The year was 1999, the turn of the millennium. eBay, the online auction site, was booming. Google, having just discovered that it could auction click-through advertising spots, had at last found its business model. Throughout the Western world, B2B trading hubs, which matched buyers with sellers, often using auctions, were given huge valuations.
Business magazines, business schools and authors piled in to chronicle the rise of the auction economy and the beginning of a new commerce based not on fixed prices, but on bidding and even bartering. The excitement was palpable. Technology, some economists said, would enable the world to return to a pre-industrial model where variable pricing was the norm, and fixed pricing a rarity.
And then what happened? eBay, to be sure, continued to boom, but a growing proportion of sales went to fixed priced sellers. Google continued to thrive, but its secret wasn’t its auction technology, but its huge traffic numbers. Most B2B markets collapsed; so too did many software suppliers. The auction economy found a nice, new niche, but there was no global revolution.
But the variable pricing revolution did not fizzle out – in fact, it has quietly taken root, aided by a combination of technologies that businesses use to set prices in almost every major market. These technologies include customer profiling and analytics, data warehousing, pricing optimisation, revenue management, and, recently, event processing systems. And perhaps, even more important than any single element, all of these components are being integrated in such a way that businesses can target individuals, and respond to events.
These technologies are expensive and difficult to implement and integrate, but the benefits go far beyond those of online auctions. Suppliers are able to take much greater control over their pricing – varying prices not just for the benefit of the customer but according to their own objectives. They can price dynamically to respond to demand, to create demand, to reduce waste and to turn over inventory more rapidly.
Indeed, in some cases, online retailers even see customer profiling and targeted pricing as a direct and powerful response to the threat created by online price comparison websites, such as Kelkoo or Pricegrabber, which threaten to pull down margins. By making special ‘one-to-one’ and private offers to customers, retailers can keep many of their real prices opaque, invisible to the price crawlers that trawl through their sites and inevitably eat into their profits.
The result of all this technology is that pricing has been quietly revolutionised over the past five years, with much more to come. People understand that the price of a flight to Paris, or an advert on Google can vary. But prices are also starting to vary widely, from customer to customer and even from hour to hour, for a much wider range of products and services. Some surprising examples: mobile phone calls, electricity, Internet domain names, the cost of insuring a drive to the shops, or even a basket of shopping.
Now, says Dr Judy Bayer, director of advanced analytics for Teradata, which provides technology for building large analytics databases, the biggest barrier to widespread adoption of variable pricing is no longer technology, but psychology. “There are many things you can do operationally and technically, but you have to be aware of what effect that can have on the customer. Customers don’t like to feel manipulated or out-witted, or even that they are being treated unfairly. I look on pricing as an area of psychology,” she adds.
The end of price
Almost all those companies that have adopted some form of dynamic pricing have experienced some resistance, not least because they are often breaking new ground. But this drawback is easily outweighed by the business benefits.
Variable pricing takes many forms. But while stock markets and auctions are well known, it is the rapid spread of targeted pricing that is having the most impact outside of specialist markets.
Targeted pricing focuses on both groups and on individuals. A striking example is Aviva’s Norwich Union company, the insurance group which recently introduced a pay-as-you drive insurance scheme. Young drivers are charged according to each individual journey, and the risks these journeys pose, rather than being charged one very high annual premium based on a demographic profile. For example, journeys made during the afternoon are cheaper than those made late at night. As a result, safety has been improved and the amount individuals pay has plummeted.
Anthony Lovick, a pricing actuary at Norwich Union, says that the company built one of the largest data warehouses in the world, with possibly the first ever “tera-row” database, in order to model the prices and risks in real time.
Retailers have also invested enormously in dynamic pricing systems, enabling them to vary their prices to stimulate demand, turn over inventory and maximise margins. Although a few experimented with auctions, or even reverse auctions, their focus now is mostly on targeted pricing, based on customer analytics.
Wal-Mart in the US and Tesco in the UK are both outstanding examples. At the recent Information Age BI conference, Dave Annis, data solutions director at Dunnhumby, Tesco’s analytics partner, gave one simple example of how pricing can vary: a customer with a Capital One card is classed as a more price sensitive buyer than one with an American Express card. And if they are buying jewellery, for example, they tend to buy aspirationally. Such buyers might be offered discounted branded goods, whereas the Amex card holders would not.
Although most retailers will change certain prices during the day, it is difficult to influence buying behaviour within the store. As Bayer points out, they can’t keep changing the prices on the shelves, and it is usually only at the checkout that the stores know who the customers are. But many are getting round this by emailing offers to customers before they shop, issuing coupons according to an individual profile. In the US, in particular, there is now a huge amount of promotional pricing. “Often, there will still be a regular price, but it will be a fiction,” says Bayer. Increasingly, offers are driven by inventories and sales, as well as by customer profiles.
Richard Gum, planning and distribution director for Grupo Cortefiel, a pan-European fashion retailer, is keen to stress that variable pricing has its limits. Although the company has invested in powerful data warehousing technology to calculate prices and their impact on profitability at a regional level, discounting in some sectors needs to be very large to affect buying behaviour. Real-time or targeted pricing is a secondary consideration – getting the pricing right on a seasonal or monthly basis is the first task.
A little-publicised example of targeted pricing comes from the mobile phone sector. Operator Vodafone can analyse phone call patterns in a way that is similar to shopping basket analysis, working out who in a social group may be the most influential. When it comes to contract renewal time, that information can be combined with call volume and profitability data to put a value on a customer and hence influence pricing.
