The Performance Impact of online Customer

MSI Proposal
Conditions for Owned, Paid and Earned Media effectiveness:
The Performance Impact of online Customer-Initiated actions for Better
versus Lesser Known Brands of Search and Experience Goods
Ceren Demirci
Koen Pauwels
Ozyegin University, Istanbul, Turkey
Shuba Srinivasan
Boston University
Gokhan Yildirim
Carlos III University of Madrid, Spain
March 28, 2011
Executive Summary
Within the growing variety of online marketing, customer-initiated communication (CIC) such
as search and social media stands out in terms of current spending and future growth potential.
Many companies have invested in “paid” CIC, such as paid search and affiliate marketing,
“earned” CIC such as Facebook and Twitter campaigns, and “owned” CIC, such as search
engine optimization leading to higher organic search results. However, the relative
effectiveness of these different customer initiated actions remains an unanswered question.
This proposal takes a contingency perspective by considering customer initiated
digital media in relation with brand strength and the search versus experience nature of the
category. We hypothesize and demonstrate how these conditions favor different types of
customer-initiated online media. In particular, owned CIC should be more sales effective for
known brands, while paid CIC should be more sales effective for lesser known brands.
Likewise, earned CIC should be more important for experience goods than for search goods.
Combining these hypotheses results in a 2x2 matrix, which yields specific predictions for
managers running a better/lesser known brand in a search/experience good category.
Our empirical analysis uses Bayesian Vector Autoregressive (BVAR) models to
estimate long-term sales elasticities, while imposing restrictions to avoid over-parametrization
and wrongly-signed coefficients. Preliminary findings yield results consistent with the
hypotheses in two cases, and Marketing Science Institute’s support will enable us to investigate
the remaining cases. This proposal touches on all three key areas of the MSI call, in particular
the questions: “How does the impact of advertising vary by context?”, “How can firms
compare the effectiveness of diverse marketing communication activities?” and “What is the
role of the brand in an environment characterized by the emergence of social media and
complex, multiple touch points between brands and customers?”
Research Statement
Since the introduction of the first banner ad in 1994, online advertising has redefined the global
advertising landscape. Spending in the sector has continued to grow, reaching $55.2 billion
globally in 2009. Expectations are for this expansion to continue, reaching $61.8 billion in
2010 and up to $96.8 billion by 2014 (EMarketer, 2010). However, managers still struggle to
quantify the return on investment for activities as different as banner ads and social media.
Moreover, the current hype for new media makes it difficult to establish boundary conditions
for success stories. Which online media work best for which kinds of brands and products?
Several studies have started to explore the sales effectiveness of online advertising.
One emerging consensus is that customer-initiated actions; such as search and social media,
have a higher sales elasticity than firm-initiated actions, such as banner ads and email (Gartner,
Inc., 2008). Indeed, Manchanda et al. (2006) found that banner ads have a lower elasticity
(0.02) than offline advertising for a health and beauty product, while Wiesel et al (2010) found
that paid search has a higher elasticity (4.35) than offline advertising for an office furniture
supplier. Likewise, improvements in “earned” (i.e. social) media metrics such as Facebook
likes may substantially increase sales (Pauwels et al. 2010). Finally, many companies have
invested substantially in “owned” media, including search engine optimization to rank higher in
organic search. These forms of paid, earned and owned media have in common that potential
customers initiated the online search and/or conversation. However, the relative effectiveness
of these different customer initiated actions remains an unanswered question.
The purpose of this proposal is to understand how the effectiveness of paid, earned
and owned online media varies by brand and category context. Our hypotheses in Table 1 focus
on brand strength, and on the search versus experience nature of the category.
