What is dynamic pricing?
Dynamic pricing is the practice of changing prices often, in response to demand, remaining capacity, timing, and competitor prices. Airlines and hotels have priced this way for decades. Machine learning extends the method to businesses that set thousands of prices, such as online retailers, parking operators, and freight brokers.
A pricing system is a form of prescriptive analytics with two parts. A demand forecasting model estimates how many units will sell, or how likely a quote is to win, at each candidate price. An optimization step picks the price that best meets the goal, such as margin, volume, or clearing perishable stock, within floors and ceilings the business sets. Some systems test prices and learn from the results, a form of reinforcement learning.
Dynamic pricing in practice
A freight broker quotes hundreds of spot-market truckload shipments a day. A model trained on past quotes, wins, losses, carrier costs, and seasonal capacity recommends a price for each lane and shows the estimated chance of winning. The rate desk adjusts the recommendation when it knows something the model does not, such as a storm closing a highway, and sends the quote.
What to watch for
- Prices that vary by customer can conflict with consumer protection, fair lending, or insurance rules, so have counsel review which inputs are allowed.
- Frequent swings erode customer trust, so limit how far and how often a price can move.
- Pricing tools that pool competitors’ data have drawn antitrust scrutiny, so confirm where every input comes from.