A homeware store I worked with once ran a manual "competitor check" every Monday: someone opened five rival sites, wrote prices into a spreadsheet, and adjusted their own by hand. By Wednesday the spreadsheet was already stale. By the time a supplier price hike hit, they were three weeks behind and selling a bestselling lamp at a loss without realising it. That lag — not the pricing logic itself — is the actual problem dynamic pricing solves.
Dynamic pricing has a bad reputation in some corners, largely thanks to airlines and ride-hailing apps changing prices mid-session in ways that feel punitive. Done properly in eCommerce, it's closer to what a good store manager already does instinctively — clearing slow stock, protecting margin on scarce items, matching the market — just automated and continuous instead of weekly and manual.
What "AI-powered" adds over rule-based pricing
Plenty of stores already do "dynamic" pricing without AI: a rule that says "match the lowest competitor price minus 2%" or "drop price 10% if stock age exceeds 60 days." These rules work, but they're reactive and single-variable — they respond to one signal at a time and can't weigh trade-offs.
An AI pricing model, by contrast, is typically trained to optimise for a target (revenue, margin, or sell-through) across many correlated signals at once:
- Historical demand elasticity — how much unit sales actually moved the last time this SKU's price changed
- Competitor price movements, scraped or fed via a price-intelligence API
- Inventory position and days-of-stock remaining
- Seasonality and promotional calendar effects
- Cost inputs — supplier cost, shipping cost, FX rate for imported goods
The model outputs a recommended price (or price band) per SKU, updated on a schedule — hourly, daily, or event-triggered — rather than a human rewriting a spreadsheet rule every time the market shifts.
Two broad approaches merchants actually use
| Approach | How it works | Best fit |
|---|---|---|
| Rule-based dynamic pricing | If/then logic on 1-2 signals (competitor price, stock age) | Small catalogues, predictable demand, low engineering budget |
| ML-based dynamic pricing | Model trained on demand elasticity + multiple signals, outputs optimised price | Larger catalogues, seasonal or fashion goods, teams with margin data to train on |
Most PrestaShop merchants I've seen succeed with dynamic pricing start with rules, prove the ROI on a handful of SKU categories, then layer in a trained model only once they have enough clean sales history (price, quantity, date) to make training worthwhile — usually at least 12-18 months of transaction data per category.
Where the margin actually comes from
Dynamic pricing doesn't magically create margin; it reallocates it. Three realistic scenarios, with illustrative figures:
- Clearing aging stock faster. A seasonal SKU sitting at 90+ days might get an automatic 15% markdown trigger instead of sitting until a manual quarterly sale — recovering some margin instead of writing it off entirely at a deeper end-of-season discount.
- Protecting margin on low-competition items. If a model detects a SKU has no close competitor match and steady demand, it can hold or even nudge price up slightly rather than defaulting to the store-wide discount cadence — this is the flip side merchants often forget about.
- Reacting to cost shocks same-day. A supplier price increase or FX swing on imported stock can be reflected in the storefront price within hours instead of the multi-week lag of manual repricing, protecting margin that would otherwise silently erode.
None of this requires selling different prices to different people — the version most merchants should actually implement changes prices by SKU and time, not by customer. That distinction matters a lot for the compliance section below.
The EU rules that actually constrain this
This is where dynamic pricing gets misunderstood. There's a real legal difference between dynamic pricing (price varies by time/inventory/demand, same price shown to everyone at that moment) and personalised pricing (price varies by who's looking — browsing history, device, perceived willingness to pay).
- Consumer transparency. Under the Omnibus Directive (EU) 2019/2161, amending the Consumer Rights Directive, if a price was personalised based on automated decision-making, you must clearly inform the consumer that the price was personalised. This obligation doesn't apply to pure dynamic pricing based on market conditions, but the line between the two is exactly where merchants get careless — if your logic ever factors in an individual visitor's data (device type, past purchase willingness, loyalty tier used as a price lever), you've crossed into personalisation and the disclosure duty kicks in.
- Non-discrimination. Pricing logic can't function as a proxy for discrimination based on protected characteristics — nationality-based repricing via IP geolocation is a recurring compliance trap, and it can also conflict with the EU's geo-blocking rules for cross-border sales within the single market.
