The Pitfalls of Pricing Algorithms
More and more companies are relying on pricing algorithms to maximize profits. The use of artificial intelligence and machine learning enables real-time price adjustments based on supply and demand, competitors’ activities, delivery schedules, and so forth. But constant price shifts have a downside: They may trigger unfavorable perceptions of a firm’s offerings and its brand.
It’s vital, therefore, to understand and manage the signals being sent by the algorithms. The authors offer real-world examples of companies that have succeeded in this endeavor and others that have not. And they recommend four steps to avoid harm: Determine an appropriate use case for algorithmic pricing and explain its benefits to customers; designate an owner to supervise and be accountable for the system; set and monitor guardrails, both to protect against wild surges and to learn how price changes affect all aspects of the organization; and override the algorithms when necessary.
Many companies use algorithms to set prices and adjust them in real time so as to maximize profits. But constant price changes can alienate customers, undermine their loyalty, and damage brand reputation.
Pricing algorithms rely on artificial intelligence and machine learning to weigh variables such as supply and demand, competitor pricing, and delivery time. However, they often fail to consider the ways that frequent price changes affect customers psychologically, making them question the motives of companies and the value of their products and services.
To better control what dynamic pricing says to customers and how it impacts customer relations, firms should develop a proper use case and narrative for implementing algorithms, assign an owner to manage them, set and monitor pricing guardrails, and act quickly to override the automation when necessary.
On June 3, 2017, blue lights flashed toward London Bridge as police cars responded to reports of a terrorist attack. They blazed past thousands of people who were enjoying a Saturday night at restaurants and pubs in the area. Many of those who were out on the streets, sensing danger, attempted to order an Uber and head home to safety. But for 43 minutes after the first emergency call came in at 10:07 PM, Uber’s dynamic pricing algorithm caused rates in that part of the city to jump more than 200%.
The London episode is just one of many troubling examples of Uber’s price surges during moments of collective anxiety. Similar spikes occurred during a 2016 bombing in New York City, a 2017 taxi drivers’ strike in protest of U.S. anti-immigration policy, and a 2020 Seattle mass shooting—the last of which sent fares soaring as much as 500%. Uber’s algorithmic pricing has consistently sparked criticism from the ride-sharing company’s 93 million active users. Even on the night of the London Bridge attack, after Uber manually halted surge pricing near London Bridge, it remained in effect for the surrounding areas of central London for another 50 minutes.
An economist might applaud Uber’s pricing engine: As demand increases relative to supply, the price of a ride climbs. For customers, however, the cost of using the service can seem as unpredictable as the spin of a roulette wheel.
Uber isn’t the only company facing this problem. Firms in many industries—including advertising, e-commerce, entertainment, insurance, sports, travel, and utilities—have employed dynamic pricing with varying degrees of success. A classic and well-known example is Coca-Cola, which experimented in the late 1990s with temperature-sensitive vending machines that would increase the price of a beverage on a hot day. The company quickly abandoned the project in the wake of public outrage.
Pricing algorithms are intended to help firms determine optimal prices on a near real-time basis. They use artificial intelligence and machine learning to weigh variables such as supply and demand, competitor pricing, and delivery time. Unfortunately, algorithms occasionally go rogue and come up with figures no one would ever pay—from $14,000 for a cabinet listed on Wayfair to almost $24 million for a textbook offered on Amazon. But such snafus are just one of the risks when companies entrust decision-making to computers.
The constant changes in price points send strong signals to customers that need to be properly managed. Yet many organizations fail to appreciate this. They know that prices affect decisions about when and what to purchase, but they overlook the fact that continual ups and downs may trigger unfavorable perceptions of their offerings and, importantly, the company itself.
Brands thus need to consider more than simple math when employing algorithmic systems. These systems can create an uncomfortable tension between earning customer loyalty and earning money. But implemented correctly, they can maximize revenue while also making customers feel as if they have paid the appropriate amount for a product or service.
In the worst cases, algorithms turn the already delicate task of asking customers for money into an experience that drives them away.
In this article, we explore the psychology at play when companies ask customers for money. We examine real-world examples of algorithmic pricing and the ways in which it benefited or harmed the associated brand. We also detail the advantages of proper oversight and management, including determining which business unit should own the effort and what parameters should be set to limit the potential for misuse.
Let’s start with the case of Root Insurance, which sells auto policies in 30 U.S. states. To better educate and foster relationships with its customers, the company devised a dynamic pricing program that treats each driver in a personal and transparent manner. Unlike its competitors, Root doesn’t segment pricing using large, relatively anonymous risk pools generated from demographic data. Instead, it offers drivers a smartphone app that measures their day-to-day behavior behind the wheel. This data gets fed into an algorithm to calculate individual safety scores. Root then bases insurance premiums primarily on how well drivers perform, while giving some weight to traditional factors such as credit scores and insurance fraud statistics. To reduce bias against underresourced customers, Root avoids considering anyone’s education or occupation (other common industry factors), and it has committed to dropping credit scores from its rates by 2025. The company also insures only those people who pass its safety test. By weeding out bad drivers, Root claims it can reduce the expenses associated with accidents and lower the price of insurance for all its customers.
