Algorithmic pricing and competition: Discussion paper

June 10, 2025

On this page

• Introduction
• About the Competition Bureau
• Section 1: What algorithmic pricing is and how it is used
  • 1.1 What algorithmic pricing is
  • 1.2 How it works
  • 1.3 What type of data is being used
  • 1.4 Where the data comes from
  • Discussion questions
• Section 2: Potential impacts on markets and consumers
  • 2.1 Potential issues under the Competition Act
    • 2.1.1 Algorithmic price-fixing and competitor collaborations
    • 2.1.2 Practice of anti-competitive acts
    • 2.1.3 Deceptive marketing practices
  • 2.2 How it factors into competition analysis
    • 2.2.1 Competitive entry and expansion
    • 2.2.2 Personalized prices and consumer switching
    • 2.2.3 Innovation and market efficiency
  • Questions for discussion
• Conclusion
• Footnotes

Introduction

Algorithms and data analytics are increasingly used to influence business strategies, including to set and adjust prices. Advances in technology – including artificial intelligence – and greater access to data are driving this trend. In Canada, over 60 companies offer services that use algorithms and that claim to help companies optimize pricing.

Algorithmic pricing can be broadly defined as the process of using automated algorithms to set or recommend prices for products or services, often in real time, based on a set of data inputs. It is gaining momentum across sectors and industries worldwide, with reports of its use in sectors from hospitality^{Note de bas de page 1} to concert tickets^{Note de bas de page 2} to ridesharing^{Note de bas de page 3}.

Algorithmic pricing has also emerged as a topic of interest among competition authorities. Last year, both the United States Federal Trade Commission^{Note de bas de page 4} (US FTC) and the United Kingdom Competition and Markets Authority^{Note de bas de page 5} (UK CMA) launched studies into personalized pricing and dynamic pricing, respectively. The Competition Bureau is similarly interested in learning more about algorithmic pricing, including the practices of personalized and dynamic pricing, and its impact on competition in Canada.

The objective of this discussion paper is to build a robust understanding of algorithmic pricing to ensure that the Bureau is ready to respond swiftly and effectively to this emerging trend. At this stage, it is not intended to result in policy recommendations.

Section 1 explains what algorithmic pricing is, how it works and the different types of applications.

Section 2 explores the positive and negative impact of algorithmic pricing on competition.

We invite you to share your thoughts and experiences related to algorithmic pricing, including in response to the discussion questions we have outlined in each section. Your feedback is invaluable in helping us better understand this topic and its challenges for competition.

About the Competition Bureau

The Competition Bureau is an independent law enforcement agency that protects and promotes competition for the benefit of Canadian consumers and businesses.

We administer and enforce Canada's Competition Act, a law of general application that applies to every sector of the economy. We investigate and address abuses of market power, anti-competitive mergers, price-fixing, and deceptive marketing practices. The Bureau also advocates for pro-competitive laws and regulations.

Section 1: What algorithmic pricing is and how it is used

1.1 What algorithmic pricing is

Algorithmic pricing generally involves the following components:

• Automation: It runs on an algorithm, which is a set of instructions or steps that is designed to perform specific tasks without human intervention. Some algorithms are guided by a predefined set of rules; others are driven by artificial intelligence (AI) techniques. We explore this distinction later in this paper.
• Price optimization: The algorithm analyzes data and determines the optimal price for products or services to maximize profits. It does that by either directly setting the prices or suggesting them to a human decision-maker. It often performs this task in real time.^{Note de bas de page 6} In other words, prices can be adjusted quickly once the algorithm has the latest data available.
• Data as an input: Algorithms rely on various data sources to inform their pricing decisions. The data may be from or about consumers, such as demographic information, online behaviour, and transaction history. It may also be about market conditions, such as supply and demand conditions, competitor prices, and inventory levels. An algorithm may use internal or external information.

Pricing algorithms are just one type of algorithms that businesses may use for marketing their products or services. For example, some algorithms might monitor the market to follow competitors’ decisions such as price changes; others might recommend certain products based on data about a user or make those products stand out more on an online platform.^{Note de bas de page 7}

1.2 How it works

There are generally at least two ways in which algorithms might work to set or recommend prices.

The first category of algorithms is rules based. These algorithms make their pricing recommendation based on a set of predetermined parameters or conditions, i.e. the rules. If a rule is triggered, the change in price is automatic. The algorithm does not learn from the data it collects over time as it adjusts prices; the rules will only be changed by human action.