Research Background and Hypotheses
Differences among Online Customer-Initiated Marketing Actions
Customer-initiated actions (CICs) differ from Firm-initiated actions in that they
require (potential) customers to actively search for and/or engage in online conversations about
the firm’s offerings (Gartner 2008, Hoffman and Fodor 2010; Wiesel et al. 2010). For instance,
emails sent by firms to consumers are regarded as firm-initiated, while customer click-through
on paid search is regarded as customer-initiated. In terms of spending, customer-initiated search
took 47.1% of $22.7 billion US online advertising market (EMarketer, 2010), while customerinitiated social media is expected to grow from 6% to 18% of the online marketing budget in
the next 5 years (Alinean 2011). Despite this current size and growth projections, sales and
profit effects have yet to be extensively quantified, with Shankar (2008) singling out “spending
on unmeasured media such as search marketing” as an emerging research area.
Paid customer-initiated media include affiliate marketing and paid search. Affiliate
marketing involves merchants sharing a percentage of the revenue when a customer arrives at
the company’s website by clicking the content in the merchant’s website (Gallaugher, Auger
and Barnir, 2001). Already 11 years ago, Hoffman and Novak (2000) found a low effectiveness
of online banner ads, and proposed affiliate marketing as a more efficient way of customer
acquisition. More recently, paid search has gained popularity, with US companies spending
more than 40% of the total online advertising dollars for paid search (Animesh et al. 2010). In
paid search like Google’s AdWords, advertisers bid for to be in a place closer to the top in the
listing of the paid search results which are displayed on the top or side of organic search results.
Owned media includes the online assets owned by the company, such as its websites
and their search engine optimization qualities. The strength of owned media shows up in the
company’s ranking in organic search (Yang and Ghose 2010), and in the amount of ‘direct
visits’, i.e. visitors that type the company’s name directly into the URL. Such ‘type-in’ traffic
may include loyal, repeat customers and late-stage buyers who have already visited the site
through other means but needed time to make the purchase decision (Bustos 2008).
Earned (social) media for a brand is created, initiated, circulated and used by
consumers (Blackshaw & Nazzaro, 2006). Social media activities include blogging,
microblogging (e.g. Twitter), cocreation, social bookmarking, forums and discussion boards,
product reviews, social networks (e.g. Facebook) and video- and photosharing (Hoffman and
Fodor 2010). Consumers are motivated to participate in these activities due to their desire to
connect, create, control and consume (ibid). Foresee Results (2010) reports that, while search
brings more potential customers to company’s websites, purchase conversion rates are higher
for visitors coming from social media. Recently, the medium has drawn criticism given the
poor sales results of Burger King and Pepsi, despite social media campaigns that scored in
terms of traffic and engagement (Baskin 2011). However, one needs to control for other sales
drivers: for a brand losing share due to higher prices and lower distribution, Pauwels et al
(2010) found that earned media (brand engagement on Facebook) helped stem these losses.
Conditions that favor paid, owned and earned media effectiveness
Brand strength: Better known brands versus lesser known brands
From an economic perspective, consumers search or sample as long as the marginal benefits of
doing so outweigh the costs (Ekelund et al., 1995). Known brands have rich associations in
consumer minds (Keller 1993) and thus carry less perceived risk and less costs for search. As
they are more easily accessible in consumer memory, known brands are more likely to obtain
direct traffic to their site (Bustos 2008). Likewise, search engines have included authority and
trust into their organic search ranking algorithms, making it easier for known brands to raise to
the top of the organic search listings (Crandall 2009). Because the organic search rank is a
strong predictor of both click-through and sales conversion (Yang and Ghose 2010), known
brands have the advantage in organic search effectiveness. In this context, paid search gives
better marginal opportunities for lesser known brands, such as Inofec (Wiesel et al. 2010) to
reach potential customers. Based on an engaging title and message, paid search links are
evaluated higher than organic search links by users in terms of relevance (Jansen 2007). By the
same token, affiliate marketing with a popular merchant should be more beneficial to lesser
known brands. In contrast, several known brands have cut back on affiliate marketing, as they
believe it cannibalizes site traffic they would have obtained anyway (NewMediaAge 2010).
