AI Dynamic Pricing Engine

Used motorbike prices in Vietnam move fast. To price one, staff had to weigh brand, model, year, mileage, condition, local demand and recent sales. It took an experienced person up to 60 minutes per bike, and two people would still come up with different prices.

Golden Sea trained a TensorFlow model on more than 100,000 past transactions from the client's own marketplace. It learned how prices behave across hundreds of make and model combinations, and it adjusts for the current market, the season and demand in each region.

Staff now photograph the bike, enter the basic specs and get a suggested price within five minutes, with a confidence range and recent comparable sales. Each suggestion also lists the five factors that moved the price most, so staff can see the reasoning and override it when they know better.

After rollout, pricing time fell from 60 minutes to under five per bike. Average profit margin rose 30%, because fewer bikes were priced under market or left unsold for being too expensive. The project paid for itself in under 90 days.