Root’s model is an effective example of how pricing algorithms—and transparency about them—can improve customer relationships. First, before a customer ever sees the price of a Root policy, she knows what the company does and doesn’t take into consideration. Second, the customer knows why she was offered a specific price that differs from what someone else might pay. Third, she knows what Root did on her behalf to minimize the final cost of insurance.
Making customers understand the mutually beneficial nature of algorithmic pricing is key to its success. That’s because overpaying for something can be painful—literally. Research conducted by neuroscientists at Carnegie Mellon, Stanford, and MIT has shown that pain centers in the human brain are activated when people see a product with an excessive price tag.
The mere act of asking for money—regardless of when or how—instantly shifts the focus of the customer relationship from pursuing aligned interests to reconciling opposing interests. In the worst cases, asking for money can be alienating to customers. The challenge for the customer-centric organization is to minimize the risks and limit the damage that occurs when market norms drive price increases and intrude upon an otherwise well-nurtured relationship.
Tsilli Pines
Before pricing algorithms became widely used, prices were stickier and differed little from one seller to another. Customers had relatively stable expectations and did not perceive prices as personal. Whenever price changes created discrepancies between actual and expected cost, it was easier for customers to rationalize the increases, believing that they were being implemented universally as part of a carefully crafted corporate strategy.
Technology has made the clashes more frequent, more arbitrary-seeming, and more startling in size—which unsettles customers and makes it harder than ever for them to reconcile what they see with what they expect. At the same time, many firms have come to believe that whenever customers’ price expectations are stable and disruptions are minimal, the company must be leaving too much money on the table. In line with market norms, firms have increasingly turned to algorithms to maximize their profits. Today even the slowest-moving B2B industries are replacing Excel spreadsheets with powerful algorithmic-pricing tools.
Technology has enabled firms to deepen their relationships with customers and, in parallel, become more efficient and proficient in extracting money from them. This combination, however, often leaves customers wondering what they should think and which companies they should trust. With their price sensitivity heightened, they work overtime to try to make sense of price changes. What do the fluctuations say about the quality or desirability of the product or service they’re buying? About the motives and values of the seller? What does that firm really think of their patronage?
If price changes reach an equilibrium, the urgency of these questions can fade. But if the frequency and magnitude of intrusions remain uncertain, these questions will linger and ultimately force customers to draw their own conclusions, without explicit guidance from the seller. That is when customers start reacting to the algorithm’s messages, not the firm’s—a risky proposition for any business.
To better control what algorithmic pricing says to customers and how it impacts customer relationships, we offer four recommendations, along with illustrative examples that help clarify how each recommendation can be applied.
In 2020 the Swedish furniture retailer IKEA launched a novel initiative at its Dubai location. For a limited period, the company allowed customers to pay different prices for products according to the time they spent driving to the store. Every item—from a sandwich at the restaurant to a complete bedroom set—had a price expressed in two units: the local currency and a time amount. A family that drove, say, 45 minutes to the IKEA store earned a certain value tied to the distance of its trip. At checkout, the family could show the cashier a Google Maps Timeline readout (using a feature of the Google Maps cell phone app that tracks and records all the routes one takes). The cashier would run an algorithm that factored time spent, distance traveled, and the average hourly wage of a Dubai worker to calculate the monetary value of the ride. The store then offered that value as a form of currency. The longer the trip, the more time credits the family got and the less money it needed to fork over.
The clear inference shoppers drew from IKEA’s program was that the retailer wanted to incentivize them to travel great distances to its stores. Although different customers would pay different prices for the same items, and individual customers might see different prices each time they visited (depending on where they came from), they nonetheless felt they had agency in how much they would pay. That contrasts with the helplessness people often experience during pricing surges. Best of all, because customers’ out-of-pocket costs could only decrease—in conjunction with distance traveled—as opposed to increasing as a result of heightened demand, no one ever paid more than the price advertised on the company’s website. In other words, IKEA used the distance-based algorithm to reward customers rather than penalize them. It might have lost some immediate revenue: Shoppers who drove far enough could get steep discounts or even obtain some products free. But by choosing a proper use case, with built-in incentives for people to visit the store, the company probably attracted more remote customers and increased all customers’ loyalty (and theoretical lifetime value).
