How to avoid dynamic pricing: what 25 years of evidence says works
Every guide to beating dynamic pricing gives the same advice: clear your cookies, open a private window, switch on a VPN and watch the price fall. Almost none of them cites a test. I went looking for the tests. This page compiles the 19 best documented cases of dynamic and personalised pricing between 2000 and 2025, each with a link to the study, the regulator's decision or the company's own statement, and then asks a narrow question of each one: what input moved the price?
The answer is awkward for a company that sells VPNs. Location matters a great deal in online pricing. Your IP address matters much less than the folklore says. Knowing the difference is what saves money.
What is dynamic pricing, and what is surveillance pricing?
Dynamic pricing is the practice of changing the price of the same product over time in response to demand, stock or the clock, so that everyone who looks at the same moment sees the same price. Airline fares, hotel rates and Uber surge work this way and have for years.
Personalised pricing is a different thing. Two people look at the same moment and see different prices, because the seller knows or guesses something about each of them. The US Federal Trade Commission calls this surveillance pricing, and its January 2025 staff findings describe the raw material on offer to retailers: precise location, demographics, browsing history, even mouse movements and the items left in an abandoned cart.
Most "how to avoid dynamic pricing" advice mixes the two up. Tricks aimed at your identity do nothing against a price that is rising for everyone, and patience does nothing against a price that was set for you.
Where has dynamic or personalised pricing actually been documented?
The table lists cases in date order. I included a case only if someone measured prices, a regulator ruled, or the company itself made a statement on the record. "What moved the price" is the input the source identified, which is the column that tells you what to do about it.
| Year | Sector | Who documented it | What was found | What moved the price | Source |
|---|---|---|---|---|---|
| 2000 | Retail (DVDs) | Amazon, after customers compared notes | The same DVDs were shown at different prices to different visitors. Amazon called it a random price test, ended it and refunded the buyers who paid more. | Random test group | Amazon statement |
| 2012 | Office supplies | Wall Street Journal | Staples.com showed the same stapler at $15.79 or $14.29 depending on the visitor's estimated location. The discount usually appeared when a rival store was within about 20 miles. | Location estimated from IP address | WSJ |
| 2014 | Hotels, car rental, retail | Northeastern University | 16 sites tested with the accounts and cookies of more than 300 real users. Personalisation on nine. Orbitz and Cheaptickets gave logged-in members lower prices on about 5% of hotels, $12 cheaper on average. Home Depot showed Android users higher prices on about 6% of products, by an average of $0.41. | Login status, device type, A/B test group | Hannak et al. |
| 2014 | Airlines | KU Leuven and Inria researchers | 25 airlines queried for three weeks with varied browser profiles. No evidence of systematic price discrimination. Fares were volatile for every profile. | Nothing personal found | Vissers et al. |
| 2014 | Car rental (EU) | European Commission | The Commission wrote to six rental companies about sites that rerouted customers to a dearer national site once their IP address was identified, or raised the price after they entered a country of residence. | IP address, declared country of residence | Commission press release |
| 2015 | Ride hailing | Northeastern University | 43 emulated Uber apps ran for four weeks in San Francisco and Manhattan. Surge multipliers updated on a five-minute clock inside fixed surge areas, and fewer than 10% of surges lasted more than 20 minutes. Riders could cut the surge 10 to 20% of the time by requesting from an adjacent area. | GPS pickup point, demand in the last five minutes | Chen et al. |
| 2015 | Test preparation | ProPublica | The Princeton Review quoted $6,600 to $8,400 for the same online SAT tutoring package depending on ZIP code. Customers in areas with many Asian residents were nearly twice as likely to get the higher quote. | ZIP code typed by the customer | ProPublica |
| 2015 | Theme parks (EU) | European Commission complaint, press-reported figures | The same Disneyland Paris premium package was reported at €1,346 for French buyers, €1,870 for British and €2,447 for German. The Commission asked French authorities to investigate. The park said the gaps came from national promotions. | Country of residence and national site | Irish Times |
| 2016 | Cross-border retail (EU) | European Commission mystery shop | More than 10,000 websites tested. Shoppers reached the final stage of a cross-border purchase on only 37% of them. The rest blocked at registration, delivery or payment. | Registration address, delivery country, payment card | Commission |
| 2018 | Retail and travel (EU) | European Commission market study | 160 websites in eight member states. 61% personalised the ranking of offers, 92% of airline ticket sites among them. Prices differed in only 6% of identical-product checks, with a median gap under 1.6%. | Little found on price. Ranking was personalised. | Commission study |
| 2019 to 2022 | Retail app | KARE 11, then California district attorneys | The Target app listed a Samsung TV at $499.99, then $599.99 once the phone was in the store car park. Target paid $5 million in March 2022 and may no longer change app prices by geofence. | Phone GPS location | KARE 11 |
| 2021 | PC games (EU) | European Commission | Valve and five publishers were fined €7.8 million for Steam activation keys that worked only in certain member states, which kept cheaper keys out of dearer markets. | Country of activation | Commission decision |
