Skip to main content
Research and analysis

Price personalisation: evidence from the UK online market: executive summary

Published 19 November 2026

Background

The growth of the digital economy, along with advances in data collection and the use of algorithms, has greatly expanded the capacity for companies to practice price personalisation.

Price personalisation is a pricing strategy in which the same seller charges different consumers different prices for the same good or service, with these differences driven by consumer characteristics and behaviours that signal their willingness to pay (Department for Business, Energy and Industrial Strategy; Organisation for Economic Co-operation and Development (OECD)).

Price personalisation is effectively a form of price discrimination that takes into account consumers’ individual characteristics and behaviours.

Price discrimination can take several forms, notably:

  • first-degree discrimination, which involves substantial variation in prices across individuals
  • third-degree discrimination, which applies different prices to distinct consumer groups[footnote 1]

In the past, sellers had limited ability to identify consumers’ willingness to pay and to set personalised prices. Today, this is increasingly possible, as companies can collect extensive data on:

  • consumer demographics
  • location
  • online browsing and purchase history
  • social media activity

Companies can now systematically test prices – for example, through A/B testing[footnote 2] – to uncover differences in consumers’ willingness to pay.

Despite this, and despite some media coverage and anecdotal evidence, convincing evidence of widespread price personalisation practices remains limited[footnote 3]. In particular, only a small number of studies have examined the prevalence of price personalisation practices, and those that have often found little or no evidence[footnote 4]. There are, however, some notable exceptions.

The US Federal Trade Commission conducted a study in 2025 on surveillance pricing. Interim results suggested that companies track consumer characteristics and behaviours – such as location, demographics, and even mouse movements – to inform pricing strategies.

An international study of an online dating app conducted in 2021 across 6 countries (New Zealand, the USA, the Netherlands, South Korea, India, and Brazil) found that older consumers were charged significantly more.

Participants aged 30 to 49 paid, on average, 65% more than those aged 18 to 29 for a monthly subscription. An investigation by Which? in the UK found similar practices. Individuals aged 30 to 49 were likely to pay 48% more (an average difference of £24) for a 12-month plus subscription, while those over 50 paid 47% more (£23.59).

Finally, a large-scale international study conducted in 2012 (covering the USA, Germany, Spain, South Korea, and Brazil), which analysed 200 sellers and 600 products, found some evidence of price personalisation based on the origin of the URL.

For certain product categories, users visiting a seller’s website via a discount aggregator were offered prices up to 23% lower than those visiting the same site directly[footnote 5].

Although price personalisation is often perceived as a negative practice, it does not necessarily harm consumers and can, in some cases, be beneficial. When personalised pricing is applied, some consumers may benefit from discounts, lower prices, or increased access to products[footnote 6].

Others may pay more, yet the overall impact could still be positive if the benefits to those paying less outweigh the losses to those paying more. Nevertheless, concerns may arise even in such cases if fears about online price personalisation discourage consumers from making online purchases, or if the consumers who ultimately pay higher prices are particularly vulnerable (Office of Fair Trading).

Research objectives

The aim of this study is to provide systematic and up-to-date evidence on the prevalence of price personalisation practices in the UK online business-to-consumer sectors. We explore the prevalence of price personalisation across a broad range of sectors, as well as in selected sectors in more depth.

A secondary objective is to explore whether price personalisation, where it occurs, is associated with specific consumer characteristics and vulnerabilities, for example, whether certain groups of consumers are more likely to be exposed to personalised pricing.

In particular, we focus on vulnerabilities or traits related to (i) financial difficulties, (ii) mental health challenges, and (iii) neurodivergence[footnote 7]. This focus is informed by prior research on harmful online choice architecture, which suggests that consumers with vulnerabilities are disproportionately affected by such practices[footnote 8]. We also examine how prices shown to consumers vary according to demographics, device types, browsers, and behaviours.

Establishing a formal or statistical relationship between price differences and consumer characteristics typically requires a large consumer sample, particularly when pricing strategies are driven by complex algorithms that require extensive data to be reverse engineered. Given the relatively limited sample in this study, the objective is not to establish a formal link, but rather to identify preliminary evidence that can inform and guide more systematic investigation in future research.

Methodology

Data collection involved gathering prices for the same products across different consumer profiles. We used 2 complementary methods:

  • a survey-based approach (phase 1), in which a large sample of participants recorded prices shown via direct links
  • a synthetic profile approach (phases 2 and 3), in which researchers used controlled profiles to record prices in settings where prices are only revealed after user interaction (for example, apps and multi-step journeys)

Data analysis involved assessing the degree of price variation and exploring whether observed differences were associated with consumer characteristics.

We used:

  • quantitative analysis to measure the extent of variation
  • descriptive analysis to examine patterns by profile characteristics
  • case studies combining descriptive and statistical analysis for selected products or sellers where we identified substantial and/or repeated price variation[footnote 9]

An important aspect of the research design was to control for the impact of dynamic pricing as far as possible. Dynamic pricing is a pricing strategy in which prices are adjusted in real time based on market demand and supply.

