Thus, we first investigate whether blogging activity leads to (differential) market outcomes. We then examine whether managerial communication (magazines advertising) and blogging are synergistic. We assemble a unique data set from fashion containing market outcomes (sales), new media (blogs) and traditional media (magazines advertising) for a brand of clothing, and a brand of shoes. Each category has at least one product launch during the duration of our sample periods.
We specify a simultaneous equation log-linear system for market outcomes and the volume of blogs. Our results suggest that blogs are predictive of market outcomes, new and traditional media act synergistically, pre-launch magazines advertising spurs blogging activity but become less effective post-launch and that market outcomes have some effect on blogging. We find detailed support for some of these findings via a unique and novel text mining analysis. We discuss the managerial implications of our findings.
I-Introduction Consumer generated media (CGM) such as blogs (a contraction of the term “Web logs”) have witnessed explosive growth in the last few years. For example, the number of blogs worldwide is estimated to be 184 millions with a readership of 346 million (March 2010). In contrast, in March 2003, the number of blogs was essentially zero. Other types of CGM have also seen similar growth patterns, e. g. , Facebook, which started in February 2004, now has about 400 million members worldwide (February 2011).
There are also indications that blogs are now being seen as similar to mainstream media sites – the number of blog sites in the top 100 most popular sites (blogs and mainstream media) worldwide was twenty-two in 2008 and blogs were being viewed by consumers as “sites for news, information, gossip etc. ” (2008). In 2010, four of the top ten entertainment sites were blogs (March 2010). It is clear from these statistics that there is considerable activity (multi- media posting, blogging, visits, traffic etc. ) on the part of consumers.
However, an important question, from a managerial perspective, is whether this activity leads to (differential) business outcomes such as sales or profits. In addition, little is known about the relationship between traditional or old media (where the company creates content and delivers it to consumers) and consumer generated, or new, media (where consumers create content and there in an exchange of this content between other consumers and potentially, the company). That is, are there any synergies between new media and old media?
In this research, we take the first step towards answering these questions. Blogging is perhaps the most established and largest form of consumer generated media at this point in time. The total worldwide viewership of blogs is estimated to be about 346 million (March 2010). Wikipedia defines as a blog as “a Web site, usually maintained by an individual with regular entries of commentary, descriptions of events, or other material such as graphics or video. Entries are commonly displayed in reverse-chronological order. Blogging is a worldwide phenomenon with the two biggest blogging markets being the United States and Japan. The number of blogs in the United States is about 23 million (about 12% of all US Internet users) and about 8 million in Japan (about 5% of all Japanese Internet users) in 2009. However, if one examines the total number of posts by language, Japanese language posts account for 37% of all posts worldwide followed closely by English language posts at 36%. Finally, readership of blogs in these two markets is ery high – about half of all Internet users in the US and about one-fifth of all Japanese Internet users have read a blog in the past year. While there are many informal opinions on the effectiveness of CGM in general (and blogs in particular) vis-a-vis market outcomes, there is limited empirical research that sheds light on this issue, especially for the launch of new products. The majority of the existing research has focused on online chatter (newsgroup postings, reviews and ratings) and its effect on market outcomes. There is some evidence that volume of online user ratings is positively correlated to sales.
Blogging, on the other hand, has been seen as a unique type of user generated content as being a highly personal, non-directed communication tool. As Kumar (2005) note, blogs are unique for sociological reasons – they comprise a “highly dynamic, temporal community structure” that “focuses heavily on local community interactions” – and for technical reasons – blogs “offer us a ready-made view of evolution (of content) in continuous time. ” In addition, blogging activity was probably the most pervasive CGM activity on the web during the time of our data.
Given these unique characteristics of blogs as opposed to reviews, it is not obvious that bloggers’ activity should affect market outcomes. Surprisingly, there is very little research that has tried to quantify the effect of blogs on market outcomes, especially in the presence of traditional media and/or an examination of pre- and post-launch changes in the role of old and new media. Two recent empirical papers have focused on blogs and market outcomes. Dhar and Chang (2009) explore the relationship between music album sales (imputed via sales ranks on Amazon. om) and online chatter (as seen in blogs and on social networks). Using 108 music albums in early 2007 (before four weeks and after four weeks of their release), they find a positive correlation between both the number of blogs and Myspace member intensity with future music sales. Gruhl (2005) propose a new methodology to automatically generate a query of blog keywords to detect spikes in Amazon. com’s book sales rank. They conclude that their new algorithm could adequately predict the changes and spikes of future sales ranks.
Thus, while these two studies suggest that there may be a correlation between blogging activity and market outcomes, they do not use actual sales data but only sales ranks from Amazon. com. To the best of our knowledge, the second issue that we outline above – the positive relationship between traditional media and new media – has not been investigated in the literature. Our expectation is that there will be a positive correlation between the quantity of traditional media and new media as traditional media is likely to provide discussion materials for bloggers.
