AI dynamic pricing, with guardrails your team controls
AI dynamic pricing changes prices as demand, stock, season and time to departure or expiry change, using a model that estimates how customers respond to price. DT Nova builds dynamic pricing with artificial intelligence for online stores, travel, events, rentals and services: the model proposes, your rules decide the floor, the ceiling and how often prices move, and the EU rules on announced reductions and personalised prices are built in from the start. Data stays in the EU as standard and the code is yours.
- You see the result before it goes live
- Fixed price for the agreed scope, in writing before work starts
- Source code, domain and accounts in your company's name
Where dynamic pricing earns its keep
- Perishable capacity. Hotel rooms, seats, rental cars, appointment slots and event tickets, where an unsold unit is worth nothing after the date. For hotels, see also artificial intelligence in hospitality.
- Short shelf life. Fresh food and seasonal goods, where markdowns should start early enough to sell through.
- Large catalogues. Online stores with thousands of products, where no one can review every price by hand; see online store development.
- Uneven demand. Services with clear peaks and troughs by hour, weekday or season.
- Clearance and stock balance. Moving slow stock before it ties up cash and warehouse space.
It earns less where prices are fixed by contract, where customers compare one price over years, or where a handful of products and a clear price list work fine. We say so when a rule-based markdown plan or a better price list is the right answer.
How AI dynamic pricing works
The system has three layers. A demand model estimates how many units will sell at different prices, given the date, stock, history, channel and other signals. An optimiser picks the price that best meets your goal, such as margin, revenue, sell-through by a date or a mix, within the limits. And a rules layer applies your business limits before any price goes live:
- floor and ceiling per product or group, including minimum margin;
- maximum change per step and per day, so prices never jump;
- products, customers and channels excluded from changes;
- price relationships, such as the larger pack never costing less per unit than the smaller one;
- quiet periods, such as no changes during a campaign already announced.
The demand model is the part that uses machine learning. It is trained on your sales history and, where history shows too little price variation, on controlled tests: a small share of products gets slightly different prices for a set period so the effect of price can be measured rather than guessed.
The EU price rules built in
Dynamic prices are legal in the EU, and two rules decide how they are shown.
Announced reductions. Article 6a of Directive 98/6/EC (consolidated text read 11 November 2026) says any announcement of a price reduction must indicate the prior price, defined as the lowest price applied by the trader during a period of not less than 30 days before the reduction; Member States may set different rules for goods that perish quickly or products on the market for less than 30 days. A pricing system that changes prices often must therefore store the full price history per product and compute that lowest price automatically whenever a sale badge or struck-through price is shown. We build that into the price history from day one, because it cannot be reconstructed later.
Personalised prices. Directive 2011/83/EU, Article 6(1)(ea) as added by the Omnibus Directive (EU) 2019/2161 (both read 11 November 2026), requires informing the consumer, where applicable, that the price was personalised on the basis of automated decision-making. Recital 45 of Directive (EU) 2019/2161 explains that this duty does not apply to dynamic or real-time pricing that responds to market demand without personalisation based on automated decision-making. Our default is dynamic pricing that is the same for every customer at a given moment. Personalised prices are built only on request, with the disclosure in the checkout, a separate review under the GDPR rules on profiling and automated decisions, and a legal opinion from your counsel.
Competition law and competitor prices
Watching competitors' public prices is normal. Agreeing prices with them, or using a shared tool that lets competitors align their prices through it, is not, and pricing algorithms do not change that. We use competitor prices only from lawful public sources, never share your prices or rules with other clients' systems, and keep each client's model and data separate. When a product's price is driven mostly by one competitor's moves, the rules layer limits how closely it follows, so two automatic systems cannot spiral prices up or down against each other.
Trust and transparency
Customers notice prices that change while they shop. We recommend, and build by default, prices that stay stable within a session and a basket, no change between product page and checkout, clear wording where prices depend on date or demand, and no price increases triggered by a customer's own repeated visits. These choices cost little margin and avoid the complaints that end dynamic pricing projects.
Connecting to your systems
Prices have to reach the store, the booking engine, the marketplace listings and the point of sale consistently. We connect to the e-commerce platform, ERP or property management system through their interfaces, read sales, stock and returns, and write approved prices back with a log of every change. Shelf labels and printed menus need their own timing, which the rules layer respects. For integration work in general, see API integration; for store apps, e-commerce app development.
From proposals to automatic prices
- Goal and limits. What the prices should achieve, the floors, ceilings and exclusions, and who approves changes.
- Data check. Sales and price history, stock, calendar and channel data; where history has little price variation, a plan for controlled tests.
- Proposals. The system suggests prices with the reason for each; a person approves them in a simple review screen.
- Measured test. Dynamic prices on part of the range, a comparable part kept on the old prices, and the difference in margin and sell-through measured.
- Automatic within limits. Groups where proposals proved reliable move to automatic changes; the rest stay on approval.
- Run. Retraining, drift alerts, a monthly review and a single switch back to the manual price list.
AI rules that apply
Pricing models for goods and services are generally not among the high-risk uses listed in Annex III of the EU AI Act, whose obligations apply from 2 December 2027; the Act's general application date was 2 August 2026. Insurance pricing for life and health and credit decisions about individuals are different and are handled as their own projects. Whatever the classification, we document what the model uses, how it was tested and who can override it, because that is what your team will need when a customer or regulator asks. Broader AI work is covered on our AI consulting page, and demand forecasting for stock in our guide to predictive analytics for retail. If you also score customers for marketing, see predictive analytics for marketers. See prices.
You own it
The models, the rules, the price history and the code are yours, in your repository and your cloud account, hosted in an EU region as standard. The rules are written in plain terms your commercial team can read and change, so pricing decisions stay with your company.
Questions we get
What is AI dynamic pricing?
Changing prices over time based on demand, stock, time to departure or expiry, competitor prices and other signals, with a machine learning model estimating how demand responds to price. The model proposes prices; rules set by your team decide the floor, the ceiling, how often prices may move and which products are excluded.
Is dynamic pricing legal in the EU?
Yes. Prices may change with demand. Two EU rules shape how: when a price reduction is announced, Article 6a of Directive 98/6/EC requires showing the prior price, the lowest price in at least the 30 days before the reduction; and when a price is personalised for a consumer on the basis of automated decision-making, Directive 2011/83/EU requires telling the consumer before the purchase. Competition law also applies, so prices are never set by sharing information with competitors.
What is the difference between dynamic and personalised pricing?
Dynamic pricing changes the price for everyone as conditions change, such as demand on a given date. Personalised pricing shows different customers different prices based on data about them. Recital 45 of Directive (EU) 2019/2161 says the disclosure duty for personalised prices does not apply to dynamic or real-time pricing that responds to market demand without personalisation based on automated decision-making. We build dynamic pricing by default; personalised pricing only with disclosure and a separate legal review.
How much data does a pricing model need?
Enough price variation in the past to see how demand responded. If prices have rarely changed, the model cannot learn price sensitivity from history and we start with controlled price tests on a small share of products. Sales history, stock, calendar and, where relevant, competitor prices from lawful public sources are the usual inputs.
Will the system change prices on its own?
Only within the limits you set. Most clients start with the system proposing prices that a person approves, then allow automatic changes for products and ranges where the proposals have proved reliable. Every change is logged with its reason, and a single switch returns all prices to the manual list.
Get a written quote
Tell us in a few lines what you need. We reply within one working day with a written quote: scope, timeline and a fixed price or an hour estimate. Asking is free and commits you to nothing.