From Institute for Operations Research and the Management Sciences
Old models help new internet retailers in tough times, O.R. article says
As dotcoms face an uncertain future, success on the Internet is still resulting from reliable models born years before the World Wide Web, according to a study published in a journal of the Institute for Operations Research and the Management Sciences (INFORMS®).
"The newness and excitement over the Internet led many to overlook the fact that ideas like viral marketing have been improving business since the sixties," explains Alan L. Montgomery of Carnegie Mellon University. "In these more sober times, dotcoms that survive are using operations research methods in promotion, pricing, control, and forecasting to work smarter and increase sales." Professor Montgomery's article, "Applying Quantitative Marketing Techniques to the Internet," appears in the special e-commerce issue of Interfaces: An International Journal of the Institute for Operations Research and the Management Sciences. The special issue, which will be published on April 30, 2001, will be available free online at http://pubsonline.informs.org/. A companion site is at http://www.informs.org/ebiz/interfaces.
Hotmail's Viral Marketing
Prof. Montgomery says a case study is Hotmail, which coined the term viral marketing but mirrored 30-year old research in the field of diffusion models.
Spurring traditional methods, Hotmail grew from 20,000 to 12 million subscribers over two years by word of mouth, offering free e-mail services and earning profits in advertising dollars. Prof. Montgomery notes that a study by Frank M. Bass, "A new product growth model for consumer durables," (Management Science, 1969, 15:1) taught researchers to predict the number of users that will adopt an innovation. "There is remarkable agreement between the actual subscriber base and the model's predictions for Hotmail," Prof. Montgomery says.
The Bass model expresses the number of adopters at any time as a function of time, total market potential, and parameters that measure the effects of advertising and word of mouth. Other Internet retailers spur sales with diffusion models by targeting highly integrated groups of consumers like teenagers, he says.
Modeling a 150% Jump in Profits
Dotcoms face the problem of extracting marketing insights from huge amounts of data, says Prof. Montgomery. Operations research models have long existed to transform this type of information into potentially profitable decisions.
A technique called Bayesian modeling, he says, allows operations researchers to overcome the limits of uncertain data and make useful forecasts. It is helping retailers identify individual consumers' online purchasing behavior and target campaigns as effective marketers have for some time.
For example, he says, a brick and mortar retailer issuing a discount coupon relying solely on demographics, would send all consumers in an area a 10 cent coupon on a product and see increased profits of 10%. Going a step further and basing a "coupon drop" on information from the last purchase increased relative profit to 60%. But using Bayesian modeling to include all information from the purchase history and demographics would result in a much larger 160% increase in profits.
"These techniques can be applied directly to the Internet, for example by an online bookseller offering a virtual coupon," he says.
"Predictions from these types of models are generally quick and efficient from a computational standpoint. There is clearly a great knowledge base from which to draw in solving Internet marketing problems using quantitative operations research models."
Other quantitative models that market researchers are now applying to the Internet, he says, are: · Models employed by companies like Media Matrix that analyze clickstream data to target the placement of banner ads and create personalized web pages for customers.
· "Autoregressive tobit" models that use past usage patterns to forecast future usage. · Collaborative filtering, which companies like CDNow use to recommend new artists to customers. · Computer agents that are used to create virtual marketplaces for buying and selling goods. · Markov models, which consider the path visitors travel online and help e-tailers determine customers' price sensitivity.
· NBD models, which model repeat purchases and can be used to determine visitor frequencies at a website.
· Trial-and-repeat models, which are used by consumer packaged goods manufacturers to analyze initial trial purchases and subsequent repeat purchases.
· Learning models, which consider the time users spend online and can provide an estimate of "stickiness."
The Institute for Operations Research and the Management Sciences (INFORMS®) is an international scientific society with over 10,000 members, including Nobel Prize laureates, dedicated to applying scientific methods to help improve decision-making, management, and operations. Members of INFORMS work in business, government, and academia. They are represented in fields as diverse as airlines, health care, law enforcement, the military, the stock market, and telecommunications. The INFORMS website is at http://www.informs.org.