From a managerial perspective this issue is crucial, as managers have no direct control over CGM (blogs in our case). However, if there is indeed a synergistic relationship between traditional media, which are under managerial control, and new media, which are outside managerial control, then managers can leverage this relationship. Specifically, they can carry out “better” resource allocation and media planning (to traditional media) as they can take the spillover effect (from traditional to new media) into consideration.
We examine the role of new media with respect to market outcomes as well as the relationship between new media and traditional media using data of two different clothing and shoes brand that are both promoted in fashion blogs. We consider the number of units sold, customers or subscribers (all a proxy for demand) as market outcomes, blogs as representations of consumer generated media and magazines advertising as traditional media. We specify a simultaneous equation model that links sales to advertising and blogs as well as a model that links blogs to advertising.
Our results, after controlling for many temporal and cross-sectional factors, suggest that first, the volume of Blogstock (cumulative sum of past blog posts) is positively correlated with market outcomes (volume of clothing sold, and the volume of shoes sold) post launch. Second, the interaction between blogs and magazines advertising has a positive effect on market outcomes. Third, we also find that traditional media (magazines advertising) positively affects new media (the volume of blogs) pre launch. In other words, bloggers consume advertising, independent of the product, and this ncreases their blogging activity. Finally, we find that the effect of blogs varies between pre and post launch. In general, the positive relationship between magazines advertising and the volume of blogs pre-launch becomes weaker after launch. This result suggests that while magazines advertising can independently increase blogging pre-launch via the provision of information and content, post-launch (i. e. , once the product is available), consumers may rely less on traditional media, leading to a much weaker relationship between new and old media at that point.
These last three sets of results shed light on the possibility that, broadly speaking, advertising and blogs act synergistically (with the relationship changing somewhat post-launch). The process explanations for our findings is not obvious. We take the first step in eliciting process explanations by carrying out a novel text mining analysis of the blog posts for the two markets (shoes and clothing) for which we have access to the textual content data. The findings from the text mining analysis suggest that blogs may affect market outcomes as they represent a rich source of product information and consumer opinion for other consumers.
Also, bloggers do use advertising as a subject for blogging pre-launch but turn their attention to product attributes post-launch. II-Data Our data come from fashion market. We consider data from two brands – clothing and shoes. We first describe the market outcome data for each product market and then we describe the measurement of traditional and new media. III-Market Outcomes The daily sales of clothings were made available for the total fashion market based on a nationally representative consumer panel.
The data include daily sales of two new pieces of clothing introduced in the period from January 2013 to March 2013. For shoes, the outcome variable we use is based on the same principle. We have data of two new models that were released (launched) in the period from January 2013 to March 2013. IV-Traditional Media The traditional marketing variable we use is magazines advertising. This was measured in units of daily or monthly Gross Rating Points (GRPs). There are some differences in the patterns of magazines advertising pre and post launch across the two brands.
For clothing, most of the advertising is post launch. Typically, commercial ads in this market begin to air about five days pre launch and then the heavier advertising kicks in post launch. In contrast, for shoes, pre-release magazines GRPs are larger (on average) than the post-release magazines GRPs. Specifically, peak advertising for shoes was, not surprisingly, a week before its launch date in order to generate high demand at the time of the opening. V-New Media We obtain blogging data from blog 1 (www. leblogdebetty. com) for clothing data and blog 2 (www. sorayabakhtiar. com) for the shoes data.
Both the brands scan and index the two blogging sites on a daily basis using keywords with coverage of about 64% of all blog articles. They then aggregate the data and provide the count of the daily number of blogs that mention a particular keyword on a specific temporal period such as day or month (multiple mentions in the same temporal unit are counted as one). As is typical for most blogs, its contents appear in a reversal chronological order and also include the blogger’s profile, “trackbacks” (links showing other websites, typically other blogs, that a blog is linked to), and comments.
Buzz Research archives the contents of all blog posts. It also carries out lexical analysis of the contents of each tracked blog by using a proprietary text- mining method and classifies each blog as positive, negative and/or neutral with respect to a given keyword. We therefore have access to the actual content of all posts as well as the daily percentage of positive, negative and neutral blogs for the movies and cellular phone service markets. There is big increase in the average number of blogs per period post launch in all two brands.
Interestingly, for the two brands markets where we have valence data, the biggest growth is in the percentage of neutral blogs post launch. To illustrate the relationship between marketing outcomes and both traditional and new media, we pick a product across our two brand markets. The figure suggests that magazines advertising, blog volume and shoes buyer are temporally correlated. Dividing the data temporally at the date of release we see that magazines GRPs and the number of blogs exhibit an increasing trend pre-release, but a decreasing one post-release.