| 2021 to 2022 | Dating app | Mozilla and Consumers International | Tinder Plus was quoted at up to 31 different prices within one country. Across six countries, users aged 30 to 49 were charged 65.3% more on average than those aged 18 to 29. Dutch quotes ran from $4.45 to $25.95. | Age in the account profile, other factors undisclosed | Mozilla |
| 2024 | Streaming subscription | Google statement | Google confirmed it checks the signup country against where the account is used and asks mismatched members to update their billing country. Users who had signed up through low-price countries reported cancelled memberships. | Billing country checked against usage location | Android Authority |
| 2025 | Cross-sector (US) | Federal Trade Commission staff | The FTC sent orders to eight pricing intermediaries in July 2024. Documents from the six covered in the staff findings showed tools that can set individual prices from location, demographics, browsing and purchase history, mouse movements and abandoned carts. The study describes what is sold to retailers, not the prices shoppers paid. | Capability documented, not measured | FTC |
| 2025 | Groceries in store (US) | UC San Diego, UT Austin and Northwestern economists | More than 180 million price observations from 114 stores that installed electronic shelf labels between 2022 and 2024. No surge pricing appeared after the labels went in. | Nothing found | Sanders et al. |
| 2025 | Airlines (US) | Delta, in a letter to three senators | Delta said no fare product it has used, is testing or plans to use targets customers with individualised prices based on personal data. Its new pricing software recommends fares to analysts from market data. | Company statement: demand only | Delta |
| 2025 | Concert tickets (UK) | Competition and Markets Authority | The CMA found no evidence that Ticketmaster used real-time algorithmic pricing in the Oasis sale. Standing tickets came in two price tiers that were not disclosed to the queue, and some "platinum" seats cost almost 2.5 times the standard price with no extra benefit. Ticketmaster signed undertakings on 25 September 2025. | Sell-out of the cheaper tier | CMA |
| 2025 | Grocery delivery (US) | Consumer Reports and Groundwork Collaborative | 437 volunteers shopped the same Instacart stores at the same time. About three quarters of items appeared at more than one price, up to 23% apart. One Seattle Safeway basket cost $114.34 for some shoppers and $123.93 for others. Instacart stopped the tests after publication. | Random test group | Groundwork report |
Read the right-hand columns and count. Location moved the price in nine of the 19 cases, which is more than any other input. But only two of those nine used the IP address: Staples in 2012 and the car rental sites in 2014. The other seven took location from somewhere a VPN never touches, such as the phone's GPS, a ZIP code the customer typed, or the country of a billing address, delivery address or game activation. Two cases priced on account data or device. Two were random experiments in which your identity played no part at all. Four investigations looked hard for personalised prices and found little or nothing, and the last two entries are a regulator describing what the software can do and an airline denying it does it.
The surprise for me was how small the measured effects of the classic suspects are. The device penalty at Home Depot was 41 cents. The cookie effect in the EU's 160-site study had a median under 1.6%. The large gaps in the table, 65% on Tinder and 23% on Instacart, came from an age field in a profile and from a coin flip.
What are the best ways to avoid dynamic pricing?
Each input in the first table has its own countermeasure. This second table turns the evidence round: for each thing a seller can price on, what the record shows and what changes the number you see.
| Price input | Evidence above | What changes the price you see | Does a VPN help? |
|---|---|---|---|
| Country or region inferred from IP address | Staples 2012, EU car rental 2014 | Viewing the site from another region, or opening another country's version of it directly | Yes. This is the one input a VPN changes. |
| GPS location inside an app | Uber 2015, Target 2019 | Denying the app location permission, using the website instead, or for ride hailing moving the pickup pin to the next surge area | No. GPS does not pass through the tunnel. |
| Billing, delivery or activation country | EU mystery shop 2016, Valve 2021, YouTube Premium 2024 | Only a real address and payment method in that country. Inside the EU, the Geo-blocking Regulation gives you the right to buy from any member state's site. | No, and faking it breaches the terms of service. |
| Login status and account profile | Orbitz and Cheaptickets 2014, Tinder 2021 | Comparing the price logged in and logged out. Members sometimes pay less, so check both ways. | No. |
| Device and browser | Home Depot and Travelocity 2014 | Checking a second device. Travelocity told the researchers it gave mobile users discounts, so the phone can be the cheaper one. | No. |
| Cookies and search history | Airlines 2014 and EU 2018, both close to nothing | A private window is free, so use one, but the measured effect is small | No. |
| Demand and time | Uber 2015, Oasis 2025, airline fares | Waiting, booking at another hour, setting a price alert. Most Uber surges in the study were gone within 10 minutes. | No. |
| Random price tests | Amazon 2000, Expedia and Hotels.com 2014, Instacart 2025 | Comparing with a second person or account at the same moment. It is the only way to see that a test is running. | No. |
So the working routine is short. Before a purchase that matters, look at the price three ways: logged in on your phone, logged out in a private window on a laptop, and from another country's version of the site. If the three numbers match, the price is not personal and only timing will move it. If they differ, you have found the input, and you buy through the cheapest route you are entitled to use.