- GDPR. Any pricing model that uses individual browsing behaviour or purchase history as an input is processing personal data, which means a lawful basis, a privacy-notice entry, and — if the pricing decision meaningfully affects the consumer with no human involved — a look at the Article 22 automated-decision-making provisions.
- AI Act. Regulation (EU) 2024/1689 explicitly prohibits AI practices that exploit vulnerabilities or use manipulative techniques that materially distort a consumer's decision-making in a way that causes harm. A pricing model that detects and exploits, say, urgency or low price-sensitivity signals to systematically overcharge specific individuals sits uncomfortably close to that line, even if it isn't formally captured elsewhere.
None of this bans dynamic pricing — it's routine and legal across EU retail. It bans doing it opaquely and doing it to individuals rather than to the market. For the fuller compliance map, our <a href="/blog/eu-ai-act-online-stores-impact/">guide to the EU AI Act's impact on online stores</a> and <a href="/blog/complete-eu-compliance-checklist-online-stores/">complete EU compliance checklist</a> cover the adjacent obligations you'll want on the same radar.
A pre-launch compliance and guardrail checklist
- [ ] Confirm your pricing logic varies by SKU/time/inventory, not by individual visitor — if it's the latter, build in the personalisation disclosure
- [ ] Set hard price floors and ceilings per SKU so the model can't undercut cost or spike price beyond a defined band
- [ ] Exclude protected-characteristic proxies (geolocation-as-nationality-proxy in particular) from any input feature set
- [ ] Log every automated price change with the signals that drove it, retained long enough to answer a regulator or customer query
- [ ] Review model outputs weekly for the first two months — automated pricing left fully unsupervised is how a scraper bug turns into a €0.01 bestseller
- [ ] Keep a documented, plain-language explanation of the pricing logic for your privacy notice and internal compliance file
Getting started on PrestaShop
You don't need a data-science team for a first version. A practical rollout: pick 20-50 SKUs with reliable sales history and at least one identifiable competitor, connect a price-intelligence feed or a competitor-scraping module, set floor/ceiling guardrails, and let a rule-based engine run for a full sales cycle before considering an ML layer. Measure sell-through and margin against a control group of SKUs left on static pricing — without a control, you won't actually know if the engine is helping.
Testing dynamic pricing safely
Before evaluating any dynamic-pricing module, export 12 months of order data for your top 50 SKUs by revenue and check how many have at least one identifiable online competitor with trackable pricing. That's your addressable set for a first pilot — everything else isn't ready for automated repricing yet.
Frequently asked questions
Is dynamic pricing legal in the EU?
Yes. Adjusting prices based on demand, inventory, or competitor moves is standard retail practice and not restricted by EU law. What's regulated is personalised pricing based on individual consumer data, which requires clear disclosure under the Omnibus Directive (EU) 2019/2161.
What's the difference between dynamic pricing and personalised pricing?
Dynamic pricing changes the price shown to everyone based on market conditions like time, demand, or stock levels. Personalised pricing varies the price by individual visitor, using data like browsing history or device type — and it triggers a specific EU disclosure requirement that dynamic pricing alone does not.
Do I need machine learning to start, or can I use simple rules?
Start with rules. Most stores get 70-80% of the benefit from straightforward if/then logic on competitor price and stock age before a trained model is worth the data and engineering investment. Move to ML once you have clean historical data and proven ROI from the simpler approach.
Can AI dynamic pricing accidentally discriminate against customers?
Yes, if inputs like IP-based geolocation function as a proxy for nationality or region, or if the model learns to charge more to visitors who show urgency signals. Auditing your feature inputs and outputs regularly is essential, not optional.
How often should prices actually update?
It depends on the category — fast-moving or high-competition items might update daily or even hourly, while stable categories are fine on a weekly cadence. Update frequency should match how often the underlying signals (competitor prices, stock levels) actually change, not an arbitrary schedule.
What happens if the pricing model makes a mistake?
This is why hard price floors and ceilings aren't optional. Set guardrails so an anomalous input (a scraping error reading a competitor's price as €0) can't push your live price below cost, and review automated changes closely during the first few weeks of any new pricing engine.
Related reading
- Predictive Inventory Management Using AI
- How AI Will Transform PrestaShop Stores
- AI Search Is Changing SEO: What Merchants Need to Know
- Omnibus Directive Compliance Guide
- How the EU AI Act Will Affect Online Stores