Models like IKEA’s are rare. Companies typically employ dynamic pricing to further their short-term financial goals with little regard for customers’ perceptions. Yet the sheer volume and the intensity of price changes implemented by algorithms send unequivocal signals to buyers about everything from a company’s mission and values to the quality of its offerings. These signals can crowd out other efforts to shape the narrative in a brand’s relationship with its customers. In the worst cases, algorithms turn the already delicate task of asking them for money into an experience that drives them away. That is why firms cannot leave the management of pricing technology to data scientists alone.
Tsilli Pines
The path to improvement is not just technical but organizational and psychological. As paradoxical as it might sound, a better algorithm might make matters worse—by exploiting circumstances and stirring resentment, as happened with Uber during the London Bridge attack.
Overcoming the organizational challenge starts with recognizing that algorithmic pricing is not simply a means to generate prices that bring supply and demand into balance. It is, in fact, a principle that needs to align with one’s organization from top to bottom.
When customers have the impression that a firm bases its prices solely on supply and demand, the inferences they draw can be harmful. Think of an innovative firm with highly differentiated offerings. When that firm emphasizes supply and demand in its pricing algorithm, it is essentially telling customers that the value of its product is mostly related to whether it is available or not—not how well it solves customers’ problems or performs relative to competitors. Additionally, customers can learn to game the system and time their purchases to coincide with a moment when they believe the price is low. This again drives commoditization. By contrast, IKEA’s dynamic pricing model focused on attracting unlikely customers rather than penalizing likely customers because of a lack of supply.
In 2019 United Airlines eliminated the mileage tables that frequent fliers relied on to redeem their reward points. It replaced the tables with an algorithmic pricing model, explained why it was necessary to tie award travel to supply and demand, and emphasized how customers could benefit (by spending fewer award miles for off-peak flights).
The new system did result, though, in higher award prices for high-demand flights. That certainly frustrated rewards customers, but the airline communicated all the changes in an easily understandable way, and it focused its efforts on a specific (and presumably loyal) customer base. In doing so, it was able to mitigate significant reputational damage. Additionally, because it delegated management of the new algorithm to the team that supervised the loyalty program, United gave clear ownership of the pricing system to a department that was highly attuned to the sensitivities of the most steadfast customers. That strategy enabled the airline to monitor and quickly respond to glitches with the algorithm or challenges with customer relationships.
It is easy to blame the algorithms themselves when they go haywire, but the root causes of the problems usually lie in other areas—inadequate organizational attention or a failure to appreciate customer psychology. Most firms have an incomplete understanding of what really happens when they ask customers for money. They focus too intensively on the numbers, which they view as little more than the passive outcomes of the market forces that shape supply and demand. To use Adam Smith’s term, the “invisible hand” does the work, not the firm itself.
This myopia leads companies to overlook all the other information that prices convey. Even when organizations do recognize the power of this information and its implications, most firms cannot manage it effectively, because pricing is an organizational orphan, with no clearly defined leadership, responsibility, and accountability.
When companies blithely hand off the heavy lifting of pricing to automation, they cede to the algorithms not only the control of the math but also the messaging. While the data scientists, data analysts, and pricing specialists focus on optimizing the numbers, who is making sure that the messages are optimal? The answer in many organizations is no one.
A pricing algorithm on its own has two weaknesses. First, it lacks the empathy required to anticipate and understand the behavioral and psychological effects that price changes have on customers. Second, it lacks the long-term perspective required to ensure compliance with a corporate strategy or overarching purpose. By emphasizing only supply-and-demand fluctuations in real time, the algorithm runs counter to marketing teams’ aims for longer-term relationships and loyalty. This conflict between long-range thinking and real-time price changes does not merely intensify the clash between earning goodwill and earning money; it also increases the urgency of finding a solution before the brand suffers irreversible damage.
If a firm does not manage its price setting and messaging proactively and strategically, it can trigger and even accelerate the commoditization of its offerings by heightening price sensitivity, undermining price-value relationships, and tarnishing brand reputation. But by empowering a team that can plan its initiatives and make in-the-moment decisions about them, the company can pivot quickly when predicaments occur.
Think about a typical poor experience at a theme park. Guests have to suffer through long lines for rides, food, and restrooms, plus a lack of personal attention from overwhelmed or undertrained support staff. Such an off-putting experience leaves many customers wondering whether their steep investment in tickets, parking, refreshments, and lodging is even worth it. Guests would have a more pleasant visit if they encountered shorter lines and wait times and had better interactions with park personnel.
To increase customer satisfaction, Walt Disney World, in Orlando, Florida, changed its dynamic price structure from a manual to an algorithmic one in 2018. The new program, which raised multiday-ticket prices overall but decreased the price of tickets for off-peak dates, encouraged customers to plan their trips well in advance or book trips during off-peak periods in order to take advantage of lower prices.
