Big Data and personalised pricing: consider yourself gamed

Chances are your every shopping move is being tracked. Phil Campbell/Flickr, CC BY

Imagine yourself the CEO of a company that mainly sells one product. One of your goals is to maximise profits. You know you can charge a flat price, or modestly raise profits by using quantity discounts or by group pricing, for example with seniors discounts.

But you feel frustrated. If you could somehow adjust prices to match your customers’ valuations, you could raise profits further. You suspect some consumers would be willing to pay much more for your product. Wouldn’t it be great if you could identify them and charge just those consumers a higher price?

You also know some people won’t purchase at the current price but would be willing to pay slightly less. You wish you could charge just those consumers a lower price. So much is left on the table.

Now cast yourself as a consumer. You don’t want to pay more than others, or more than you have to, for the product. So clearly you are not going to volunteer the information - that you have a high value for the item, to encourage retailers to charge you a higher price. To get a lower price, you would happily say you have a lower valuation, whether or not that is true.

Now suppose the firm is sneakier, offering different prices over time, and learning how much you will pay. There are companies that help you be so sneaky. Catalina Marketing Corp has used this strategy with moderate success, installing printers at store checkouts that produce coupons tailored to the purchases of each customer who uses a store card.

Such methods, while apparently effective, have their limits. If you, the consumer, follow one simple heuristic – “don’t buy at high prices,” the firm will infer you have little value for the product, and will offer you lower prices in the future. As a result you might want to forgo instant gratification to get lower prices later. Research confirms that when consumers are forward looking, many will do just that, making such pricing unprofitable.

Historically, this reluctance of consumers to reveal their willingness to pay, knowing it will be exploited, has been all that is needed to limit the effectiveness of person-specific pricing. But is the effectiveness limited anymore?

A market of one

What if you could learn more about consumers by hiring one private investigator per consumer, and tracking everything they do all day every day? Do they read celebrity gossip? Do they attend auto shows? These types of behaviours, and thousands of others, might reveal information that helps personalise prices for movie tickets or luxury cars.

But surely it would be ridiculously expensive to hire one private investigator per consumer?

Not anymore. Relevant behaviours can be acquired cheaply without a private investigator, for example by purchasing web-browsing histories from ISPs, browsers, individual websites, or third party advertisers. Then, the firm can observe whether a consumer visits celebrity gossip websites, or watches YouTube videos featuring sports cars, etc.

Moreover, location by time of day can be obtained from smartphones or cameras which automatically read license plates. Or, automatic facial recognition could allow cameras to track consumers. Facebook can already track users to some extent, through the time and geo-location data contained in uploaded photos and facial recognition software and tagged photos.

Where does it stop?

Firms could be more creative. For example, take Nest, a thermostat which over time learns owners’ habits, and reduces energy usage when the home is likely vacant. Nest could “ping” homes by raising the temperature remotely via the internet. If the homeowner is there, they would presumably adjust the thermostat, confirming their location. Maybe this is why Google recently paid US$3.2 billion for Nest.

By 2020, Cisco’s research estimates there will be about 50 billion internet connected devices, or 6.58 per person. In principle, all such devices can collect data on individuals, and report back to centralised databases. Is the idea of following each consumer around, learning their habits and preferences such a crazy idea anymore?

In the coming years, if personalised pricing becomes common, will it be possible for a consumer to avoid being charged high prices for the products they love most?

Suppose a consumer avoids all products with networked sensors, thinking that if they aren’t tracked they can’t be charged more. This strategy can easily be overcome by firms – charge astronomical prices to anyone for whom data is sparse.

Instead, suppose a consumer tries modifying behaviour to get lower prices. Would they really want to? They might have to change thousands of behaviours just to get a low price on one product. Moreover, they might not be able to, if they don’t know which subtle behaviours determine prices. What’s worse, those same behaviours could cause them to pay more for other products.

In fact, many companies today are using personalised prices. Staples Inc., for example, was observed charging lower prices to consumers living near a brick and mortar competitor. Freshplum Inc., a venture-funded startup, implements personalised discounts for many companies using a proprietary algorithm.

In my research, I investigated the impact of personalised pricing based on web browsing behaviour. For feasibility, I focused on Netflix, asking what would happen had it personalised prices. The answer was a 12% increase in profits. Good news for Netflix, but less so for some consumers who would be left paying roughly double the prices others do for the exact same product.