• Selected
• Entire List

Abstract:
The authors study the effect of word-of-mouth (WOM) marketing on member growth at an Internet social networking site and compare it with traditional marketing vehicles. Because social network sites record the electronic invitations sent out by existing members, outbound WOM may be precisely tracked. WOM, along with traditional marketing, can then be linked to the number of new members subsequently joining the site (signups). Due to the endogeneity among WOM, new signups, and traditional marketing activity, the authors employ a Vector Autoregression (VAR) modeling approach. Estimates from the VAR model show that word-of-mouth referrals have substantially longer carryover effects than traditional marketing actions. The long-run elasticity of signups with respect to WOM is estimated to be 0.53 (substantially larger than the average advertising elasticities reported in the literature) and the WOM elasticity is about 20 times higher than the elasticity for marketing events, and 30 times that of media appearances. Based on revenue from advertising impressions served to a new member, the monetary value of a WOM referral can be calculated; this yields an upper bound estimate for the financial incentives the firm might offer to stimulate word-of-mouth.

Word-of-Mouth Marketing, Internet, Social Networks, Vector Autoregression

Abstract:
Clickstream data are defined as the electronic record of Internet usage collected by Web servers or third-party services. The authors discuss the nature of clickstream data, noting key strengths and limitations of these data for research in marketing. The paper reviews major developments from the analysis of these data, covering advances in understanding (1) browsing and site usage behavior on the Internet, (2) the Internet's role and efficacy as a new medium for advertising and persuasion, and (3) shopping behavior on the Internet (i.e., electronic commerce). The authors outline opportunities for new research and highlight several emerging areas likely to grow in future importance. Inherent limitations of clickstream data for understanding and predicting the behavior of Internet users or researching marketing phenomena are also discussed.

Internet, Marketing, E-commerce, World Wide Web, Clickstream Data

Abstract:
In paid search advertising on Internet search engines, advertisers bid for specific keywords, e.g. "Rental Cars LAX," to display a text ad in the sponsored section of the search results page. The advertiser is charged when a user clicks on the ad. Many of the keywords in paid search campaigns generate few, if any, sales conversions - even over several months. This sparseness makes it difficult to assess the profit performance of individual keywords and has led to the practice of managing large groups of keywords together or relying on easy-to-calculate heuristics such as click-through rate (CTR). The authors develop a model of individual keyword conversion that addresses the sparseness problem. Conversion rates are estimated using a hierarchical Bayes binary choice model. This enables conversion to be based on both word-level covariates and shrinkage across keywords. The model is applied to keyword-level paid search data containing daily information on impressions, clicks and reservations for a major lodging chain. The results show that including keyword-level covariates and heterogeneity significantly improves conversion estimates. A holdout comparison suggests that campaign management based on the model, i.e., estimated cost-per-sale on a keyword level, would outperform existing managerial strategies.

Internet, Advertising, Paid Search, Bayesian Methods

Abstract:
In paid search advertising on the Internet, marketers can bid for search engines to serve their text ads in response to keyword searches that are either generic or branded. A generic keyword does not contain branded words, e.g., "Hotels LA"; a branded keyword does, e.g., "Hilton Hotels LA." Managers operating search campaigns face the question how to allocate resources across these types of keywords. For example, the advertising cost associated with a sale is typically much higher when it is tied to a generic versus a branded search. However, if advertising spillover occurs from generic search to branded search, metrics for generic keywords should be adjusted.

The purpose of this study is to investigate the nature and extent of potential spillover effects in paid search advertising. To do this, the authors model response to paid search advertising using the Nerlove-Arrow Goodwill Model. Exposure to brand-related information served after a search, e.g., seeing a paid search text ad or company website, is modeled as advertising which affects goodwill. The model incorporates the potential effect that goodwill has on search-related activity (e.g., impressions, clicks and sales) over time, thereby revealing whether spillover effects are present.

Using a Bayesian estimation approach, the authors apply the model to an aggregate dataset containing daily information on impressions, clicks and reservations from a paid search campaign for a major lodging chain. (Such information is commonly available to advertisers and managers rely upon it to operate search campaigns.) The results show that generic search activity affects branded impressions, clicks and reservations through positive goodwill. Branded search activity, however, does not affect generic activity, indicating that the spillover is asymmetric.

Internet, Advertising, Paid Search, Spillover, Goodwill, Nerlove-Arrow Model, Bayesian Dynamic Linear Model (DLM)

Abstract:
Due to customer segmentation, multiple types of dynamic business scenarios (business-as-usual, escalation, hysteresis, and evolving business practice; Dekimpe and Hanssens 1999) may co-exist within a single product market. The authors develop an approach to model this phenomenon with time series panel data. Unit-root tests are used to group panelists by whether or not outcome (e.g., sales) and marketing activity (e.g., advertising, promotion) variables are stationary or evolving. This produces four clusters corresponding to each business scenario. Next, panel-data vector autoregressive models appropriate for each panelist cluster are estimated to assess the dynamics and the magnitude of the response to marketing effort. The approach is applied to physician panel data on drug prescriptions and direct-to-physician promotions. Estimation results show markedly different response dynamics (as captured by impulse response functions) and elasticities across the physician groups. The approach also produces better in-sample and holdout fits than pooled data models. For firms that track customer-level marketing activity and response over time, a segmentation based on dynamic business scenarios provides a new tool for targeting and efficient marketing resource allocation.

Time series analysis, panel data, segmentation, marketing strategy

Abstract:
The success of Internet social networking sites depends on the number and activity levels of their user members. While users typically have numerous connections to other site members (i.e., “friends”), only a fraction of those “friends” may actually influence a member’s site usage. Since the influence of potentially hundreds of friends needs to be evaluated for each user, inferring precisely who is influential - and therefore of managerial interest for advertising targeting and retention efforts - is difficult. We develop an approach to determine which users have significant effects on the activities of others using the longitudinal records of members’ login activity. We propose a non-standard form of Bayesian shrinkage implemented in a Poisson regression. Instead of shrinking across panelists, strength is pooled across variables within the model for each user. The approach identifies the specific users who most influence others’ activity and does so considerably better than simpler alternatives. For the social networking site data, we find that, on average, about one-fifth of a user's friends actually influence his/her activity level on the site.

Internet, Social Networking, Bayesian Methods