Forecasting in Viral Marketing Campaigns
Shared by: tls14265
Forecasting in Viral Marketing Campaigns by Ralf van der Lans, Gerrit van Bruggen, Jehoshua Eliashberg, and Berend Wierenga1 Word-of-Mouth communications are considered as one of the most important antecedents of consumer behavior. The introduction of internet has dramatically increased the possibility of consumers to exchange information, and electronic word-of-mouth (eWOM) is now a common phenomenon. Marketers have recognized the opportunities of eWOM and viral marketing campaigns are getting increasingly popular. Nowadays we observe a wide range of different campaigns, from emails offering free products and promotions such as Hotmail, to emails inviting friends to watch funny commercials or play online video games on a company’s website. In this research we focus specifically on viral marketing campaigns using online gaming. The popularity of these viral games is growing rapidly, especially because many consumers like these games and therefore spend time on a company’s website. Viral games may have several goals. First, similar to advertising, viral games create awareness, positive attitudes, and may promote products. Second, because viral games are interactive, companies may obtain interesting consumer information, their preferences and opinions. Third, during viral games a participant is asked to invite a friend by email to visit the game as well. Thus, viral games result in large databases with consumer information, including email-addresses that the company may use to target consumers. For example, in the viral game that we study in this research, the company invited consumers to visit their branch office to pick up a price and to receive information about their products. Critical elements for the success of a viral marketing campaign are 1) the start up of the process, i.e. initial participants need to be informed so that they start spreading eWOM, and 2) the design of the campaign. Marketing tools to start a viral marketing campaign are: sending emails to consumers (i.e., seeding), placing banners on websites, or traditional advertisements and coupons. Important game design elements are the structure of the game, e.g. a competitive element so that consumers can compete against friends, and the prizes consumers may win for providing information or inviting friends. An interesting question is of course how many consumers will participate in the viral marketing campaign, given its startup strategy and the design of the campaign. An early forecast of the success of a campaign is critical to estimate the reach of the campaign in relationship to its operational costs. Furthermore, at the beginning of the process, marketers still have the opportunity to guide the process to the desired direction. 1 Ralf van der Lans (presenting author) is Assistant Professor, RSM Erasmus University, Gerrit van Bruggen is Professor of Marketing, RSM Erasmus University, Jehoshua Eliashberg is Sebastian S. Kresge Professor of Marketing and Professor of Operations and Information Management, The Wharton School, University of Pennsylvania, and Berend Wierenga is Professor of Marketing, RSM Erasmus University. The authors thank Patrick Filius and Klaas Weima of Energize for their helpful suggestions during the project, and for providing the dataset. In this research we propose a dynamic stochastic model that uses the detailed information available in viral marketing campaigns, to predict the success and reach of the campaign in the early stages. The model is based on the theory of branching processes, which is used extensively in epidemics and biology to predict the spread of viruses and the growth of populations. A branching process is a specific Markov process in which a consumer may generate new consumers, in this case by sending emails. The states in this Markov process are the number of consumers that currently have an unopened email in their mailbox. More specifically, the proposed model simultaneously predicts the number of invited consumers, the number of visitors to the website generated by these invitations as well as by the different marketing activities. We applied the model to an actual viral marketing campaign that was online for 5 weeks, and which attracted over 500,000 participants to a viral game in which more than 200,000 participants registered themselves, and invited over 1 million friends. Initial results show that the model is able to predict these numbers of participants accurately already in the very early stages of the campaign. This start of the campaign is most important, since during this period a marketer still has the possibility to influence the process. Critical model parameters are the probability that someone accepts an invitation from the company, and from a friend respectively, the number of friends a participant invites, and the time someone needs to open an email.