← Selected work
C-08 Civic & mobility

Bike-share rebalancing prototype

A user-driven approach to rebalancing a city bike-share network, backtested against millions of real trips before anyone touched the street.

Real project · client withheld · outcomes discussed under NDA

  1. 01 Backtest against real trips
  2. 02 An imbalance recommender
  3. 03 Pilot simulation
1730 / C-08 / Decision model Build path / 3 stages Data modelling

Bike-share networks drift out of balance — bikes pile up where people ride to and run dry where people want to start. Operators usually fix this by trucking bikes around, which is expensive and always a step behind demand.

The problem

Testing a different idea — nudging riders themselves to help rebalance — is risky and slow if the only way to try it is on the real network. The question was whether such a scheme would actually move the needle, and that had to be answerable before a pilot, not after.

What we did

  1. 01
    Backtest against real trips

    The approach was evaluated against millions of historical trips, so its effect could be estimated on real demand patterns rather than a toy model.

  2. 02
    An imbalance recommender

    A component that identifies where the network is going out of balance and what would help, as the basis for user-facing nudges.

  3. 03
    Pilot simulation

    A simulation of how a real-world pilot might play out, so the idea could be pressure-tested and sized before committing to it on the street.

Relevance

Where this pattern applies.

  • You need to test an operational policy or product idea against historical data before a real pilot.
  • The decision depends on backtesting, optimization, or simulation — not a polished interface alone.
How we would start

A comparable engagement starts with the data you already have — historical trips or usage logs — and a backtest of the idea against it, so you learn whether it is worth piloting before building anything operational.

Start a conversation → Next Bilingual artist portfolio →