Demand-based price fluctuations are nothing new: even a century ago cinemas charged less for matinee screenings than for more popular evening shows, and room rates at a beach resort naturally varied by season. The practice however has now become an exacting science, and algorithmic dynamic pricing is transforming transportation, E-commerce, entertainment, and a wide range of other industries.
The concept is straightforward: have prices automatically fluctuate based on data about supply, historical demands, and competition in order to quickly adapt to changes in the marketplace and improve profitability. As algorithms become more powerful and more data becomes available, companies’ product and service prices can automatically respond to demand and competition in real time.
In this report, we will look at the application of algorithmic dynamic pricing through case studies of ride-hailing startups Uber and Lyft and E-commerce giant Amazon. We will also discuss potential issues involved in personalized dynamic pricing.
Uber and Lyft Real-Time Dynamic Pricing
The dynamic pricing strategy contributes to the growing revenue of the ride-hailing companies. At the same time, it also raises the risk of losing customers’ trust in the system.
Today’s tech-driven dynamic pricing has its roots in the transportation industry, where American Airlines is considered a pioneer of the strategy. Forty years ago the airline introduced low “super-saver” fares based on projected passenger demand and seat availability. The scheme was quickly adopted by most other airline companies.
Now ride-hailing companies like Uber and Lyft are evolving the concept by leveraging their massive data in real time. Users see different prices in different areas of a city and at different times of day, dynamically adjusted based on real-time data on driver supply and predictions on customer location, traffic, weather and so on.
Lyft’s prime time pricing
Uber’s surge pricing
In essence, both Uber’s “surge pricing” and Lyft’s “prime time” mechanisms are no different than what American Airlines did decades earlier. All are based on the classical economic model of supply and demand. During the workday rush hours or on Saturday nights, when demand for ride hailing services in specific areas rises, the price will be pushed high enough to discourage some customers, who can either wait for the price to go down or take public transit. As a result, demand is reduced. Correspondingly, the higher price will also attract more drivers to the area to serve the remaining demand, which establishes a new equilibrium. In this way the platforms can proactively affect both supply and demand to maximize their service efficiency.
The ride-hailing platforms’ AI-driven real-time dynamic pricing systems for example will respond to increased demand for cars around a venue where a concert or movie has just finished, and push the price up. A half hour later, as the higher price has attracted more drivers to the venue and the excessive demand has been met, the price will revert to a normal level. Such real-time adjustments rely on the massive traffic data on these platforms and their powerful algorithms’ ability to predict future demand. Without such real-time adjustments, prices would not reflect market conditions quickly enough for the process to be effective.
Uber and Lyft’s dynamic pricing schemes however still face several challenges. For example, they tend to divide a city into blocks and set the price adjustments accordingly. Some savvy passengers reduce the fare by simply walking to an adjacent block to book their ride. In addition, the system does not have a hard price cap. Last December an Uber passenger in Toronto was charged a whopping surge price of CDN$18,518 for a 6-km drive. Under social media pressure Uber later cancelled the charge, but the incident damaged public trust in the company.
Amazon: Dynamic Pricing Spanning Online to Brick-and-Mortar
Amazon is well-known for its practice of dynamic pricing, which is applied on the platform itself, by third party sellers on Amazon Marketplace, and in the new Amazon Go retail stores.
As the largest E-commerce website in the US and Europe, Amazon manages a huge volume of online transactions and is constantly monitoring the competition’s prices on hundreds of millions of products. Amazon thus self-generates the critical data required to implement its dynamic pricing. A study showed that the price of 100 random products on Amazon fluctuated by up to 260 percent over the course of a year. For example, on the UK site the same paddle board was offered at prices ranging from £235 — £699.
In additional to the scale of change, the frequency of change is also impressive: on average, a product’s price changed every five days, while certain products’ prices changed up to 300 times in the year. Many of the price changes were automated and determined based on products’ sales predictions based on Amazon’s massive amount of shopping data and powerful machine learning algorithms. Like Uber and Lyft, Amazon aims to optimize sales volume and price based on its supply, demand, and competition data. In 2016, Amazon quietly ended its pricing-matching policy, freeing it from the promise of offering the lowest price and shifting to a calculated price model based largely on customer behavior.
About 10 percent of third-party sellers in the Amazon Marketplace are also adopting AI-powered automated algorithmic pricing to get an edge on the competition. A critical battlefield for algorithmic pricing is Amazon’s Buy Box, the product page section where Amazon showcases best sellers’ offers. Research shows that sellers using algorithmic pricing are more frequently featured in the Buy Box, which will gives them higher exposure, more orders, and more user reviews. As Amazon evaluates its third-party sellers based on sales volume and customer feedback, these algorithmic sellers are more likely to be selected for the Buy Box and their advantage will amplify over time. Dynamic pricing can thus produce a “winner-take-all” scenario in certain product categories.
With the help of electronic labelling, dynamic pricing can also be expanded from online retailers to physical stores. Kaufland, a supermarket group in Germany, has already adopted digital labelling technology. This not only eliminates the human cost of printing and revising physical price tags for example during promotions, but can also enable instantaneous price changes to hundreds of different products in thousands of stores. In the near future, it’s likely we’ll see dynamic pricing in brick-and-mortar Amazon Go stores.
Amazon Go shop
Potential Issues With Personalized Dynamic Pricing
A different and more controversial angle in dynamic pricing is setting different prices for different customers. Many major E-commerce companies prefer not to disclose whether they do this, or may do it discreetly, as the practice can be regarded as a form of discrimination pricing. It’s not unusual for example to see a different price for a travel package when visiting a booking website on your laptop versus the price displayed on a friend’s computer, or even on the booking company’s app on your smartphone.
Online retailers’ dynamic pricing systems build and respond to individual users’ pricing profiles, which can be based on their zip code, device type, the type of products they have browsed and ordered, and other data. Like a savvy car salesman, sellers endeavor to size up the customer to determine how much they can afford; it is thus natural to set a higher price for those who can be expected to pay it.
Back in 2000, Amazon was found to be charging different people different prices. The company apologized and promised it would not set prices based on customer demographics. However even if pricing remains consistent on a specific product, there are other variations that can create personalized pricing scenarios:
Create different pricing tiers tailored to different customers
Customize product bundles based on users’ pricing profiles
Target different customers with different product suggestions at different prices, etc.
Such subtle adjustments in presentation and pricing, augmented by algorithmic dynamic pricing, can also result in some shoppers paying more than others, which benefits the seller.
The expansion of sales platforms using dynamic pricing is making it increasingly difficult to detect potentially unfair pricing schemes or protect oneself against them. Consumers would do well to apply the age old “caveat emptor” not only to the product they are purchasing but also the price tag it carries — even if one’s wits alone will not unravel the mystery and complexities of algorithmic dynamic pricing.