While we illustrate a typical data pattern through this example, the pattern is not identical for all brands across product markets. In conclusion, these data are novel in the sense that they combine marketing data for both traditional and new media along with market outcomes from a market where new media have proven to be important (at least in terms of activity). Our data are also novel in the sense that they enable us to focus on new product launches. In addition, the fact that we have data from two different brand markets (frequently purchased consumer goods) with varying characteristics (e. . , more versus fewer new product launches) will help us determine if the relationship between market outcomes and new media as well as the relationship between new media and traditional media generalizes across product markets. Finally, the availability of the actual blog post text (for two categories) opens up the possibility to conduct a deeper text-mining analysis. VI-Managerial Implications So far, we have discussed the findings purely from a statistical point of view. However, it may be useful to translate these findings in a manner that uantifies the effect sizes from a managerial point of view. We therefore ran two experiments – the first to get a sense of how managers could change resource allocation and the second to see how managers could use blog data to improve sales forecasts. In the first experiment, we use the estimates from the clothing market data. To illustrate short-term effects, in the experiment, we assumed there were only three periods, two in the pre-release and one in the post-release. Recall that blogging is outside the control of managers.
We therefore used the marketing instrument under managerial control in our data set – traditional magazines advertising. In the experiment, we increased the Adstock by one percent in the first pre-release period. The output we measured was the percentage increase in the size of the daily volume sold in the post-release period. A ten percent increase in the Adstock results in a 3. 3 percent increase in the number of blogs at the second pre-release period. As a result of this increase in the Adstock, we find that the net increase in the sales volume is 2. 1 percent.
A decomposition of this overall increase due to traditional media versus new media suggested that the increase in the Adstock directly enhances the sales by 0. 13 percent while the interaction between blogging and advertising increases the sales by 0. 1 percent. Furthermore, the largest and most significant increase in the sales volume at post-launch is led by the indirect impact from advertising via blogging activity, which accounts for 1. 9 percent. Similar experiment for the other product markets also support these findings with the overall effect being slightly smaller for shoes (0. 4%). In addition to simulating the short-term effects of advertising, we use a simulation setting similar to the above experiments and expand the time horizon from one period to ten periods. The largest indirect effect of the ten percent increase in Adstock decays slower than do the other two effects across two product categories. The peaks of the indirect effects are located at the third period for the clothing and at the second period for the shoes. These are resulted from the larger estimates of the carry-over constants of
Adstock and Blogstock at post-launch in the blog equations. In the second experiment, we hold out the last observation from each brand and re-estimated the model. We then use the model estimates for prediction and computed the difference in the predicted value and the actual data across all the held out observations. We do this for the full model and a restricted version of the full model where the response coefficients for the number of blogs and the cumulative number of blogs were set to zero.
Thus, the difference in prediction (based on the Root Mean Square Deviation) between these two models shows the extent to which the use of blog data can improve sales forecasts. The improvement in RMSD is very high for shoes, and modest for clothing. VII-Conclusion, Limitations and Directions for Future Research This paper adds to the very limited, but rapidly growing field of research into the effectiveness of new media, especially in the case of new product launches.
Using a unique dataset from two product markets (a major new media market), we are able to combine into a single source, data on market outcomes, traditional media (magazines advertising) and new media (volume and content of blogs). We used a simultaneous equation model to capture the effect of new media on market outcomes and the effect of market outcomes on new media. While this in itself is somewhat novel, we were also able to include the major marketing activity (mgazines advertising) in both equations, both directly and via interactions.
Thus this allows us to investigate two open questions in this domain – (a) whether new media (blogging activity in our case) leads to (differential) market outcomes and (b) whether traditional marketing actions (i. e. , magazines advertising) and new media act synergistically. We also make a first attempt, to the best of our knowledge, to use the content of the blog posts to shed “process” light on our econometric findings via a careful and methodical text mining analysis.
Using data from clothing, and shoes brands, we find that patterns across the two categories showing clear linkages between traditional media, new media and market outcomes. In general, we find that cumulative blogs (Blogstock) are predictive of market outcomes, blogs and magazines advertising act synergistically, pre-launch advertising spurs blogging activity (that is predictive of marketing activity) but becomes less effective in inducing blogging activity post- launch and market outcomes also do have some effect on blogging activity.
Our text mining results provide additional support for some of these findings. From a managerial point of view, in the experiment using clothing estimation results, we find that a one percent increase in the traditional marketing instrument (magazines advertising) leads to a median increase in market outcomes of 0. 2%, with a majority of the increase coming from the increase in blogging activity generated by the advertising pre-launch.
Our analyses do also have a few limitations (driven mostly by the nature of the data). First, as noted earlier, the aggregate nature of our data makes it very hard to offer micro-level causal explanations of the effectiveness of new media and the synergistic relationship between new and traditional media. While our text mining analyses shed some light on our findings, it would be very beneficial to obtain datasets that link individual activity to market outcomes for a larger variety of new media.
Second, our measures of new media are at present limited to blog content – volume – and in two product markets, keywords and valence. ). Third, our model could be improved with the potential use of non-parametric models to model the effects of both old and new media and the associated interactions. Finally, our data do not contain information on all marketing instruments and hence we use proxies (such as lagged sales in the case of distribution). We hope that with better data, future research will be able to address these limitations.