While the number of users on Twitter, Facebook and other social-networking sites continues to grow, business intelligence practitioners remain sceptical about the value of knowledge such services could generate, if one survey by a data warehousing firm is any indication.
Kognitio surveyed 125 people on its sales contact list about the potential value of social-networking tools, in terms of providing raw data that could be analysed. Respondents were ambivalent about this possible new source of intelligence, however.
"We wanted to find out if the enterprise architects, and the people in the BI trenches, are bringing in and analysing social media data," says John Thompson, CEO of Kognitio's North American operations. "And for the most part, they aren't."
Only 14 percent of the respondents said they have any desire to incorporate social-networking data into their current data analysis efforts. A total of 63 percent of the respondents were "undecided" about the potential value of aggregated social-networking data, and 23 percent called social media "overrated."
Kognitio released the results of the survey at the National Retail Federation's annual conference, held in New York this month.
Thompson says while organisations have started using social-networking sites as marketing tools, less thought has been thus far dedicated to analysing feedback such sites could generate.
Thompson speculates that most BI practitioners are too busy refining the existing systems to look into new sources of raw data. Once upper management starts to see the strategic value in analysing the chattering of the many, however, then we might see more social media-based BI.
Such social media-based conversations, recommendations and other user-generated data could be worth investigating, through such tools as text mining, sentiment analysis and geo-location. "It would be interesting for retailers to look at certain trends and to get an idea of what people are thinking, doing and talking about," he says.
A large social-networking site can provide a snapshot of what people are talking about in near real-time, Thompson says. A retailer could track if a certain brand name is being talked about, and watch the buzz as it moves around to different regions. It can also summarise whether positive or negative things are being said about the brand, through a keyword-based technique called sentiment analysis.
Setting up a social-networking monitoring feed within an existing BI system shouldn't be that difficult, Thompson says. Social-networking data is unstructured, so it is not organised in a formally defined database. By now, most BI tool vendors and service providers have incorporated some means of working with unstructured data.
BI providers have also streamlined operations so that analysis can be generated in near real-time and displayed on dashboards and widgets. However, Thompson did caution that Twitter limits the number of times in a day a search call can be made to the service.
On the show floor of the National Retail Federation conference, some vendors were showing off how their BI tools could process and make sense of social-media data. SAP spokespeople, for instance, talked up the ability of Business Objects software to aggregate, highlight and analyse Twitter data. The software could offer a dashboard detailing how much discussion a particular brand name generated, as well as how much of the conversation about the brand was positive or negative.
Jon Würfl, an SAP retail industry principal, noted that one of the company's customers already uses this setup to monitor Twitter. When the company, which Würfl did not name, found itself to be the subject of some negative talk, it was able to pinpoint the talk to a particular region and determine that the source of the negativity came from one badly-managed store.
The Kognitio survey also asked some questions about the use of BI in general. The company found that 31 percent of respondents planned to purchase new BI capabilities in the upcoming year, and 36 percent said that when setting up a new BI system, they must test, evaluate and deploy the system within a few weeks, far less time than management allowed in previous years.