Wayvero: route-overlap matching and a real trust model

The demo visibly matches the table stakes — OTP-verified profiles, two-way ratings, number masking, women-only rides, recurring commutes, trip sharing and SOS — and shows off what the incumbents don’t: the overlap drawn on the map, a “why this matched” explanation, pro-rata fares for partial trips and a company CO₂ dashboard.

1. 1Index the routesEach ride’s route is split into segments. A segment goes into every 0.1° grid cell its bounding box touches, padded by the snap radius, so a lookup reads a single cell.
2. 2Snap drop, then pickupFind segments within 3 km of the drop point first: most rides fail there, so the pickup lookup only runs when something passes the drop.
3. 3Check direction and overlapThe pickup must come before the drop along the route, and the shared distance must be 0.7–1.6× the straight-line trip, which rejects routes that only brush both points.
4. 4Price the stretchCheap filters (day, seats, women-only, visibility) run before any timezone maths. The fare is pro rata to the shared kilometres, rounded to ₹10 with a ₹50 floor.

const nearDrop = nearSegments(ctx, q.to, stats)
if (nearDrop.size === 0) return noMatches // most rides fail here, cheaply
const nearPickup = nearSegments(ctx, q.from, stats)

for (const [rideId, pickupSnap] of nearPickup) {
  const dropSnap = nearDrop.get(rideId)
  if (!dropSnap) continue
  if (pickupSnap.km >= dropSnap.km) continue // wrong direction

  const matchedKm = dropSnap.km - pickupSnap.km
  const ratio = matchedKm / tripKm
  if (ratio < MIN_OVERLAP_RATIO || ratio > MAX_OVERLAP_RATIO) continue

  // Cheap filters first; date maths only for rides that survive them.
  if (dayOffset !== null && ride.dayOffset !== dayOffset) continue
  if (ride.seatsTotal - ride.seatsBooked < q.seats) continue
  if (q.women && !ride.womenOnly) continue
  if (!canSee(ride, ctx.viewer, ctx.policy)) continue
  // …departure time at the pickup, fare, score
}

• Invisible, not disabled

  Women-only rides don’t appear for men at all — not in results, counts or nearby days. A direct link shows the same neutral “This ride isn’t available” page a full or cancelled ride would, so the ride can’t be probed.

• Numbers stay masked

  Phones show as +91 98XXX XXX21 until a booking is confirmed; then a call and WhatsApp link appear. Until then chat redacts Indian mobile numbers, emails and UPI IDs, so contact can’t move off-platform early.

• Blind ratings, both ways

  Passengers rate drivers and drivers rate passengers. Neither sees the other’s review until both are in, so nobody softens a rating out of fear of retaliation.

• Share my trip

  A link for family shows the driver’s first name, the car and a masked plate with live progress — no phone numbers, no passenger surname — plus an SOS button that dials 112.

• Payment disputes

  Collect through Razorpay before the ride, release to the driver after completion, and keep a short dispute window with a ledger per booking so support can see exactly what moved and when.

• Live GPS and route deviation

  Stream driver GPS from the app. The matching engine’s point-to-segment distance already measures how far a point is from a route; alert the rider and their shared contacts when the car stays off route beyond a threshold.

• Fraud

  An OTP ties an account to a SIM, not to a person. Add device fingerprinting, velocity limits on new accounts, and flags for rating patterns that repeat between the same pairs of users.

• Insurance and commercial-vehicle rules

  Private cars can share fuel costs; carrying passengers for profit can breach commercial-vehicle rules. Keep prices near fuel-share guidance — the offer flows already suggest one — and get a legal review per state before launch.