Haifa, Jan 12, 2005…Enhancing the online shopping experience is an issue for any business with a customer facing web site. Now an IBM researcher from Israel has come up with a way to accurately predict the potential lifetime value of each online customer, enabling businesses to provide enhanced services for the best customers and increase the effectiveness of marketing campaigns.
Although models that predict Customer Lifetime Value (CLV) already exist, they base their customer evaluation solely on the analysis of historical data from one customer, such as what items were bought in the past and at what frequency. However, existing technology can only identify a customer’s lifetime value once they have a notched up a significant purchase history within a particular online site. The new technology developed by IBM allows retailers to estimate the potential lifetime value of relatively new customers. It works by not only ‘learning’ from the spending patterns of a single customer, but also from the online activity of other more established customers. This way, by monitoring the buying behavior during the initial visits of a new customer, the model can assign the customer to a group and predict spending patterns based on those with the same 'buying profile'.
Once the model predicts that a visitor to a website has a high probability of purchasing, the company can decide, for example, to customize the experience by making relevant new offers to the right customers or prioritizing the customers’ requests so that they get faster online service. These new methods could provide a much more effective solution to web traffic management than the ‘first come first serve’ policies currently in place for most web sites. Such technologies are already being used in various domains, especially in the financial arena.
"It is a well known business principle that it is better to keep a good customer than find a new one," said Amit Fisher, the IBM researcher who began work on the model as part of his master's thesis. "By using intelligent data mining algorithms, we can predict the long-term value of a web customer, helping to optimize online businesses. At times when a website is experiencing peak demand, this type of technology could enable online retailers to prioritize the requests of the most valued customers – in a similar way to how 'gold card' loyalty schemes work in the hotel and airline industry".
The model was proven to be successful in a feasibility case study conducted on one of Israel’s top online auction sites. The model was constructed and validated based on actual customer data for an online auction site that serves thousands of customers each day. Daily data was taken for a one year period, during which time over 70,000 purchases took place, with a total value of over $18M. The new model was used to correlate customer behavior, accurately ranking customers with a better or worse potential for spending on the site—with a high degree of accuracy. In trials involving highly-populated groups of over 1,000 customers the model’s predictions were extremely accurate and came close to a correlation of 1.0 with actual data collected from the site.
This type of business optimization technology is becoming increasingly important to companies that are under pressure to add value to customers while at the same time cut costs and maximize resources. Fisher’s new system is typical of the kind of On Demand technology IBM’s business consultants are helping companies implement across various aspects of their business. The new Customer Lifetime Value model is currently being integrated into other "active technologies" being developed by staff at the IBM Haifa Research Labs. Active