About the role
Our dynamic pricing is rule-based today โ conditional logic on surface signals like weather and supply-demand ratios. It works, but it’s rigid, and we think it’s leaving money on the table.
Our platform team is strong on data engineering and business analysis. What we don’t have in-house is the applied economics or algorithmic depth to work out which pricing models are actually right for us. Other platforms have solved versions of this already. We want someone who can tell us which approach fits, prove it on our data, and hand it to our engineers to build.
6-month contract, fully remote and open to candidates anywhere, reporting to the Director, Platform.
What you will deliver
- A read on our existing pricing logic, with a number on the revenue and matching efficiency we’re currently giving up
- A shortlist of approaches โ mechanism design (VCG-style allocation), algorithmic search (optimal tree search, MCTS), predictive and ML pricing models โ weighed against our data maturity, latency limits and commercial goals
- A recommended model, tested offline on GoGoX order and supply data, with the assumptions and failure modes spelled out
- An experiment design for live testing, including guardrails against cannibalising users or driver-partners, and the metrics that call it a win
- A handover pack our data engineers can build from, plus an honest view of what we’d need to hire or train for to own this long term
Who you are
- 5+ years applying economic or optimisation modelling to pricing, matching or incentives in a live two-sided marketplace โ ride-hailing, logistics, delivery, travel or e-commerce
- Comfortable across mechanism design, causal inference (diff-in-diff, regression discontinuity, synthetic control), demand elasticity and structural modelling, and modern ML approaches to pricing
- Strong in Python and SQL, and happy working in production data next to engineers rather than handing over a paper
- Master’s or PhD in Economics, Operations Research, Statistics or similar โ or the industry track record that stands in for it
- Pragmatic about constraints. You recommend what works with the data and engineering we have, not what’s most elegant
- Can explain the economics to a commercial audience. We can’t act on a recommendation we can’t interrogate
- Fluent spoken Cantonese