H1: Owned media (organic search, direct visits) are more effective for better known brands
than for lesser known brands
H2: Paid media (affiliate marketing, paid search) are more effective for lesser known brands
than for better known brands.
As for social media effectiveness, brand strength appears a double-edged sword. On
the one hand, known brands have an easier time gathering a large social media following
(Alinean 2011). On the other hand, the performance benefits of such following are unclear:
does social media really generates additional sales for these brands? As one example, Burger
King has been receiving several social media ad awards (eg for ‘Subservient Chicken’ and
‘Facebook Whopper Sacrifice’), but fails to see sales increase (Ries 2010). In contrast, lesser
known brands consider social media as a low-investment tool with great potential payoffs. The
CMO Survey (2011) showed that smaller companies devote the largest part of their budget to
social media. Examples of successful media campaigns by lesser known brands include
“Destroy Your Printer” by Expert Laser Services (Mueller, 2011) and Crème Brulee Cart in
San Francisco (Rao, 2010). Thus, whether better or lesser known brands achieve higher sales
returns from social media actions remains an empirical question we aim to investigate.
Product category: search versus experience goods
Consumers can evaluate the quality of search goods prior to purchase, but can only determine
the quality of experience goods after purchase (Nelson 1970). This higher uncertainty makes
consumers more likely to search for additional information for experience goods (Nelson,
1974). In the online world, consumers spend more time evaluating experience goods and favor
interactive mechanisms (Huang et al. 2009). Moreover, consumers should prefer peer-based
recommendations for experience goods, but rule-based recommendations for search goods
(Agarwal and Vaidyanathan 2003). While Agarwal and Vaidyanathan (2003) failed to find a
preference for collaborative filtering for experience goods, they attribute this to the low source
credibility of virtual recommendation agents. In contrast, today’s social media is among the
most trusted information sources (Hoffman and Fodor 2010) and the shared usage experiences
of other consumers should matter especially when deciding to purchase experience goods.
H3: Earned media is more effective for experience goods than for search goods.
Combining our hypotheses with informed speculation yields the 2x2 matrix in Table 1, which
offers specific predictions for managers in each cell, and helps putting the few previous
findings into context. For instance, Wiesel et al. (2010) report the rather high elasticity of 4.35
for paid search. The company in question is Inofec, a supplier of office furniture (search good)
which does not invest in brand building (low brand strength). Given the above hypotheses, paid
search should be especially beneficial in this situation, and different brand/category
combinations are likely to reflect systematically different effectiveness. Moreover, our
framework suggests that earned media should be more effective than owned media for this
company, and thus receive a higher priority in future experiments.
3. Methodology and preliminary findings
Bayesian Vector Autoregressive Models
Dynamic system models such as Vector Autoregression (VAR) have been a popular tool to
analyze both short-term and long-term marketing effectiveness for offline activities ranging
from new product introductions to price promotions, distribution and communication
(Bronnenberg et al. 2000; Dekimpe and Hanssens 1999; Pauwels et al. 2004). Moreover, VARmodels appear especially relevant in an online context, given the multiple touchpoints brands
have with consumers over time (Trusov et al. 2009; Wiesel et al. 2010). In addition to such
‘dynamic synergies’, we also allow for same-period synergies by adding the interaction terms
among paid, owned and earned media. For one, sales may increase when potential customers
are exposed to both paid and organic search listings for the brand (Yang and Ghose 2010).
