For the tourism and hospitality industry, the importance of social media analytics can hardly be understated, according to a new report.
“The most successful brands will be those that embrace and learn to harness social media rather than underestimate or fight against its influence,” wrote Deloitte LLP consultant Robert Bryant.
But megavendors such as SAS Institute and IBM are betting that the tourism and hospitality industry isn’t the only one that can benefit from social media analytics. Both vendors have turned their attention to mining data in blogs, wikis and popular social networking sites like Facebook and Twitter to uncover valuable insights for companies in a number of industries.
Some analysts and industry watchers are wary, however, and question the accuracy and, ultimately, usefulness of the underlying analytics technology as it currently stands.
In April, for example, SAS Institute unveiled a social media analytics suite to help its retail, finance, pharmaceutical and customers in other industries understand the effect that online discussions are having on their corporate brands and design more effective marketing campaigns.
Last summer, IBM acquired SPSS, whose text analytics and sentiment analysis software is seen by some as the key to accurate social media analytics. In May, the company released the latest iteration of SPSS Modeler, which includes more than 180 industry-specific taxonomies designed to recognize language found in blog posts and Tweets that are specific to various vertical markets.
Even social networking sites themselves are getting into the data analytics game. In June, Twitter acquired Smallthought Systems, a start-up Web analytics firm. Also last month, Facebook released a newly enhanced version of its own analytics tool, Insights Dashboard, which provides Facebook page owners metrics around their content.
“By understanding and analyzing trends within user growth and demographics, consumption of content and creation of content, page owners and platform developers are better equipped to improve their business with Facebook,” Alex Himel, a software engineer at the social networking site, said in an email interview. “They can use this information to fine-tune their pages to increase engagement or further enhance applications.”
And, as the vendors tell it, customers are having success.
IBM says RTL Nederland, the Dutch TV production company, has been using social media analytics technology to understand what viewers are saying about its So You Think You Can Dance? program on Facebook and Twitter. Turns out viewers were less than thrilled with the program’s voting procedures, so RTL Nederland made a change to the show’s format.
Marcus Hearne, who oversees product marketing for SPSS, said recent advances in sentiment analysis technology have enabled companies like RTL not just to monitor what is being said about them in the blogosphere, but also to take action based on those insights.
Ten years ago, Hearne said, the only thing text analytics technology could do was identify individual words, but it provided little context. And it was rarely, if ever, applied to the burgeoning blogosphere.
But with the advent of sentiment analysis, “We’ve gone from counting words to understanding the feeling behind what people are saying [on blogs and on social networking sites]” Hearne said.
But not everyone is convinced. Consultant Katie Paine, chief executive of KDPaine and Partners LLC in Berlin, N.H., said most social media analytics technologies still do a “terrible” job at determining the sentiment of Facebook posts, Tweets and other online discussions.
“It’s very easy to collect this stuff,” Paine said of social media-related data. “[But] it’s extraordinarily difficult to make anything meaningful out of it because 93% of it is drivel or irrelevant.”
Paine cited a recent experience she had with SAS. The company used its own technology to gather and analyze social media about itself. It collected around 3,500 pieces of social media content, then whittled them down to 250 or so that were actually referring to SAS the company and not typos or other entities.
Using its sentiment analysis technology, SAS then tried to determine the sentiment of each post – positive or negative. It was able to definitively call a post positive or negative in only about 50 of the cases, Paine said. The sentiment analysis engine tagged around 200 pieces of content as “neutral,” Paine said. While it’s likely that some truly were neutral, Paine said, [for many of the pieces], the technology simply couldn’t determine its sentiment so it reverted to the default “neutral” classification.
The result is that many companies still don’t trust sentiment analysis technology enough to rely on it to make important decisions.
“The technology just isn’t mature yet,” said Angela Chen, director of BI for the financial trading firm Liquidnet. But, Chen said, she thinks sentiment analysis technology has significant potential to help companies better understand their customers if it can be improved.
“I would say it would be a waste not to mine data wherever it is,” Chen said, referring to social media-related data. “I hope [vendors] will put more focus on enhancing this technology.”
Finding talented BI and IT pros to design and manage social media analytics projects is also a problem, according to James Kobielus, an analyst with Cambridge, Mass.-based Forrester Research Inc.
“These are not tools that are widely adopted or known or used by your traditional BI professionals,” Kobielus said. Those analytics pros with the required skills, therefore, “are going to fetch a pretty good price.”
Still, despite its weaknesses and high costs, sentiment analysis technology has clearly improved over the years -- even if vendors and analysts and customers can’t agree on how much it’s improved.
Likewise, social media analytics is likely to continue to grow in importance as the blogosphere itself continues to expand, meaning SAS and IBM may be onto something.
That might not necessarily be a good thing for customers, however.
“False positives and false negatives will become an even bigger issue as social media delivers more text into your advanced analytics applications,” Kobielus said. “It hasn’t been worked out by the linguists yet.”