Price optimisation with data and AI

Price optimisation uses data to determine which price best serves your goal, such as margin, sales volume or revenue, within the limits you set. A model estimates how customers respond to price and evaluates different scenarios. Dynamic pricing, where prices change with circumstances, is one form of it.

This makes your pricing strategy a data-driven choice. It is one of the AI applications we build.

Price elasticity describes how much sales volume changes when the price changes. If sales fall sharply after a small price increase, a product is price-sensitive. You estimate it from past instances when the price differed: price changes, promotions, or differences between shops or channels.

The pitfall is that a change in season, an empty shelf or a competitor’s promotion can look like a price effect. We account for these in the model. Products also affect one another: a lower price for one item can draw sales away from another.

With dynamic pricing, companies continually adjust prices in response to changing factors, such as supply and demand and competitors’ prices. That is how the ACM describes it. With personalised pricing, customers are charged different prices at the same time based on their characteristics or behaviour.

According to the ACM, you must make customers aware of personalised pricing, and consumers consider it less fair. In September 2026, the ACM called for greater transparency about pricing algorithms. Your legal adviser can assess what that means for you.

By deciding in advance what you want to optimise for. A lower price can increase sales volume while reducing the margin per item; a higher price can do the opposite. Which outcome is better depends on stock, capacity and your position in the market.

We set this out in the model as both an objective and a set of limits. An entry-level product, for example, may need to remain competitively priced even if a scenario suggests otherwise.

In physical retail, you can use pilot shops: some adopt the new price, while comparable shops keep the old one. The difference in sales volume and margin between the groups shows the effect of the price. Online, you can run an A/B test, but first check what you need to tell customers if they see different prices at the same time.

Set the test duration and the metric in advance, so you do not stop just when the results happen to look favourable.

Sales by product, including the price paid and any discount, stock levels, promotions, cost price and, ideally, competitors’ prices. For Fokker Services Group, we brought together pricing, market and logistics data from its ERP system and external marketplaces in a single overview.

If you first want to predict how much you will sell, read about forecasting. To understand who your customers are and what they are worth, see customer analysis using data.

By comparing sales volume after each price change with what the model predicted. A new competitor, a rise in purchase costs or a different season can change the behaviour the model was trained on. We then retrain the model and evaluate the scenarios again.

We never put a price live without an employee’s approval. If you work in retail or e-commerce, you can also read about data and AI in retail and e-commerce.

When prices are fixed in contracts or price lists, when you sell infrequently and only make one-off sales, or when prices have never changed, leaving nothing to learn from. A model also adds nothing if no one is allowed or able to change the price.

In those cases, a dashboard showing margin by product, customer and channel often provides more insight. See also the Fokker Services Group case study.