This post is part of a series on corporate blogs based on data collected for my Masters thesis and it focuses on sentiment in blog memes.
All 1,553 consumer blog posts captured by the study were tested individually and the result was a statistically significant skew towards positive sentiment. The chart below shows that 887 or 57.1% percent of posts were demonstrated positive sentiment. A further 33.4% of posts expressed neutral sentiment and just 9.5% of posts were negative.
Pearson’s Chi-Square test was used to determine if the skew towards positive sentiment was statistically significant. It was based on the assumption that if there was no bias, the expected result would be an equal distribution of posts in all three sentiment categories. The test showed that the observed result was different enough from the expected result that it represented a statistically significant bias towards positive sentiment in the data.
A positive bias among consumers responding to corporate blog posts creates some tension with research that suggests consumers do not trust corporate blogs. But, as social media becomes increasingly distributed, this becomes an increasingly irrelevant debate. Although, even if it is the case, corporate blogs serve other purposes.
Corporations are taking social media seriously and there are already numerous small companies creating tools to monitor online chatter, including sentiment, across multiple platforms and most include tools for companies to interact in these streams as well.
Sentiment seems to be a particularly alluring area and one that currently appears to be reaching a wider audience, but its viability remains passionately debated. Demand stems from a view of social media as a channel for customer interaction and word-of-mouth. And it results in a broad cross-section of potential uses for companies and professional company watchers in marketing and PR, finance, law, media, consulting, academia and more.
The sentiment data for this study was produced by the Parnassus Group, a social media research firm, using a custom configuration of their Sentimine tool. All articles were given a single sentiment score at the document level without targeting any particular topic or entity such as companies, products or brands.
In the next post in this series, I’ll talk more about sentiment among early posts in memes and the most connected posts in memes, which were the primary variables of my study.
Posted on September 14, 2009
This is post is part of an ongoing series on corporate blogs based on data I collected for my Masters thesis. It’s been a while since I last posted in this series and when I was writing the last post I realized how difficult it was to describe any one cut of data in isolation. For these reasons, I wanted to provide a recap of the previous few posts on the size and structure of blog memes before moving on to my findings on sentiment.
In an ideal world, I would have liked to have created something more visual but here is the time-limited version:
- The top three companies, Google, HP, and Yahoo accounted for more than a third of the 389 blogs.
- The most active company, Google, accounted for almost half of memes by itself.
This little recap is a long way of saying that power laws were present in most cuts of the data. The presence of power laws in social networks has been known for some time and this data represents another example.
It makes analyzing the data for patterns more complex, but potentially more rewarding, than data that follows a normal (bell-curve) distribution and there are plenty of much smarter people working on ways to do that.
Posted on September 11, 2009
This is the eighth post in a series on corporate blog memes. This one focuses on the number of times distinct blogs that participated in memes by publishing blog posts that linked back to corporate blogs.
A total of 1,553 blog posts were collected from the 299 memes previously identified and these were found to represent 792 distinct blogs. The table below plots the frequency of posts per blog from 1 – 20. It shows that 626 (79%) of the 792 blogs appeared just once in memes. Another 103 blogs appeared in blog memes twice and overall the pattern mimics those seen in corporate posts, meme size and links in corporate blog memes.
Over 95% percent of blogs appeared in no more than three memes which would seem to indicate that there are very few blogs regularly responding, sharing or otherwise spreading the content of Fortune 500 companies. It also deflates the idea that some small group of heavily invested customers or power users are regularly spreading information about any given company.
However, there are a number of factors to consider. It was already shown that only 299 memes were produced from 5,887 corporate posts. In addition, 14 companies accounted for all 299 memes. It is more likely that the lack of responses is because most corporate blog posts don’t produce memes to begin with, at least not through blog posts.
At the time of data collection, the exclusion of blog comments due to complexity was considered a potentially significant omission from memes. In the year since, the rapid growth of shorter, quicker and more distributed forms of social media make the lack of fully-formed blog memes less surprising. It also makes the study of memes by medium less useful.
