Dynamic Pricing
Date: 2026-08-16
Changing price by time, demand or stock. It works, and it’s normal in travel and marketplaces. Changing price by person is a different thing wearing the same name — and that one carries trust, transparency and data-protection exposure that usually outweighs the margin.
What it is
Dynamic pricing varies price automatically according to conditions rather than holding it fixed until someone changes it.
The critical distinction, because the two get conflated constantly:
| Varies by | Status | |
|---|---|---|
| Dynamic pricing | Time, demand, stock, competitor price | Routine, expected in some categories |
| Personalised pricing | Attributes of the individual shopper | Legally and reputationally loaded |
Everyone accepts a flight costing more at Christmas. Nobody accepts paying more than the next person for the same item at the same moment.
The legitimate versions
- Demand-based. Price rises with scarcity — the travel and events model. Transparent because everyone sees the same price at the same time
- Stock-based. Discount slow lines, protect price on constrained ones. This is just markdown management run automatically — Stockouts and Availability
- Competitor repricing. Track a defined set of competitors and hold a position. Standard in marketplaces, and effectively mandatory on some — Marketplaces
- Time-based. Scheduled promotional pricing. Old, uncontroversial, and the thing most “dynamic pricing” projects actually turn out to be — Promotional Cadence
Repricing has a floor problem
Automated competitor repricing pursues a rule you must define, and the default rule is a race:
rule: match the lowest competitor
competitor drops to £45
you drop to £45
they reprice against you → £44
you follow → £44
↓
both at cost, neither gaining share
Set a hard margin floor before any repricing goes live, and accept losing the sale below it. Also cap the frequency and size of moves — high-frequency repricing trains competitors’ algorithms and destabilises the whole category — Contribution Margin.
Why personalised pricing is different
Charging different people different prices for the same item, based on who they are, runs into four things at once:
- Trust. It’s the version customers find objectionable when discovered, and discovery is a matter of when. A price that changes between devices or between a logged-in and logged-out view is trivially noticed and screenshots well
- Data protection. Setting a price from a profile is a decision made about an individual using personal data — which is UK GDPR territory covering lawful basis, transparency and, where it’s fully automated with significant effect, additional restrictions — Consent Management
- Consumer law. Requirements exist around disclosing when a price has been personalised using automated decision-making
- It leaks into acquisition. New-customer-only pricing is personalised pricing with a friendly name, and existing customers find it
Note the asymmetry: discounts to identified segments — first order, loyalty tier, win-back — are widely accepted, because a discount off a published price is a promotion. Surcharges to segments are not. The published price is the ceiling in practice.
[CHECK: current ICO and CMA positions on personalised pricing and automated decision-making before implementing anything that sets price from a customer profile. This is an area where guidance has moved.]
What to do instead
Most of the margin dynamic pricing promises is available from structures customers can see and choose:
- Tiered Pricing — self-selection captures the willingness-to-pay spread without differential pricing
- Promotional segments — target who receives an offer, not what the price is
- Bundling — different effective prices via different combinations
- Time-limited public promotions — everyone eligible, same price
- Loyalty pricing — clearly earned, clearly disclosed, and read as reward rather than discrimination
If you do it
- Publish one price per item per moment, visible to everyone
- Floor on contribution, not on cost or on competitor price
- Rate-limit changes. Customers who watch a product’s price bounce daily wait rather than buy
- Never change price during a session. A price that moves between the product page and the basket is the single most damaging version of this
- Log every price change with a timestamp. You need this for reference-price claims, for margin analysis, and for the complaint you’ll eventually answer — Annotation and Change Logs, Price Anchoring
- Watch contribution per session, not revenue. Repricing that wins share at a lower margin will show up as growth — Margin versus Volume