stated preferences
There is a structural reason why the preference layer works at scale even when individual preferences conflict. Research on preference aggregation in two-sided markets shows that making preferences explicit shifts the cost of mismatch from post-match to pre-match, where it is cheap. A declined booking costs nothing. A three-hour car journey with someone whose habits you find intolerable costs significantly more — in discomfort, in a bad rating, in reduced likelihood of returning to the platform.
BlaBlaCar's preference taxonomy was also deliberately coarse. The "Bla Bla Bla" scale for conversation level — silent, some talk, lots of talk — sacrificed precision for legibility. A five-point Likert scale for chattiness would have produced richer data and worse user behaviour: people overthink continuous scales and underthink binary ones. Three options with cartoon icons resolved in under a second and were completed. Completion is everything; an unfilled profile field is a filter that does not filter.
Smoking is the cleanest case. It is binary, it has a physical consequence (smell), and it is the one preference that cannot be renegotiated politely mid-journey. Platforms learned early that surfacing smoking preference prominently — and making it part of automated match filtering, not just visible text — dramatically reduced the complaint rate for that specific friction point. The preference field here was essentially a contract clause with an enforcement mechanism: a reported violation was grounds for a bad rating, and a bad rating affected future match probability.
Trust as product design
What the stated-preference system reveals is that trust on a platform is an engineering problem before it is a social one. The question is not whether two people will behave well together — it is whether the product architecture makes mismatches visible before they become disputes. A reputation system handles the past; a preference system handles the future. They do different work, and a platform that conflates them — relying on ratings alone to produce good matches — is using the wrong instrument.
The profile as contract also had a secondary effect: it created accountability without surveillance. Neither party needed to monitor the other; both had publicly committed to terms before the journey began. The commitment was lightweight — checkboxes, not signatures — but it was recorded, searchable, and attached to a profile that accumulated history. That combination of low friction and high visibility is precisely what makes stated preferences stickier than a verbal agreement in the parking lot, which evaporates the moment the door closes.
The match is settled before the engine starts. What the platform sold was not routing but recognition.
Photograph from the tripda.com picture kitStated preferences — how much conversation, smoking, luggage, music — turned an awkward negotiation into a filter.