For the survey-based approach, data collection took place within a single day, minimising exposure to dynamic pricing or other factors that may cause prices to vary across days. In addition, the time at which each respondent completed the survey was recorded, including both start and finish times. This allows us to examine whether any observed price variation was a function of time, and therefore potentially attributable to dynamic pricing.

For the synthetic profile approach, data was collected across 4 days. Price variation was measured only within a given day. Any price differences observed across days were not treated as evidence of price personalisation. In addition, the research team was able to identify price changes across profiles in real time and communicate these to others involved in data collection.

This allowed us to assess whether a price change represented a time-related shift potentially attributable to dynamic pricing, or variation across profiles independent of timing. In cases where there was clear evidence that a price change was driven by timing, it was not recorded as an instance of price variation.

Finally, for platforms where dynamic pricing is more prevalent, such as taxi services, the data collection window was determined to be 2 hours.

Our approach was implemented in 3 phases:

  • phase 1 involved collecting data across a wide range of products and sectors through a consumer survey
  • phase 2 used a synthetic profile approach to collect data for a targeted set of sectors where data collection via a consumer survey is more challenging
  • phase 3 focused on the transport sector using synthetic profiles, motivated by the findings from phase 2

These 3 phases are further discussed in the next section.

Phase 1

In phase 1, a large number of survey participants (1,688 in total and around 120 per product) were asked to visit 100 webpages corresponding to 100 different products and record the prices displayed[footnote 9].

In addition to price information, we collected data on participants’ demographics, location, and the device and browser used, as well as responses to questions designed to infer financial vulnerability, mental health status, and neurodivergence. The survey was conducted on the same day by all participants to minimise the potential impact of dynamic pricing or other time-sensitive factors on the prices observed.

Phase 2

In phase 2, we created 20 artificial profiles with varying characteristics (age, gender, device and browser type used to access online products). We also simulated the 3 vulnerabilities that are the focus of our research by constructing digital footprints[footnote 10] consistent with each vulnerability.

In total, the 20 profiles were used to visit 35 products across 34 sellers[footnote 11]. Each product was visited 4 times (4 waves) over the course of a month, generating 4 sets of prices for each product and synthetic profile[footnote 12].

Phase 3

In phase 3, we conducted a deep dive into the transport sector, motivated by the high price variation identified in phase 2, which we sought to explore further.

Specifically, using the 20 synthetic profiles constructed in phase 2, we collected data across 3 transport platforms and 3 alternative routes over a 2-week period, with observations gathered on 3 to 5 occasions (days).

Phases 1, 2 and 3

In total, across all phases, we collected and analysed prices for 144 products from 87 online sellers[footnote 13] across multiple consumer profiles, as well as over time. To our knowledge, this is the most comprehensive study of price personalisation in the UK and internationally.

Major findings

While the majority of products within scope did not exhibit price variation – different survey participants or synthetic profiles recorded the same price for the same product – there is strong evidence that price variation is more prevalent in certain sectors.

Specifically, we observed low levels of price variation in the retail sector, compared with substantial variation in sectors such as transport, packaged holidays, hospitality, and ticket sales. A third of the platforms in these sectors demonstrated high price variability. The majority of the variation identified is unlikely to be driven by dynamic pricing, as our methodology largely controls for it. This evidence of price variation is interpreted as a strong indication of price personalisation.

In terms of the type of price personalisation, we found a strong indication of first-degree price personalisation in one case – a taxi service platform. In particular, for the 4 routes we examined for this seller (in phases 2 and 3), we found that nearly every consumer was charged a different price.

These differences averaged between 11% and 25%, meaning that, on average, each consumer faced a price that differed from those offered to others by 11% to 25%. While some of this variation could, in principle, be attributed to dynamic pricing – which may be more intensive in this sector – the frequency and magnitude of the differences are difficult to explain solely through such mechanisms.

In addition, the correlation between prices and consumer characteristics identified for this seller is also consistent with first-degree price personalisation.

The remaining instances of price variation align more closely with third-degree price personalisation, as price variation is more discrete and/or associated with a limited set of consumer characteristics.

In the cases where we found price variation, we explored the degree to which this can be explained by consumer – survey participant or synthetic profile – characteristics. In the majority of the cases, we were unable to provide a clear link between price variation and the consumer characteristics that were under examination in our study. However, there were a few notable exceptions.

For the taxi service platform examined in phases 2 and 3, we found strong associations between prices and gender, age, and device type. However, these relationships appear to depend on the route, as the direction of the effects of gender, age, and device on prices varies across routes.

Overall, this suggests that price personalisation is more complex than the simple application of a premium or discount to specific consumers, with prices also influenced by additional factors such as the route.

For the second taxi service platform (examined in phase 3 only), we also found an association between prices and device type, with prices being 9% higher for smartphone users compared to laptop users.

We found a strong correlation between prices and device type in 3 more cases (in the survey-based approach / phase 1).

