Anthony Savagar, Misa Tanaka and Jagdish Tripathy
With increased availability of big data and computing power, more firms are adopting algorithmic and AI-powered pricing to adjust prices rapidly in response to changing economic conditions over time and across consumers. This post reviews the existing research, draws implications for central banks, and identifies areas for further research on this topic. The research reviewed here was also used to inform Lombardelli and Patel (2026). The existing research suggests that new pricing technologies will lead to faster pass-through of shocks to prices, greater market segmentation, and may improve the inflation-output trade-off for monetary policy makers. To ensure price stability, central banks will need to monitor granular, high-frequency price data to gauge the impact of shocks on prices and inflation expectations.
Have prices become more flexible?
Improvements in pricing technology, such as electronic shelf labels and real-time algorithmic pricing, reduce the cost of changing prices and increase the frequency of price adjustments, thus making prices more flexible. The average length of time retail prices are fixed in the US has roughly halved over the past decade (Cavallo (2019)). Online prices change more often than offline prices (Gorodnichenko et al (2018)), suggesting that overall prices faced by consumers could become more flexible as more transactions move online.
The lower cost of changing prices may speed up pass‑through of shocks to aggregate price levels. Using more than 20 million prices for multiple online sellers, Gorodnichenko and Talavera (2017) and Cavallo (2019) report stronger pass-through and faster convergence of prices to new equilibrium levels in response to exchange rate and gas price shocks. However, the extent of price stickiness and pass-through varies by item (eg brand loyalty), sector (eg firm-entry costs) and the market (eg degree of competition) (Gorodnichenko and Talavera (2017)). So central banks will need to monitor granular, high-frequency data to understand the speed of pass-through across different segments of the economy.
A micro lens: do algorithms raise prices or just disperse them?
Historically, dynamic pricing – whereby firms adjust prices over time in response to changing economic conditions – has been used to manage capacity through price discrimination. Airlines, for example, use dynamic pricing to reallocate demand across time (Puller and Taylor (2012)), while ticket sellers extract surplus through timing discounts rather than increasing mark‑ups (Sweeting (2012)).
The impact of algorithmic pricing – whereby firms use data-driven, rule-based processes to adjust prices – on retail prices is mixed. The possibility that algorithms interact to raise prices has been shown in simulated marketplaces (Calvano et al (2020)), but there is limited real-world evidence on this (Schwalbe (2019)). Assad et al (2024) find that algorithmic pricing increases margins by 15% in a cross-country study of the retail gasoline sector. By contrast, Brown and MacKay (2023) report that drug retailers charge lower prices when algorithms respond rapidly to competitors’ prices. Overall, existing research is inconclusive as to whether algorithmic pricing increases prices.
Algorithmic and AI-based pricing can be used not only to adjust prices across time, but also across consumers, for example by enhancing firms’ ability to personalise prices based on consumers’ characteristics. This may lead to higher price dispersion as individuals with high willingness-to-pay subsidise those with lower willingness-to-pay. There is established evidence that US retailers adjust prices in response to local demand conditions (Stroebel and Vavra (2019)). Although the extent to which pricing technology is currently used to target demand at a highly granular level remains unclear, it is likely to result in a wider array of prices faced by consumers, increasing price dispersion.
A macro lens: what happens to inflation?
If more flexible micro prices translate to more flexible aggregate price levels, then inflation will respond more strongly to real economic conditions. In a standard framework, less price stickiness yields a steeper Phillips curve, implying that central banks can lower inflation with a smaller sacrifice in terms of unemployment or output. An inflation-accelerator mechanism could also amplify inflation if firms raise markups more aggressively when inflation is already high. In Blanco et al (2024)’s framework, a self-fulfilling cycle occurs as the fraction of price changes increases with inflation, leading to more price increases. The consequence is again a steeper Phillips curve in high-inflation periods.
Market features, such as the extent of competition and returns to scale (whether a firm’s production gets more efficient with its size), also influence monetary policy transmission. Further research should examine how algorithmic pricing shapes competition and firm cost structures, which will affect aggregate price markups. For example, access to customer data may serve as a barrier to entry, strengthening the market power of incumbent firms, which is perhaps already on the rise in the US (De Loecker et al (2020)) and the UK (Savagar et al (2024)). Greater market power enables firms to price further above cost, raising the price level. Conversely, the new pricing technologies could lower costs. For example, better pricing technology could minimise waste of perishables, improve inventory management, and so mitigate upward pressure on food prices resulting from shocks to energy prices. This mechanism could be further enhanced if increasing returns to scale lower costs for the largest firms. Thus, new pricing technologies may shake-up existing market structures, change the balance between incumbents and new innovators, and alter how shocks to costs translate to prices and inflation.
Will it affect inflation expectations?
Anchoring inflation expectations is central to monetary policy effectiveness. Firms’ pricing decisions play a key role in shaping consumers’ inflation expectations. In that context, it is notable that Cavallo et al (2017) find that consumers focus on retail prices rather than official inflation statistics, with food and other frequently purchased items appearing particularly important in shaping inflation expectations (Anesti et al (2025) and D’Acunto et al (2021)).
Further research is needed to examine how algorithmic pricing – which could increase the frequency of price changes and dispersion of prices – influences inflation expectations. For instance, algorithmic pricing could increase expected inflation volatility and thus could increase precautionary savings. Similarly, rapid pass-through of cost shocks could risk de-anchoring of inflation expectations. Reis (2022) emphasises that unanchored inflation expectations during periods of inflation can extend the lifespan of otherwise transitory shocks.
If we all pay different prices, what even is inflation?
Algorithmic pricing also complicates the measurement of inflation itself. When algorithms reprice products frequently, conventional CPI sampling (monthly, store-level) will understate the true frequency and variance of adjustment (Cavallo (2019); Leung et al (2023); Davies (2021)). Moreover, posted prices may differ substantially from transaction prices once discounts and personalised offers are taken into account, straining the concept of a ‘representative’ price (Lombardelli and Patel (2026)).
As a result, official inflation measures may diverge from consumers’ lived experience. Statistical agencies, including the Office for National Statistics, are responding to this challenge by using new data sources, such as groceries scanner data which allow for high frequency, broad-based and automated collection of prices that accurately reflect prices paid by consumers. Several central banks are also using web-scraped data to study heterogeneity in realised inflation (Messner and Rumler (2024)), nowcasting (Macias et al (2023)) and high-frequency pass-through (Gautier et al (2023)).
Conclusion
New pricing technologies increase frequency of price adjustment, with ambiguous effects on price levels. It enhances pass-through of shocks to prices and enables market segmentation, but it does not necessarily imply greater macro-volatility or a worsening trade-off for monetary policy makers.
Further research is needed to understand how changes in pricing technologies and strategies are shaping the macroeconomy, as well as inflation expectations. This includes construction of high-frequency, granular data sets to enable central banks to monitor the speed of pass-through of shocks, as well as their impact on inflation expectations. More work is also needed to examine how dynamic pricing, along with agentic AI and more personalised pricing, reshape competition across sectors and affect consumer welfare.
Anthony Savagar and Misa Tanaka work in the Bank’s Research Hub and Jagdish Tripathy works in the Bank’s Centre for Central Banking Studies.
If you want to get in touch, please email us at bankunderground@bankofengland.co.uk or leave a comment below.
Comments will only appear once approved by a moderator, and are only published where a full name is supplied. Bank Underground is a blog for Bank of England staff to share views that challenge – or support – prevailing policy orthodoxies. The views expressed here are those of the authors, and are not necessarily those of the Bank of England, or its policy committees.
Share the post “What machines taking over pricing means for central banks”