The second category is AI-driven. These algorithms tend to be more advanced. They leverage AI, in particular machine learning algorithms, to set or recommend prices in real time.

The main difference from rules-based algorithms is the ability for AI-driven algorithms to learn. As these algorithms set or recommend prices, they collect data on how the market reacts to these prices. That is, the algorithm learns from and continuously updates the data used to set or recommend prices (a process known as reinforcement learning). As a result, the decision-making process of these algorithms often lacks transparency or can be difficult to understand (a problem known as the black box issue). Human intervention may be used to improve these algorithms, but it is not necessary. The algorithm’s use of real-time and historical data and reinforcement learning allows it to set or recommend pricing without human help.

1.3 What type of data is being used

In this paper, we also make a distinction between pricing algorithms based on the type of data they use—data on market conditions, or consumer data.

Some algorithms recommend or automatically set a price based on market conditions; these are known as dynamic pricing algorithms. Examples of market conditions include supply and demand, other firms’ prices, inventory levels.^{Note de bas de page 8} Factors such as time of day, season, weather and local events are also monitored since they may have an impact on supply and demand. In contrast, personalized pricing algorithms tailor prices to individuals or groups of individuals based on various characteristics. These may be found through consumer data about demographics, online behaviour, and transaction history.

In practice, it can be difficult to make a distinction between dynamic and personalized pricing algorithms, as firms might be feeding their pricing algorithms both types of data.^{Note de bas de page 9}

Dynamic pricing may allow companies to maximize their profits by responding swiftly to market changes, while personalized pricing (also known as surveillance pricing) may allow companies to target as closely as possible consumers’ willingness to pay to maximize their profits.

1.4 Where the data comes from

The data analyzed by algorithms comes from various sources; it does not necessarily come from the developer of the pricing algorithm. This paper distinguishes between three types of data sources:

• Publicly available information: Algorithms might monitor different source of publicly available information such as competitors’ online prices and promotional activities to inform strategic decision making.
• Internal information systems: This is data that comes from the business itself, such as inventory levels, sales, promotions. Businesses may also gather data on their consumers, such as their online behavior, transactions and locations. This data is then used to feed the algorithm and set personalized prices for their customers. Firms sometimes incentivize their customers to share personal data through loyalty and membership programs. They might also collect information as a condition of service.

• External information providers: Firms can also buy data from businesses offering data or analytic services. Businesses sometimes use third-party services that pool industry data to provide algorithmic pricing services. In this instance, the third party could sell its algorithm to many companies, often packaged as a complete, ready-to-use software. Each company using the software feeds data into the algorithm. A third-party supplier could also sell the data to develop and train other firms’ in-house pricing algorithms.

Discussion questions

• Are there other components that the Bureau should consider in defining algorithmic pricing?
• How prevalent are pricing algorithms in Canada? Are AI-driven algorithms more commonly used than rules-based algorithms?
• What are the benefits to using personalized or dynamic pricing from an economic perspective?
• What are the data sources for algorithmic pricing? How do pricing algorithms use this data to set or recommend prices?

Section 2: Potential impacts on markets and consumers

Algorithmic pricing can both promote and hurt competition. It may increase competition by helping businesses innovate and be more efficient. It may lower barriers for new businesses to enter the market with innovative pricing strategies. It may also make it easier for consumers to switch.

But it can also limit market entry and facilitate companies coordinating their prices. It may limit market supply and result in a higher market power for companies. Pricing algorithms may also limit consumer choices and reduce helpful information for making decisions. As personalized prices become more common with the use of algorithmic pricing, it may become harder for consumers to gather pricing information through social interactions.

The impact of algorithmic pricing depends on how it is used and its specific features, including whether it is personalized or dynamic pricing, the type of data used, who owns the data used and the market characteristics.

2.1 Potential issues under the Competition Act

2.1.1 Algorithmic price-fixing and competitor collaborations

Competitor collaborations refer to agreements or arrangements between or among competitors or potential competitors.^{Note de bas de page 12} These collaborations may be reviewed by the Bureau under the criminal or civil provisions of the Competition Act. For example, competitor collaborations involving price-fixing, market allocation or output restriction are generally reviewed under the criminal provisions, while commercialization or joint-selling agreements, or other types of competitor collaborations that may lessen or prevent competition are generally reviewed under the civil provisions.^{Note de bas de page 13} Pricing algorithms may facilitate both explicit and tacit forms of competitor collaborations, including “hub-and-spoke” agreements.

