ISMS Press Release 2014, no.1
article in Volume 33, Issue 3, May-June 2014 of
Current Issue
Home Page
Subscribe
Contact Marketing Science
Contact for press releases: Barry List, Director of Marketing and PR, (443) 757-3560; (800)
4INFORMS; cell (443) 794-5182; barry.list@informs.org
Xiong, G., & Bharadwaj, S. (2014). Pre-release Buzz Evolution Patterns and New
Product Performance. Marketing Science, 33(3), 401-421.
Study Shows Online Buzz can Forecast New Product Performance Months before Product
Release and Reveals Important Link between Buzz and Firm Stock Prices
CATONSVILLE, MD, May XX, 2014 – Companies can significantly improve the forecasting
accuracy of forthcoming products’ performance by mining online consumer buzz prior to product
release, according to a study published by Marketing Science.
Social media attention to a firm’s forthcoming products also influences its stock price, the study
shows.
Pre-Release Buzz Evolution Patterns and New Product Performance is by Guiyang Xiong and
Sundar Bharadwaj, professors at Terry College of Business at the University of Georgia.
Sales forecasting before product release has been an important but challenging task, especially
for new products without prior sales history.
Pre-release buzz refers to consumers’ online conversations (e.g., in blogs and online forums)
about a new product before its release. “Online buzz is frequent, curated, and can be
conveniently tracked. With increasing consumer chatter and declining computing and storage
costs, such buzz can be mined for evolving patterns over time,” said Xiong. These patterns
reflect changes in consumer interest, vary by product, and can predict new product performance.
For example, the game Alan Wake was gaining increasing buzz over time before its release,
making it one of the most anticipated and successful games released in 2010.
“Amazingly, the partial pattern of buzz evolution months before product release can forecast new
product sales more accurately than the cumulative buzz until the day of product release,” said
Xiong. Such accurate forecasts of a new product’s sales well before its release allow sufficient
time for managers to make critical changes to the product’s design or marketing. In contrast,
traditional forecasting relies on past sales to predict future sales, which is not useful for prerelease strategies.
In addition, the research finds a key link between pre-release consumer buzz and firm stock
prices: Stock prices react immediately and positively to increases in buzz volume before product
release. This means that firms do not have to wait until product release to capitalize on the new
product.
“Firms may consider communicating early to investors about pre-release buzz, so as to increase
investor awareness of the buzz.” Bharadwaj said.
For their study, the authors examined over 800,000 blog and forum postings of more than 600
new video games released in 2009 and 2010. Building on recent advances in statistics and math
modeling, they employed Functional Data Analysis. By analyzing the shapes of pre-release buzz
evolution, this statistical model reduces forecasting error by 15-40% compared to the model
based on product characteristics alone.
The authors also examined the factors that influence pre-release buzz, such as advertising and
new product alliances, thus providing firms with guidance on how to manage the dynamics of
pre-release buzz to enhance new product performance.
The authors are members of ISMS (the INFORMS Society for Marketing Science). ISMS is a
group of scholars focused on describing, explaining, and predicting market phenomena at the
interface of firms and consumers.
Xiao, L., & Ding, M. (2014). Just the Faces: Exploring the Effects of Facial Features
in Print Advertising. Marketing Science, 33(3), 338-352.
New Facial Selection Technique for Ads Increases Potential Buyers by as Much as 15%,
Says New INFORMS Marketing Science Study
CATONSVILLE, MD, April XX, 2014 – Merely changing the face of a model in an ad increases
the number of potential purchasers by as much as 15% (8% on average), according to a study
published by Marketing Science entitled “Just the Faces: Exploring the Effects of Facial
Features in Print Advertising.”
The research was conducted by Li Xiao, Assistant Professor of Marketing at Fudan University
(China), and Min Ding, Smeal Professor of Marketing and Innovation at Smeal College of
Business, Pennsylvania State University, and Advisory Professor of Marketing at Fudan
University.
The study reveals that using computer science techniques to screen faces when designing ads
can transform the current subjective process into a scientifically automated one. Considering the
extensive use of human faces in advertising (over 50% of print ads contain human faces), this
technique may be quite profitable.
“This technique will revolutionize the field of ad design,” predicts author Min Ding.
The main technique is eigenface method, which has been widely used for face recognition
purposes in other fields, including personal device logons, human-computer interaction, and law
enforcement’s tracking of suspects. Eigenface method aims to identify each face by a small set
of key dimensions that together explain the variations in human faces.
The authors used eigenface method to extract and represent facial features in ads with a limited
set of eigenface weightings. In an experiment with 989 participants, the authors used real
models’ faces and real ads with minimal modifications to elicit participants’ natural reactions to
print ads. Their results show that different faces affect ad effectiveness substantially and people
show substantial differences in their facial preferences across product categories.
“An 8% increase in effectiveness could produce a substantial gain for the $600 billion ad
industry,” says author Li Xiao.
“These methods can substantially increase sales in individual industries,” add the authors. “For
example, there is a potential for up to $5 billion additional sales for the automotive industry in the
US alone.”
Ad agencies would use four steps to employ this technique in ad design. (1) Create a single
database (containing perhaps 1,000 or more) faces of professional models; (2) represent each
face in the database with a set of eigenface weightings; (3) measure the facial preferences of
target customers in a product category; and (4) identify the top faces that best match target
customers’ facial preferences for the specific product category.
These steps can be automated once enough data about characteristics of various product
categories and facial preferences have been collected.
The authors are members of ISMS. ISMS is a group of scholars focused on describing,
explaining, and predicting market phenomena at the interface of firms and consumers.
About Marketing Science
Marketing Science is an Institute for Operations Research
and the Management Sciences (INFORMS) publication
(SSCI indexed). We invite authors to submit for peer review
their best marketing-oriented research. We accept many
types of manuscripts. Please consider us as an authorfriendly outlet for your research. We are THE premier
journal focusing on empirical and theoretical quantitative
research in marketing. The journal is governed by the
INFORMS Society for Marketing Science.
About ISMS
ISMS is a group of scholars focused on describing,
explaining, and predicting market phenomena at the interface
of firms and consumers.
..