Typical issues with VAR-models include overparametrization and wrongly-signed
coefficients (Ramos 2003). We can address both through shrinkage, which imposes restrictions
on the parameters of the VAR model. Such Bayesian Vector Autoregressive (BVAR) models
are formulated in Litterman (1986) and Doan, Litterman and Sims (1984), but have seen little
application in marketing (Ramos 2003). Among the possible priors, the Minnesota (Litterman)
prior has the advantage of a simple posterior inference involving only the normal distribution
(Koop and Korobilis, 2010). Using Doan et al.’s (1984) formula for the uncertainty of the
Minnesota prior means, we can specify individual prior variances for a large number of
coefficients in the model using only a few parameters (LeSage, 1999). These parameters
and w(i,j) represent the overall tightness, lag decay and the weighting matrix respectively.
We estimate the BVAR model through the “mixed estimation” technique
developed by Theil and Goldberger (1961). This method involves supplementing data with
prior information on the distributions of the coefficients (Ramos, 2003). A typical
unrestricted VAR with n endogenous variables and p lags can be written as:
Focusing on a single equation of the model:
is the vector of observations on
, the matrix X represent the lagged values of
and the deterministic components, the vector A stands for the coefficients of
the lagged variables and deterministic components and
is the residual vector. Prior
restrictions for this single equation model can be written as:
Where prior mean
prior, and
is the prior mean and
is the standard deviation of the Minnesota
. Using Theil and Goldberger (1961), we rewrite equation (3) as:
Then, we can find the estimator for a typical equation by using the following formula:
In order to find the optimum values for the parameters ,
and w, we minimize the log
determinant of the sample covariance matrix of the one-step-ahead forecast errors for all the
equations of the BVAR (Doan et al. 1984).
Preliminary findings
We applied the above model to analyze the effectiveness of marketing actions by Knewton,
a relatively new player (since 2008) in the GMAT test preparation market. Instead of the
traditional face-to-face interaction, Knewton uses adaptive learning software to provide
maximum flexibility and individual customization to each student’s progress (Di Meglio,
2009). The experience nature of the category is evident: customers will only be able to
evaluate the quality of this service after they have experienced it. This situation thus
corresponds to the unknown brand/experience good quadrant in Table 1. Consequently, our
prediction is that social and paid media would be more effective than owned media.
The available marketing data include price, paid search, organic visits, direct
visits and online banner ads (i.e. we miss earned media actions). Our dependent variable is
the number of customers that sign up for the service. Consistent with our hypotheses, we
find that paid search has a higher sign-up elasiticity (2.32) as compared to owned media, in
the form of organic search (0.56) and direct visits (0.97). Online banner ads have the lowest
elasticity; from 0.004 on a general interest site (Amazon) to 0.11 on a special interest site
(US News, which provides annual MBA school rankings).
4. Support Needs and Time Table
While our pilot study gave preliminary support to some of our hypotheses, we need data in
the remaining quadrants of our framework. Especially helpful are data on customer-initiated
activity by known brands, many of which are MSI members. Therefore, we appreciate any
support in identifying and contacting data providers. In many cases, additional information
is needed, and funding will enable us to buy such data, to clean it and get it ready for
analysis. The Bayesian VAR estimation also involves substantial efforts in program coding
and debugging. Last but certainly not least, we are looking forward to the feedback and
advice of the expert review panel on this proposal. Table 2 below details this timetable and
summarizes our support needs.
5. Expected Outcomes and their Relation to MSI call priorities
The results of these proposed studies are expected to shed new light on how the elasticity of
owned, paid and earned media differs according to brand strength and product category
dimensions. We aim to address MSI’s challenge to provide insights on how to manage an
expanded marketing communications mix that includes multiple media and channels, and
rich marketing communications. While the effectiveness of traditional firm-initiated
communications is relatively well understood, the sales and profits effects of new customerinitiated media are much debated. This proposal touches on all three key areas of the MSI
call, in particular the questions: (1) How does the impact of advertising vary by context?, (2)
How can firms compare the effectiveness of diverse marketing communication activities?,
and (3) What is the role of the brand in an environment characterized by the emergence of
social media and complex, multiple touch points between brands and customers?