For instance, pricing algorithms may help coordination between firms by using a common algorithm. Two or more competitors may use an algorithm that processes their data and dynamically sets prices to earn the highest combined profits for all firms. A pricing algorithm may also help firms detect deviations from the coordinated strategy and respond to competitors lowering their prices by lowering their own prices.^{Note de bas de page 14}

Third-party algorithm suppliers may also facilitate “hub-and-spoke” agreements by acting as the “hub” that sends shared signals to competing firms (the “spokes”). The hub may also encourage firms to follow the suggested price. If the spokes communicate directly to coordinate, this forms a “rim” which indicates an explicit agreement between competitors. Alternatively, hub-and-spoke scenarios could also lead to tacit agreements^{Note de bas de page 15} (rimless) among competitors without explicit communication between them.^{Note de bas de page 16} For example, multiple competitors may use the same pricing algorithm to get pricing recommendations. By pooling data from multiple competitors, the algorithm’s pricing recommendations could resemble collusive outcomes even if the competitors do not explicitly collaborate. Research shows that AI could help facilitate tacit coordination by autonomously (without human instruction) learning and implementing coordinated strategies.

The likelihood of pricing algorithms facilitating coordination depends on several market characteristics. For example, markets in which firms produce similar products and services have higher risks of coordination, as it is easier to find the price that maximizes their combined profits. Algorithmic coordination is also more likely in markets with high transparency due to better availability of data, and in markets with frequent interactions, as algorithms may enable firms to punish deviations more effectively.^{Note de bas de page 17} Coordination is also more easily sustained in markets with fewer firms and higher barriers to entry.^{Note de bas de page 18}

2.1.2 Practice of anti-competitive acts

Firms may engage in other types of conduct that harms competition. This section discusses how pricing algorithms may help firms leverage data on competitors, consumers, and the market to engage in harmful practices such as predatory pricing or tying and bundling strategies. Other types of conduct that are not discussed here may also raise similar concerns.

2.1.2.1 Predatory pricing

Predatory pricing is a strategy where a dominant firm takes on short-term losses by deliberately setting the prices of goods or services below cost to eliminate competitors or new entrants.^{Note de bas de page 19} This strategy typically has two stages: first, the dominant firm aggressively undercuts its competitors to force them out of the market (the predation phase), and then it uses its market power to increase prices to recover losses and make profits after competitors are gone (the recoupment phase). For predatory pricing strategies to work, the firm must keep prices low long enough to eliminate competitors.

Pricing algorithms may engage in price discrimination by helping firms target specific customers of their rivals with below-cost prices.^{Note de bas de page 20} For example, an established firm may do this to avoid losing business to a new competitor. The established firm could use an algorithm to target the customers that are most likely to switch, in an effort to retain them, instead of offering lower prices to all their customers. This could help the established firm minimize losses. These algorithms may also help firms stick to a predatory pricing strategy^{Note de bas de page 21} and establish a reputation that the firm will engage in future price cuts if challenged by entrants.^{Note de bas de page 22} Pricing algorithms may also help firms raise prices for consumers who have a higher willingness-to-pay or are less sensitive to price changes.^{Note de bas de page 23} They also enable firms to simultaneously engage in predation and recoupment.

2.1.2.2 Tying and bundling

Tying and bundling are practices used by firms to link the sale of one good or service to another.^{Note de bas de page 24} When faced with these practices, customers make choices by considering the value of the individual products (if available separately) versus the tied or bundled offer. Customers who are less sensitive to price changes are more likely to accept these offers. Pricing algorithms may help these practices by allowing firms to target higher prices for tied or bundled offers to less price-sensitive customers, while offering discounts to more price-sensitive customers to prevent them from switching to competitors.^{Note de bas de page 25} Pricing algorithms could also make it easier for a company to use its market power in one product to gain competitive advantage in the market for a second product by tying the two products together. They do so by targeting discounts to consumers who value the second product less.

2.1.3 Deceptive marketing practices

Firms need a lot of data to set prices using algorithms, which could raise issues of deceptive marketing practices.

