Dynamic pricing
Aussi : Dynamic pricing, Real time pricing, Surge pricing, Demand based pricing, Algorithmic pricing, Tarification dynamique
Automatically adjusting prices in real time based on demand, competition or user behaviour to optimise revenue, margin or conversion.
What it is
Dynamic pricing is the practice of changing the price of a product or service automatically, and often frequently, in response to live signals such as demand, remaining inventory, competitor prices, time of day, or customer context. Instead of a single fixed price set in advance, the price becomes a variable computed by rules or models.
It sits on a spectrum:
- Rule based: explicit business logic (for example, raise price 10% when stock drops below 20 units).
- Model based: statistical or machine learning models that estimate willingness to pay and price elasticity.
- Real time bidding: prices recomputed per request, common in travel, ride hailing, and online ads.
Why it matters
Pricing is one of the highest leverage decisions a business makes. A small improvement in average price captured usually flows almost entirely to profit, because the cost base is unchanged. Dynamic pricing lets a company:
- Capture more value when demand is high and protect volume when demand is low.
- React to competitors within minutes rather than quarterly reviews.
- Reduce waste of perishable or time bound inventory (flights, hotel rooms, event seats).
It also carries risks: customer perception of unfairness, regulatory scrutiny (price gouging, discrimination), and the danger of automated feedback loops that spiral (algorithmic collusion or runaway markdowns).
How it is used in practice
A typical pipeline:
1. Collect signals: demand, inventory, competitor feeds, seasonality, user segment.
2. Estimate elasticity: how volume responds to price for each segment or SKU.
3. Optimise: choose the price that maximises the objective (revenue, margin, or conversion) within guardrails.
4. Apply guardrails: floors, ceilings, fairness constraints, brand rules.
5. Monitor and learn: A/B tests and bandits feed results back into the models.
Worked example
An online hotel platform has 100 rooms for a Saturday. Two weeks out, 30 rooms are sold at a base of 120 EUR. The model detects a local concert and books filling faster than forecast. It raises the price to 165 EUR, slowing bookings but lifting average revenue per room. If a competitor drops prices and pace stalls, a guardrail triggers a markdown back toward 130 EUR to avoid empty rooms.
Result: average realised price of 148 EUR versus 120 EUR fixed, a 23% revenue uplift on that night, achieved without new inventory.