The technical obstacle to pulling this off has to do with information. It's very hard to know where a person's limits are and when they'll decide something isn't worth the price. Music artists and some software designers have tried experimenting with this by saying to consumers, "Pay what you want for our product." Radiohead famously tried this tactic in 2007 for its album In Rainbows and wound up pulling in between $5 and $8 per copy, on average.
Yet analyzing transaction data after the fact is very different from forecasting what customers might be willing to pay in the future. Luckily for e-commerce, companies are getting better at it all the time.
Via Modeled Behavior, a new paper from Brandeis University economist Benjamin Shiller shows that combining traditional demographic data with information about Web browsing habits led to much more accurate predictions as to when a consumer would commit to a Netflix subscription. Furthermore, tailoring Netflix's prices according to consumers' estimated willingness to pay — a process known as first-degree price discrimination — led to higher profits in simulations. (Second- and third-degree price discrimination respectively occur when a business either changes the price of its goods based on the amount you buy or changes the price by grouping customers into categories, such as veterans or seniors.)
Shiller began by looking at ComScore data on a range of Web users, including Netflix subscribers and non-subscribers. By correlating the subscribers with demographic data, Shiller got a baseline idea of what might predict a subscription purchase. Then Shiller tested individual Web behaviors to see which might improve the prediction's accuracy. For instance, people who used Rotten Tomatoes, Wikipedia or Blockbuster.com were all more likely than not to have a Netflix subscription.
Armed with that information, Shiller plugged the results into a probability formula to see how people might respond to first-degree price discrimination.
With the formula, "one can simulate expected profits under counterfactual prices," wrote Shiller. "Any given set of prices implies some probabilities that an individual consumes each tier [of Netflix service]."
Using demographic data alone offered a coarse prediction of how users might reasonably distribute themselves above and below Netflix's actual price point, the real-world one that's unaffected by price discrimination. By some addition, Shiller was able to calculate how much money Netflix could make from that model. Factoring in Web-browsing data to the predictive model led to even more pronounced results.
"Using all variables to tailor prices, one can yield variable profits 1.39 percent higher than variable profits obtained using non-tailored 2nd degree price-discrimination," Shiller writes. "Using demographics alone to tailor prices raises profits by much less, yielding variable profits only 0.14% higher than variable profits attainable under 2nd degree [price discrimination]."
In plain English, that means that the more information a business has about you, the better it is at predicting how much money it can extract from you before you revolt. We've seen this before in a number of examples. Orbitz last year was accused of charging Mac users more for plane tickets than non-Mac users. And a Wall Street Journal investigation found that some businesses were charging customers more based on their location.
To be clear, Netflix itself hasn't begun variable pricing; Shiller is simply conducting a thought experiment. But it might not be far off, if Netflix lives up to its reputation as a data-savvy company that knows how to capitalize on us almost without our knowing it.