Can a VPN help you avoid dynamic pricing or price discrimination?
It helps with one row of that table, and the row is a real one. Sellers that run separate national sites often pick yours from your IP address, and travel is where this shows most. The 2014 car rental case is the cleanest record of it: the European Commission wrote to six companies because customers were being rerouted to a dearer national site by IP address. Flights, hotels and rentals are still sold through country storefronts with their own prices and currencies, and a VPN is the simple way to see what those storefronts show. Our page on using a VPN for online shopping covers the mechanics, and the IP test page shows what location a site currently infers from your connection.
Three limits apply. First, the comparison is free but the purchase has rules. Subscriptions priced by region are tied to a billing country, and the YouTube Premium row shows what happens to accounts that fake one. Second, inside the EU you often do not need a workaround at all. Since 3 December 2018, Regulation 2018/302 has banned sites from redirecting you to your national version without consent and from refusing a sale because of your nationality or residence. Third, a lower foreign price can vanish at checkout once currency conversion, card fees and local taxes are added, so compare the final total and not the headline.
For everything else in the table, a VPN is the wrong tool. It will not calm an Uber surge, change the age in a dating profile or move you into the cheaper half of a price experiment. We would rather say that here than have you find it out at a checkout. What a VPN does do on every purchase is hide your IP address from the seller and encrypt the connection on hotel and airport Wi-Fi, which is a privacy reason to have it on while you shop, not a discount. For travellers, the companion guide on booking cheaper travel deals goes through the storefront comparison step by step, and the IP masking page explains what the seller sees.
Is dynamic pricing legal, and do sellers have to tell you?
Changing prices with demand is legal almost everywhere. Personalising them is legal in most places too, and the rules that exist are about disclosure.
The EU went first. Since 28 May 2022, the Consumer Rights Directive as amended by Directive 2019/2161 has required online traders to tell you when a price was personalised on the basis of automated decision-making. They do not have to say which data they used. New York followed with its Algorithmic Pricing Disclosure Act, in force since 10 November 2025 after a federal court rejected a retail industry challenge. Any price set from a shopper's personal data must carry the sentence "THIS PRICE WAS SET BY AN ALGORITHM USING YOUR PERSONAL DATA", and the attorney general can seek up to $1,000 per violation.
Disclosure laws have an obvious weakness, which is that someone has to look. Until they bite, the three-way price check above is the disclosure regime you run yourself.
What changes in 2026 and 2027?
Four open files will decide whether this page needs rewriting next year.
- California. AB 2564 would have banned retailers from setting a customised price from personal data gathered by electronic surveillance. It replaced AB 446, which its author parked in 2025 after amendments cut its scope. The Assembly passed AB 2564 on 27 May 2026 and the Senate passed an amended version 22 to 14 on 31 August 2026. The Assembly then did not take up the concurrence vote before the session ended at midnight, so the bill died. A third attempt in the 2027 session is the thing to watch.
- European Union. The Commission's Digital Fairness Act, which is expected to cover personalised pricing and the labelling of dynamic prices, is scheduled for the fourth quarter of 2026 according to the Parliament's legislative tracker. No text had been published by mid-September 2026.
- Instacart. The company ended its price tests in December 2025, and the New York attorney general demanded answers about them on 8 January 2026. Whether any regulator treats a random price test as a deceptive practice is the open question.
- Airlines. Delta's written denial of personal-data pricing is on the record with three senators. The first independent measurement of fares set with the new generation of pricing software has yet to be published, and it will be the next row in the table above.
Until then the evidence says this much. Prices online do vary, location is the input that varies them most, and the location that counts is usually the one on your card or your phone's GPS. Check the price three ways, buy through the route you are entitled to, and treat anyone promising that one app makes dynamic pricing disappear as you would treat a platinum ticket.
About the author
Le VPN Blog Editor
Alan Summers has been writing and editing for the Le VPN blog for years, covering online privacy, cyber security, and the best ways to get the most out of a VPN. He keeps a close eye on the news that affects internet freedom around the world and turns it into practical advice for Le VPN readers.
Articles by Alan Summers →