  • in 2 of these (a train ticket booking platform and a hotel booking platform[footnote 14]), smartphone users observed lower prices (by 19% and 10%, respectively) compared with desktop or laptop users.
  • in one case (a bus ticket booking platform), smartphone users observed higher prices than desktop or laptop users (by 6%)

In total, we found 5 products that showed a correlation between price and device type – 2 taxi service platforms in phases 2 and 3, and 3 in phase 1. Device type may serve as a proxy for unobserved consumer characteristics (for example, age and income) that influence willingness to pay, which may help explain why prices vary by device.

In one case (a hotel booking platform), price variation was strongly associated with login status. Consumers who were logged in to their account on the seller’s website at the time they viewed the price observed prices that were 10% lower than those who were not logged in.

Similarly, there is some indication that gender may be associated with price variation, as part of the observed variability appears to be correlated with gender. However, further evidence – such as larger samples, more platforms, and an assessment of whether consistent premiums exist across genders – is required before any firm conclusions can be drawn.

We found some evidence that is consistent with a pricing strategy that combines dynamic pricing and price personalisation, with the degree of price personalisation appearing to vary depending on market conditions.

We found no evidence of price personalisation linked to consumer vulnerabilities within the scope of this study (financial difficulties, mental health challenges, and neurodivergence status).

However, there is some indication that the neurodivergent group in our sample experienced greater price variation – often in the form of premiums – than other groups. This result should be treated with caution, given the small sample size, and further evidence would be needed to support any robust conclusions.

The remaining instances of price variation could not be linked to any consumer traits or behaviours captured in the study, suggesting that they may be associated with characteristics or data points beyond the scope of this research.

An important limitation of this study is that we were only able to examine a limited set of consumer characteristics, while several other characteristics and behaviours – such as purchasing history and complex real-world online behaviours and journeys – may have been omitted or not adequately captured.

Our synthetic profiles created digital footprints for vulnerabilities that may not approximate the real behaviours of vulnerable consumers online. The relatively small sample size should also be considered when interpreting the results, particularly given the challenges of identifying strong association between prices and consumer characteristics, which typically require larger samples.

All in all, a major finding of this study is that certain sectors are more likely to offer different prices to different consumers; that is, personalised pricing appears to be more prevalent in specific sectors. Further research would provide valuable additional insights.

This could include studies that more closely reflect real consumer behaviour and journeys, as well as research incorporating purchasing histories or larger sample sizes to enable more robust quantitative analysis.

  1. Second-degree price discrimination, or menu pricing, occurs when all consumers are offered the same set of options and choose between them based on their preferences, such as selecting basic or premium services. ↩

  2. In an online pricing context, A/B testing typically works as follows:

    (i) Consumers are randomly split into different groups.

    (ii) Each group is shown a different version of a webpage, offer, or price (for example, Price A versus Price B).

    (iii) The seller then observes how each group behaves – such as whether they make a purchase, how quickly they buy, or how much they spend.

    By comparing outcomes across the groups, firms can infer how sensitive different consumers are to price changes and identify differences in willingness to pay. This information can then be used to refine pricing strategies, including personalised pricing. ↩

  3. Also, see the recent news report on Instacart allowing AI-driven pricing tests, as well as previous allegations that Uber charges higher prices for users with low phone battery levels. ↩

  4. For example, see a study of the online airline market and another on the German online retail sector. ↩

  5. The study also identified evidence of location-based price differences, mainly for digital products such as e-books and video games, which may reflect variations in digital rights costs and competitive conditions, and not necessarily price personalisation. ↩

  6. It can enable products to be offered—or offered on specific terms—that might not be feasible without such price differentiation. ↩

  7. While these characteristics do not inherently constitute vulnerabilities, they may in some contexts be associated with increased susceptibility in market interactions. ↩

  8. For example, see a report recently commissioned by Citizens Advice. ↩

  9. This was facilitated by providing direct links within the survey and asking participants to click on each link, observe the price, and then return to the survey platform to record it. Each participant was asked to visit 7 links – not all participants had to visit 100 links. The links were randomised across participants.

    Because of this randomisation, the number of visits per link, and therefore the number of prices collected, varied across links. On average, each link was visited by 119 participants, with the minimum number of visits being 88 and the maximum 152. ↩ ↩2

  10. This involved searching for specific keywords (for example, debt help), visiting particular websites (such as depression support groups), and sending emails and posting on social media using keywords linked to the 3 potential vulnerabilities. ↩

  11. For one of the products, a dating platform, we used a different approach – see the methodology section for further details. ↩

  12. The profiles were operated by Alma researchers across multiple devices, including laptops and smartphones, with a virtual private network (VPN) configured to London. This approach allowed us to control for location while preventing platforms from assuming that all devices originated from the same place (our London office). ↩

  13. For some sellers, we collected prices for more than one product, and there was some overlap between the sellers included in our survey-based and synthetic profile approaches. ↩

  14. This is the same seller where we observed that prices vary depending on login status, although this refers to a different product (for some sellers, we explored more than one product). ↩