Table 1: Conceptual framework and hypotheses
Search Good
Experience Good
Owned > Paid, Earned
Earned > Owned > Paid
Example: to be collected
Example: to be collected
Paid > Owned, Earned
Earned > Paid > Owned
Example: Inofec office furniture
Example: Knewton test preparation
Known brand
Unknown brand
Table 2: Support Needs and Time Table
Budget Item
Time of completion
1 Complementary data collection
September 15, 2011
$ 3,000
$ 3,000
2 Program coding and debugging
200 hours @ $25 per hour
October 15, 2011
$ 5,000
$ 8,000
December 1, 2011
$ 10,000
$ 18,000
3 Data analysis and verification results
400 hours @ $25 per hour
Total funding requested from MSI
$ 18, 000
Target completion interim report
December 15, 2011
Target completion date for paper
March 15, 2012
Agarwal P. and R. Vaidyanathan (2003), “The Perceived Effectiveness of Virtual Shopping
Agnts for Search vs. Experience Goods”, in Advances in Consumer Research vol. 30, eds. P.A.
Keller and D. W. Rook, Valdosta, GA : Association for Consumer Research, 347-348
Alinean Inc. (2011) "How Do You Calculate the ROI from Social Media Marketing?"
_Marketing.pdf; accessed March 27th 2011.
Animesh, A., Ramachandran, V. and Viswanathan, S. (2010), “Quality Uncertainty and the
Performance of Online Sponsored Search Markets: An Empirical Investigation”, Information
Systems Research, 21 (1).
Baskin, J. (2011), “Do Campaign Failures, High-Profile Firings Signal the End of Social
Media?”, AdAge, March 22nd, , accessed March 26th, 2011.
Blackshaw, P., and Nazzaro, M. (2006), Consumer-generated Media (CGM) 101: Word of
Mouth in the Age of the Web-Fortiļ¬ed Consumer, New York: Nielsen BuzzMetrics.
Bronnenberg B., V. Mahajan, W. Vanhonacker. 2000. “The Emergence of Market Structure
in New Repeat-Purchase Categories: A Dynamic Approach and an Empirical Application”,
Journal of Marketing Research 37(1) 16-31.
Bustos, Linda (2008), “The Forgotten Metric: Direct Traffic Reveals Brand Strength”,
WebAnalytics, July 31st,, accessed
March 27th 2011.
Down”,, accessed March 27th.
Dekimpe, M. G., D. M. Hanssens (1999), “Sustained Spending and Persistent Response: A
New Look at Long-term Marketing Profitability,” Journal of Marketing Research 36(4)
De Salvo, N. (2011), “Big Brand, Small Brand – The Great Social Media Equalizer”, Ezine,,-Small-Brand---The-Great-Social-MediaEqualizer&id=3821845 , accessed March 27th 2011.
Di Meglio, F. (2009), “GMAT Test Prep: A User's Guide.”, BusinessWeek Online. - August
8, p. 9.
Doan, Thomas, Robert Litterman, and Christopher Sims (1984), “Forecasting and
Conditional Projection Using Realistic Prior Distributions,” Econometric Reviews, 3, 1-100.
Ekelund, R. B. Jr. , Mixon, F.G. Jr., and Ressler, R. W. (1995). “Advertising and
information: an empirical study of search, experience and credence goods”, Journal of
Economic Studies, 22 (2), 33-43.
at: .
Gallaugher, J. M, Auger P. and Barnir, A. (2001), “Revenue Streams and Digital Content
Providers: An Empirical Investigation”, Information and Management, 38 (7), 473-485.
Gartner, Inc (2008), A Checklist for Evaluating an Inbound and Outbound Multichannel
Campaign Management Application. Report ID Number: G00160776.
Hoffman, D.L. and Novak, T. P. (2000), “How to Acquire Customers on the Web”, Harvard
Business Review, May-June, 78 (3), 179-183.