Confirming the existence of vocal customers or power users and measuring their effect will take a lot of further study across multiple forms of social media. It also assumes the development of suitable tools that are capable of capturing distributed social media keeps pace with changes in social media publishing.
The complete table of data includes outliers values not shown in the chart. Two blogs appeared 187 and 188 times across all memes. In both instances, these blogs posted a (usually) daily summary digest of links and frequently posts appeared in multiple blog memes. For example, a blog on search engines posted a daily digest of links that included several corporate blogs on search and these posts subsequently appeared in all associated memes.
Posted on August 5, 2009
This is the third post in quick succession and seventh overall in a series on Fortune 500 corporate blogs. This is the final post in the series on blog memes and focuses on which companies produced the most memes.
The chart below shows all 299 memes by company and it is a short list. All 299 of the memes I collected were attributed to just 14 out of the 76 Fortune 500 companies with corporate blogs.
The concentration among such a small group was surprising because this data was derived rather than actively collected. It was also a little disappointing because it immediately limited the extent to which any findings could be generalized across all companies and corporate blogs.
As the graph clearly shows, Google’s corporate blogs accounted for almost half of the memes by itself. Although, it is important to note that all corporate blogs associated with each of the 76 Fortune 50 companies found to have blogs were tracked.
At the time of data collection Google had 62 blogs, more than any other company. The next highest was Hewlett-Packard with 53 blogs followed by Yahoo! with 28. Although Google dominated blog memes, it was also part of a larger skew towards Internet and technology companies in general.
The pie chart below shows that almost two third of memes were produced by Internet Services and Retailing companies while related industries such as Computers/Office Equipment and Peripherals were also present.
This was a little less surprising than the company spread because the overall number of Fortune 500 blogs also skewed towards technology-related industries. The complete table of data for memes by Fortune industry sector is available here.
These findings affect the interpretation of all of the data on blog memes presented in this series. It has to be analyzed from the perspective that a small number of companies, clustered in technology-related industries and dominated specifically by Google were disproportionately represented in memes.
Similarly, it is also useful to remember that most memes were made up of just two posts connected to a corporate blog entry. All of the patterns and relationships in the data are interrelated and deriving meaning has to be done holistically. The previously mentioned and general limitations also apply.
Posted on July 24, 2009
This is a short post on the number of incoming links in corporate blog memes and represents part six in a series on Fortune 500 blogs. The data is based on the blog memes I collected using Blogpulse.
The chart below shows the frequency of incoming links in corporate blog memes, starting at zero. By definition, at least one post in every meme had no incoming links as all memes end somewhere. Since only memes containing at least two tiers were collected, all tier one posts had to have at least one incoming link to be included in the study.
Almost 61% of posts had no incoming links and a further 28% contained one incoming link meaning 89% of posts had no more than one link. The maximum number of incoming links found for any post was 17.
The same sharp downward trend found in previous charts was also evident in the number of incoming links per post. The space restriction makes the data points for higher numbers of links difficult to see but they’re not zeros, the complete tables of data is available here.
One particular point that is relevant to make again is that not all tier one posts were collected so this data represents a subset of complete blog memes. If all tier posts were collected the sharp downward trend would be even more accentuated between zero and one incoming link per post.
The low number of posts with higher numbers of incoming links suggests memes were “thin” in that they don’t spread widely at each tier. This was true but previous posts in this series have shown both that the majority of posts were tier one and tier two posts and most memes contained only two posts: one tier one post and one tier two post. As a result, it is hard to generalize the findings in this way.
The data presented here is subject to all of the limitations I’ve mentioned in previous posts as well as the general limitations of my study. It is also specifically subject to the comprehensiveness and accuracy of the Blogpulse search index.
Posted on July 23, 2009
My name is Phillip Baker and this is my personal blog about finding value in a world of free information.