Event-based pricing
Targeted pricing is a well-understood, if difficult, application of customer profiling. But what about targeting according to real-time demand, or in response to events?
This is at a much less advanced stage, because of the sophistication of the technology required. Because pricing is so important to overall profitability, prices are usually calculated either as an application based on an integrated database/data warehouse or by a specialist pricing package that draws feeds from other sources.
Adding in real-time data driven by real events, in addition to forecast data and models, is difficult and expensive, and requires an ‘active’ data warehouse – the ability to feed in data in real time as changes occur. Most early applications of active data warehousing do not touch on, yet still drive, pricing, although there are a few technical boundaries to overcome.
Alternatively, complex event processing (CEP) systems are just beginning to be adopted: These are easily programmed to enable systems to rapidly respond to real-time events – such as a sudden rush of demand for seats or a sudden change in the weather. Usually, they ‘pull’ historical data from the data warehouse, treating this as event information. Stephen Brobst, the CTO of Teradata, says organisations are beginning to do both, using both active data warehouses and CEP systems. But it is still early days.
Bertrand Vianney-Liaud, head of BI for the French supermarket group Casino, says his company uses an active data warehouse for customer targeting and profiling, but does not use real-time pricing. It does, however, use a pricing optimisation system, to calculate prices according to financial metrics.
Event-based pricing is of particular interest to the airline industry, where prices already fluctuate strongly according to demand. They are not, however, driven by real-time demand, says Alicia Acebo, the acting CIO of Skybus, and formerly a senior analytics expert with Continental Airlines. Instead, they rely on proven algorithms that are widely used in the airline industry to forecast demand and set prices accordingly.
Puneet Arora, the CTO of Tibco Software, is currently talking to several airlines about how its event processing technology could be used for real-time pricing. This will enable them to set their prices based on real events, such as adverse weather, rather than relying on historical models.
But it is unlikely to stop there. The best way to price goods and services, say analysts, is to use analytics and pricing optimisation software, combined with real-time input and some human judgment. With all this in place, prices will change like the weather.
The psychology of variable pricing
The technology of dynamic pricing can be complex and expensive, but at least its predictable. Customer psychology is another thing altogether. Dynamic pricing can mean one customer is charged a different price at different times; or it might mean different customers are charged different prices at the same time. Either way, some customers don’t like it, while others like it so much they start to second guess the system, skewing the models and sometimes undermining the business objectives.
Coca Cola found this out. It trialled a simple vending machine with temperature sensors built in. The higher the outside temperature, the higher the price of the nice cool cola. There was an outcry, a threatened boycott, and rival Pepsi promised it would never mess with customers in the same way. Coca Cola backed off.
Amazon, the online retailer, uses all kinds of variable pricing systems. This has created many suspicions among customers, with academics and bloggers trying to find out what it is up to. One academic study sampled thousands of prices offered to hundreds of individuals. It found a huge variety of prices, but couldn’t find any patterns. Amazon now sells historical pricing data to enable partners to understand ‘the long tail’. But experts say that Amazon’s own algorithms can’t be “reverse engineered”.
Many customers think that variable pricing is unfair, causing many retailers to be very wary. Quoting a different price to a known customer online is one thing; charging two different customers two different prices in a supermarket could create a nasty squabble at the checkout.
Airlines are very sensitive to this particular issue, and often advise customers not to discuss ticket prices with fellow passengers. Financial services companies, meanwhile, have been lambasted for consistently charging new customers less than their existing ones. Perhaps not surprisingly, credit card and mortgage companies now struggle with diminishing customer loyalty.
The price is right: Dynamic pricing models
Real-time matched trading
Products are traded in a real-time market, with a seller offering a price and a buyer accepting it. Prices are set by the last trade, or by aggregating recent deals. Used for commodities, where goods are well understood. Buyers are sophisticated and rapid transactions are critical. Works best with high volumes.
Technology: A scalable real-time trading system publicises offers, matches buyers to sellers, tracks prices, produces data and audit trails, manages participants and links to back-office fulfilment systems.
Examples: Stockmarkets; commodity spot markets; online gambling websites.
Auctions
Buyers compete to buy goods at the best prices in a managed, timed auction. Most effective where demand is high and supply is scarce. Reverse auctions, where suppliers bid to supply goods to a buyer for the lowest price, are increasingly popular in business-to-business markets.
Technology: An electronic auction system advertises goods or services for sale, manages security, conduct and membership, runs auctions, matches sellers and buyers and tracks all trades.
Examples: eBay, Google Adwords, B2B e-procurement (reverse auctions).
Variable pricing for targeted customers.
Suppliers build sophisticated profiles of customer segments, or of individual customers. Targeted offers are then made, or prices are set especially for these customers, either with the goal of enticing them to buy, or to maximise the supplier’s margin.
Technology: A customer relationship management system or database collects, holds or integrates customer data; an analytics system is usually used to profile and segment customers, and for mining data to identify patterns in customer behaviour. Some companies use specialist pricing optimisation software.
Examples: Retail special offers (Tesco), Pay-as-you-go insurance (Norwich Union).
Demand driven or real-time pricing
Prices are driven by demand, either real or forecast, and by yield management – the need to ensure that business goals are met and that margins are held high as possible.
Technology: A means of capturing demand-side events in real time, such as single or multiple sales, and feeding this back into a database that is also linked to supply data. Examples might be an analytics system that has been pre-programmed according to business goals to react to events.
Examples: Airline seats; electricity supply (in some markets).
Further reading in Information Age
Science for winners – January 2007
Customers in the tank – August 2006
A hard sell – August 2005
More articles can be found in the Business Intelligence Briefing Room