Hoffman D.L. and M. Fodor (2010) "Can You Measure the ROI of Your Social Media
Marketing?" MIT Sloan Management Review, 52 (1), 41-49.
Huang, P., Luries, N.H., and Mitra, S. (2009), “Searching for Experience on the Web: An
Empirical Examination of Consumer Behavior for Search and Experience Goods”, Journal of
Marketing, Vol.73, p. 55-69.
Jansen, B. (2007), “The comparative effectiveness of sponsored and non-sponsored links for
web e-commerce queries”, ACM Transactions on the Web, 1 (3) 1-25.
Keller, K.L. (1993) “Conceptualizing, Measuring and Managing Customer-Based Brand
Equity”, Journal of Marketing, 57(1), 1-22.
Koop, G. and D. Korobilis (2010), “Bayesian Multivariate Time Series Methods for
Empirical Macroeconomics,” Foundations and Trends in Econometrics: 3 (4), 267-358.
LeSage, J.P. (1999), Applied Econometrics Using Matlab, available at , accessed March 27th 2011.
Litterman, R. (1986), “Forecasting with Bayesian Vector Autoregressions-Five Years of
Experience,” Journal of Business and Economic Statistics, 4, 25-38.
Manchanda, P., J.P. Dube, K.Y. Goh and P. Chintaguntat (2006), “The Effect of Banner
Advertising on Internet Purchasing”, Journal of Marketing Research, 43(1), 98-108.
Mueller, M.P. (2011), “Taking a Crack at Social Media”, accessed March 27th 2011:
Nelson, P. (1970), “Information and Consumer Behavior”, Journal of Political Economy,
78, 311-29.
------------ (1974), “Advertising as Information”, Journal of Political Economy, 82, 729-54
New Media Age (2010), Orange discovers ads’ effect on sales beyond last click, April 8th,
Pauwels, K., S. Srinivasan, O. Rutz and R. Bucklin (2010), “Are Online Metrics Leading
Indicators of Brand Performance?”, Presentation at the 2nd Google/WPP Awards, New York,
November 9th.
Pauwels, K., J. Silva-Risso, S. Srinivasan, and D.M. Hanssens (2004), “New Products, Sales
Promotions, and Firm Value: The Case of the Automobile Industry,” Journal of Marketing,
68 (October), 142–56.
Rao, L. (2010), How Social Media Drives New Business: Six Case Studies, available at:
Ries, A. (2011), “Social Media will Usher a Golden Age of Global Marketing – if marketers
get the message right”, AdAge Blogs, January 11th,, accessed March 27th 2011.
Ramos, F. F. (2003), “Forecasts of Market Shares from VAR and BVAR Models: A
Comparison of their Accuracy,” International Journal of Forecasting, 19, 95-110.
Shankar, V. (2008), “Strategic Allocation of Marketing Resources: Methods and Insights”,
in: Marketing Mix Resource Allocation and Planning: New Perspectives and Practices,
Roger Kerin and Rob O’Regan, eds., American Marketing Association, 154-183.
The CMO Survey (2011), Accessed March 27th 2011 at:
Theil, H. and A. S. Goldberger (1961), “On Pure and Mixed Statistical Estimation in
Economics,” International Economic Review, 2, 65-78.
Trusov, M., R. Bucklin, K. H. Pauwels. 2009, “Effects of Word of Mouth versus Traditional
Marketing: Findings for an Internet Social Networking Site”, Journal of Marketing, 73(5),
Wiesel, T., K. Pauwels, and J. Arts (2010), “Marketing’s Profit Impact: Quantifying Online
and Offline Funnel Progression,” Marketing Science, forthcoming, published online in
Articles in Advance, December 30th, DOI: 10.1287/mksc.1100.0612.
Yang, S. and A. Ghose (2010), “Analyzing the Relationship Between Organic and
Sponsored Search Advertising: Positive, Negative or Zero Interdependence?”, Marketing
Science, 29 (4), 602-623.