Personalized pricing on the net

September 2000

Is personalized pricing an opportunity to be exploited on the Net, or a hazard that can permanently harm customer relations? recently charged customers different prices to buy the same DVD at the same time, which started a furor chronicled in Wired News and Computerworld. At first, some customers thought that these prices were personalized prices, in which the prior behavior of customers influenced the prices charged to specific customers.

Amazon subsequently stated that the prices were randomly generated in order to estimate how customers responded to different prices. The furor did not subside with this explanation and Amazon has changed its policy somewhat. Now Amazon will charge different prices to different customers when testing the effect of different prices, but it will rebate anything paid over the lowest price after it is done testing.

On one level, it is hard to understand why anyone would get upset. Firms routinely charge different prices to different customers. Offline firms charge different prices to customers all the time. If you use a coupon to buy a box of crackers, you are paying a different price than everyone else who buys the crackers without a coupon. If you use a frequent-purchaser card at a grocery store, you pay different prices than others who don't use a card and you get coupons that are related to your purchases. If you purchase online, you may well get a coupon with the purchase that gives you money off your next purchase. There's no furor over any of this.

An important difference is that, in the examples above, the different prices are available to everyone who is similarly situated and the customer decides whether he or she wants a lower price. Anyone can clip coupons out of the paper, anyone can use a frequent-purchaser card and anyone can use a coupon from a prior purchase.

Random prices are not readily and apparently available to similarly situated consumers by taking some action. Not surprisingly, those who pay higher prices resent it; they do not have the opportunity to pay a lower price. Moreover, it does not help much that the higher prices are random. Random here really means determined arbitrarily by a computer. Being charged a higher price randomly is like getting a bill for higher taxes randomly.

Resentment at being charged different prices for no good reason may be enough to explain the continued furor over random pricing, but it is not adequate. The furor started over a surmise that Amazon was using personalized pricing: charging different prices based on the customer's prior purchases. A good explanation is liable to explain the furor over both random prices and personalized prices.

If people don't object to personalized prices offline, why would they object to them online? Frequent-purchaser cards are common in grocery stores and no one has started demonstrations against frequent-purchaser cards. While the lower cost of offline demonstrations than online demonstrations may be part of the answer, there is a better explanation.

If some customers pay more than others, those who pay more have an incentive to try to get the lower price.

In the case of coupons and frequent-purchaser cards, buyers segregate themselves into those who pay more or less by their own actions. You either clip coupons or you don't; you either get a frequent-purchaser card or you don't. If you don't want to be bothered with coupons and cards, you're indicating that you're willing to pay the higher price and you pay it.

Online, the seller automatically has a record of purchases that is easy to keep and obvious to the buyer. Every customer of Amazon can view a record of his or her purchases, which makes it obvious that Amazon has this record. It is not clear whether Amazon would charge higher or lower prices to frequent purchasers.

Suppose that Amazon charges higher prices to frequent purchasers. This was the initial complaint. Then frequent purchasers have an incentive to attempt to pretend that they are someone else who is not a frequent purchasers.

Avoiding the higher prices involves costs, whether it is the cost of pretending to be someone else, getting a different e-mail address, using a different credit card or having the books shipped to someone else. This may seem far-fetched, but if a buyer spends $200 a month at Amazon and the difference in price is only ten percent, there's $20 a month involved — an extra book, CD or DVD every month.

These issues do not arise for frequent purchasers if Amazon charges lower prices to them. If infrequent purchasers pay more, these infrequent purchasers of, say, books have an incentive to ask someone who is a frequent purchaser to include the book in their next purchase. It will be worth some people's while to do this for $5 and it won't be worth other people's while.

In sum, it is unlikely to be in a seller's interest to try to charge more to frequent purchasers. Any such attempt by the seller creates an incentive for frequent purchasers to pretend to be infrequent purchasers: this effort makes the frequent customers worse off and does not make the seller better off.

Any seller that wants to use personalized pricing has the problem of making it verifiably clear to frequent customers that they pay less.

Online firms have a problem making it clear to frequent purchasers that they are not paying more. There is no card that a customer can choose to use or not use, such as a grocery store's frequent purchase card. The easiest way for an online firm to make it verifiably clear to buyers that they are not paying more is to send e-mails with discount codes. This is what online firms do.

What about personalized pricing based on something besides prior purchases? For example, at least one firm asks a buyer if he or she is student when one of the books purchased is a textbook. Is this likely to persist?

Similar reasoning leads to the conclusion that pricing based on customers' self-description is not likely to be useful. When the seller asks the buyer whether he or she is a student, what is the best response? It seems to me unlikely that a bookseller will attempt to charge students more for books. The best response then is to say: I'm a student. In short, self-identification is unlikely to be helpful for personalized pricing because of the low cost of pretending to be in a low-price group.

Finally, what does this emphasis on the costs of avoiding higher prices say about random prices presented at a Web site? Easy. If you are uncertain whether you have received a price that is randomly higher, you have an incentive to log on and check whether you can get a lower price. This is a pain for the buyer with no offsetting benefit.