The handcrafted infographics of Tsilli Pines explore the power that numbers have in our lives and the topography of money in human consciousness. Tsilli Pines
Disney’s program has several merits: First, it shows that dynamic pricing can serve other objectives besides increasing revenue or volume. Even if total revenue and overall guest count stay constant over time, the pricing structure makes the flow of customers steadier, which means less volatility in Disney’s needs for staff and other resources. That can lead to significant cost savings. Second, the customer experience improves dramatically because guests can enjoy more rides, visit more attractions, and better use their time in the parks. Finally, the dynamic pricing program can be explicitly publicized as a commitment to long-term customer satisfaction (despite an overall increase in prices).
When Disney World switched to its algorithmic system, it also determined that it would be in its best interest to no longer dynamically price single-day entry to its individual theme parks (Magic Kingdom, Epcot, Animal Kingdom, and Hollywood Studios). Pricing for single-day tickets across all four properties was set from $109 to $129 no matter what time of year a customer chose to visit, regardless of demand. That guardrail limited the amount that Disney could charge for a single-day pass, but it set clear parameters that helped customers anticipate their costs and plan their visits. And by observing how they self-selected their trips, Disney could sharpen its communication about the park experience and design additional service packages to cater to different customer segments.
Other companies can use guardrails in a similar way— not just to protect customers from wild price swings but also to judge how pricing impacts every area of the organization. When establishing the initial guardrails and continuing to deploy them, firms should encourage information sharing among different lines of business. That’s the best way to extract key learnings and use them for the company’s benefit. We see three primary areas for closer collaboration across functions to glean insights from algorithms:
Controlled, periodic testing of prices can help a company measure the extent to which customers value a product or service, or any of its features, and understand the context of when and how they derive that value. Indeed, pricing experimentation can be far more powerful than traditional market research, because customers are reacting to actual offerings and making real transactions. Their responses to price shifts help firms discover what works, what doesn’t, and at what point buyers first make their purchase decisions.
Firms can develop a new key performance indicator or compare existing indicators to ensure that the frequency and magnitude of price changes are not eroding customer loyalty or brand reputation. No company wants to be perceived as unfair, manipulative, or greedy. Thus it’s important to take measures to constrain and manage the output of the pricing algorithms, and vital to think through the messages and their consequences in advance. This enables firms to avoid extreme and free-floating prices by implementing hard floors and ceilings, as Disney did with its fixed single-day pricing.
This is essentially a long-term, integrated view of the first two elements. Are the firm’s product development, branding, positioning, and pricing all working in harmony—or with the least amount of friction—to fulfill the company’s strategic objectives? The firm must strive to ascertain, directly or indirectly, how customers perceive its mission and purpose and whether its price actions reinforce or harm the reputation it’s trying to establish. The messages that customers infer from prices should sync up with the explicit messages that a company communicates through its nonprice activities to promote itself and its products.
When firms pay attention to all the various ways that price changes can alter what customers believe and how they behave—beyond the immediate buy-or-no-buy decision—they can enhance the customer relationship rather than diminish it, even when they raise prices. Firms can tap into the power of price changes to improve their operations and at the same time create a better overall experience for customers.
Far from the “set it and forget it” approach to pricing that was common in the past, organizations with a dynamic strategy must take a more proactive, creative stance to achieve the desired results. For Disney, IKEA, and United Airlines, the aims were simple: The brands wanted to make it worth the customers’ while to transact, even under less-than-ideal circumstances (on less convenient days, or despite long commutes to brick-and-mortar locations). They also wanted to benefit from being able to manage how, when, and why changes in pricing were communicated.
The best pricing algorithms can analyze customer data and other information to generate optimal prices for any given customer at any given time. But from whose perspective are those prices optimal? That question gets at the conflict between earning customer goodwill and earning more money, which presents a complicated organizational challenge that should be overseen by a clear owner and managed when necessary. Sometimes the algorithm might need to be tweaked; other times its use might need to be temporarily suspended.
The day after the London Bridge attack, Uber announced that it had refunded the payments of all riders who had hired a car in the affected area. It also boasted that its drivers had helped tens of thousands of people flee the scene. Both announcements would likely have enhanced the company’s reputation had it not just been tarnished by the swift backlash to the price surge. Although it is difficult to quantify the lasting negative impact of that surge on Uber’s relationship with its customers, it’s clear that a faster response or a more proactive mechanism for preventing the soaring prices would have benefited the brand and the riders served that evening.
All companies should understand what their pricing algorithms are communicating to customers and how best to control that message. To effectively do so, they must develop a proper use case and narrative for implementing algorithmic pricing, assign an owner to monitor pricing guardrails, and empower that owner to manage or override the automation when necessary. By doing so, companies will be able to optimize dynamic pricing in real time without sacrificing customer loyalty or harming their reputation.
The Pitfalls of Pricing Algorithms
Research & References of The Pitfalls of Pricing Algorithms|A&C Accounting And Tax Services
Source
0 Comments