Every lender in the United States uses credit scores to figure out how likely it is that a person will pay back a loan. But you may not know about the Customer Lifetime Value (CLV) that many companies use to tailor ads, prices, products, and service levels to each customer, or the Affluence Index, which ranks households by how much money they can spend. These are just a few of the scores that have come out recently as a result of the large amount of consumer information that can be found online. These consumer scores use information about a person’s age, race, gender, household income, zip code, and purchases to come up with numbers that stand in for important consumer traits or behaviours. But consumers can’t get these scores, which are different from traditional credit scores. Can a consumer benefit from collecting data even if the scores that come out of it end up being used “against” her, like by allowing companies to set prices for each person? Would knowing her score help? And how would businesses try to respond to what the customer said?
Problems with “Scoring”
One thing that makes these scores stand out is that the data brokers who make them also sell them to companies as part of their strategies for market segmentation. So, these scores don’t just affect how a consumer deals with one company. The information carried by the score creates links between how a consumer deals with different companies and industries over time. The case for it is that collecting data adds value by making trade more profitable, and scores are an easy way to package data. But bad effects on welfare can happen. For example, if a customer makes a big purchase, which raises her “profitability” score, she may have to pay more the next day.
The focus on price discrimination comes from the fact that e-commerce targeting and product-steering techniques are getting more and more precise, which makes price discrimination a real possibility. Prices are set using scores that are based on signals from past purchases. This is because purchases give information about willingness to pay, and willingness to pay goes up over time. In this situation, analysis looks at how the different levels of consumer sophistication (does the consumer know about the scores and the links they make?) and score transparency (can the consumer see what their current score is?) affect consumer welfare.
Good and bad effects
Price discrimination hurts consumers who aren’t smart enough to see the connections between their purchases, but it can help consumers who are smart enough to plan ahead. In the naive case, consumer welfare goes down with the quality of the signals firms have. In turn, firms do better. More surprising is that making a score out of data doesn’t protect consumers at all. This is because firms can combine information about purchases into a score that belongs to the class without losing the ability to predict. This class is set up by how much weight each score gives to past behaviour signals. A score with a lot of weight in the past has high persistence.
On the other hand, a strategic consumer can benefit from scores, even if firms use them against her in the end, because she can reduce the amount she wants to buy to change her score. Take a look at the diagram below, which shows a traditional monopoly problem with a consumer whose demand goes down and a single firm, which we’ll call Firm X. If there’s only one interaction, the consumer doesn’t change what she does, so Q units are bought at price P. But let’s say that a second company talks to the consumer tomorrow after seeing that they bought something in the first period. Because the consumer knows that her choice in the first period affects the price in the second, she will try to lower Firm Y’s signal, and therefore her score, by lowering her demand, which brings her purchases down to Q’.
When a consumer buys less, she loses. She also gets different prices in the future, which isn’t shown, because Firm Y gets information about how much she is willing to pay (the intercept of her demand function). But after the strategic demand drops, Firm X is forced to lower its price. If the consumer is willing to pay a lot, the discount on many units gives her enough of a benefit that she wants to be tracked.
Managing the strategic response of customers
The strategic reduction of demand means that purchases are less affected by changes in how much people are willing to pay. So, signals don’t tell us as much, and using scores to set prices is less effective. These losses can’t be stopped. If companies use scores that are the best predictors “ex post,” that is, based on the data they already have, then smart consumers will change their behaviour, making the data less useful to begin with. There is a complicated “cat and mouse” game going on, with consumers trying to “hide” while companies try to figure out what they want.
The study shows that companies choose a less-than-ideal way to use the data they have to improve the quality of the data they use. In particular, businesses can reduce their losses if they use persistent scores, which give too much weight to information from the past. This may seem strange, since the long-term effects of a score that stays high for a long time suggest that consumers might be afraid to share information and pay high prices for a long time if they do. But a score that gives more weight to the past also has less of a relationship with current willingness to pay. This means that changes in the score have less of an effect on prices at first. Scores that stick around longer than those that come up in a cat-and-mouse game can be more profitable because they give people a reason to share more information.
Score “Openness” Is Very Important
Our second contribution is to show that making scores public is a big part of how data collection could help consumers save money through lower prices. To show this, we look at the way the market works now, where the score is hidden from the customer.
When signals of purchases aren’t perfect, a smart consumer won’t be able to figure out her score just by looking at what she’s done in the past. But prices will tell you something. When a consumer sees a high price today, it tells her that firms think she is willing to pay a lot. This means that prices will stay high in the future because the trend is persistent. If the consumer then plans to buy a small number of units, she is less likely to cut her demand because the discount only applies to a small number of units. So, the consumer doesn’t care as much about price as they would if they could see the score. (In this last case, the consumer would be able to tell that a price is “abnormally” high if it is higher than what her score says it should be. This would allow her to pass up bad deals.
With a lower sensitivity, prices are more affected by the score. Even though this makes demand drop even more and makes people buy fewer items, prices are still higher, which hurts consumers. Also, consumers who are smart but have their scores hidden from them can end up worse off than those who aren’t. Our results can help shape policy. For example, if consumers are aware of the possibility of price discrimination and their scores are easy to understand, they may be better off.