It is illegal to make false or misleading representations to promote a product, a service or a business interest to Canadians. Companies that design digital products or services to collect consumer data for a commercial purpose^{Note de bas de page 26} are promoting their own business interests, even if they do not charge for their product or service. Representations must not mislead consumers into making decisions that they may not have made without the misleading or deceptive information.

Representations may also be false or misleading if they lead consumers to give companies access to data that they would not have otherwise provided, or acquire digital products or services they might not have otherwise selected. How a company collects, uses, handles and shares consumer data may influence a consumer’s decisions.

False or misleading representations that are most likely to raise issues are those about:

• Whether consumer data will be collected;
• What data will be collected;
• How often data will be collected;
• Why the data is collected and what it will be used for;
• Whether the data will be sold to, or otherwise shared with third parties; and
• Whether consumer data will be retained, and how it will be maintained and deleted,

False or misleading representations about how much control consumers have over their data will also raise concerns.

2.2 How it factors into competition analysis

2.2.1 Competitive entry and expansion

Algorithmic pricing may also affect the ease of entering a market, depending on the circumstances.

On one hand, pricing algorithms may need access to proprietary data and the computational power to process such data, especially for AI-based algorithms.^{Note de bas de page 27} Established companies have the resources to get this data and computational power, as well as to develop the AI technologies and train their employees to use them. New entrants often don’t have these resources, which could make it harder for them to enter markets and industries that are using algorithmic pricing.^{Note de bas de page 28}

On the other hand, pricing algorithms may help new businesses enter a market by letting them use innovative pricing strategies that better leverage consumer and competitor data. For example, a new firm might create a new pricing algorithm that uses market information to target specific groups of consumers who are more likely to switch over from existing firms.^{Note de bas de page 29}

2.2.2 Personalized prices and consumer switching

Pricing algorithms may influence whether consumers decide to stay with a firm or switch to a competitor. They do this by offering different prices to different consumer types: those who are willing to pay more than the offered price and are less likely to switch (inframarginal consumers), and those whose willingness to pay is closer to the market price and who are more likely to switch to a competitor (marginal consumers).

New or smaller firms may use pricing algorithms to target the marginal consumers of established firms by providing them different prices, creating an incentive for them to switch. Established firms could use the same technology to keep their customers by cross-subsidizing losses between consumer groups and/or markets. For example, a firm that engages in predatory pricing could offer below-cost pricing to one group of consumers to stop them from switching, while making up for these losses by charging higher prices to another group^{Note de bas de page 30}. Similarly, a firm might recover its losses from offering below-cost pricing to marginal consumers in one market by charging higher prices to inframarginal consumers in another market.

2.2.3 Innovation and market efficiency

With pricing algorithms, companies may adjust prices more quickly based on changes in supply and demand, and manage their inventories better.^{Note de bas de page 31} This could improve resource allocation and lower production costs. Algorithms may also encourage disruptive innovation by helping new firms enter the market with better pricing strategies, leading to new products and services and increased competitive pressure.^{Note de bas de page 32} The diversity of products and services may also help prevent collusion, since cost differences across product categories and firms would make it difficult for firms to calculate the price that would increase their combined profits.^{Note de bas de page 33}

As personalized prices become more common, it may make it harder for consumers to find good deals through social interactions. For example, consumers often learn about products and services from friends and family.^{Note de bas de page 34} Algorithms that set personalized prices could make it harder for consumers to share useful information about prices and deals because businesses may offer different prices to different consumers. Also, the speed at which the algorithms adjust prices may make price comparison more difficult. For example, an online retailer using a pricing algorithm offers a product for $10. The consumer decides to shop around and visits other websites to compare price options. If they come back to the first website, the price might no longer be $10—it could be higher.

Questions for discussion

• What are the competition concerns and procompetitive effects of using pricing algorithms?
• How does algorithmic pricing impact different consumer groups? Could it increase the vulnerability of some consumer groups?
• What are the challenges for competition law enforcement agencies when it comes to algorithmic pricing, especially with AI-driven algorithms?
• What role does AI play in reshaping algorithmic pricing strategies? How is it changing the competitive landscape?

Conclusion

The rise of algorithmic pricing has revolutionized the way businesses operate. By leveraging data and advanced algorithms, companies may set prices dynamically and personalize offers to meet individual consumer needs. While this innovation brings some benefits, it also poses challenges for competition. As we navigate this evolving landscape, it is important to balance the advantages of algorithmic pricing with the need to ensure the benefits of a competitive marketplace for Canadians.