News Stay informed about the latest enterprise technology news and product updates.

Sentiment analysis turns social media into customer intelligence

Sentiment analysis tools are creating a kind of customer intelligence that online marketers are finding beats traditional market research.

Intuit Inc. didn't expect any major customer pushback when it announced in 2008 that the new version of TurboTax...

would limit the number of tax returns per installation and add a modest charge for each additional tax return prepared.

The tax preparation software firm had gathered plenty of customer intelligence via polling and research before it decided on the changes. "Few customers were doing more than one or two returns with the software, which after all was designed for consumers, not professional preparers," said Brian Andrews, Intuit's vice president for customer experience and business excellence.

What Intuit didn't take into account, however, was the power of social media as a consumer soapbox. A few, very vocal customers posted furious messages on such blogs as Canadian Capitalist and forums such as Rekords Rekords. Reviews on the website gave the new version just one star. "It was clear our ability to effectively message our change wasn't working, so we quickly reversed our decision. We sacrificed millions in revenue in order to protect and foster our reputation as a customer-centric company," Andrews said.

Depending primarily on word of mouth to sell its products, Intuit has been a voice of the customer (VOC) pioneer since it was a two-person startup nearly 30 years ago, Andrews said. Now, however, "Word of mouth is even more important, because social media sites like Yelp, Facebook, and Amazon have given customers a megaphone."

Customer intelligence has pluses and minuses

That megaphone poses both a challenge and an opportunity to professionals who work with customer intelligence. On the positive side, it gives them access to a rich, constantly updated trove of customer insights. On the negative side, that good stuff is buried in vast quantities of irrelevant and dubiously sourced material. And as companies like Intuit have discovered, time is of the essence when it comes to zeroing in on and responding to negative posts and taking action before they go viral.

Intuit uses Radian6 Insights, one of a large and growing body of services that monitor social media for customer posts and analyze them for relevance and sentiment, then categorize the data by such factors as influence, sentiment, geographic location and topic. End users then can access the results and do further analysis via a so-called social dashboard.

Intuit is in good company. During the past few years, according to James Kobielus, a senior analyst at Forrester Research Inc., customer support, sales, marketing and reputation managers have been trawling the Web for answers to such questions as these: How do they feel about us and our products? Do they have a strong propensity to become or stay our customers or buy something? Initially this search was a time-consuming manual task that often produced inconsistent results.

Now, however, much of the grunt work has been eliminated by such service providers as BuzzMetrics (now part of The Nielsen Co.), Crimson Hexagon, Cymfony Inc., MediaMiser Ltd., NetBase Corp., Radian6 (now part of Inc.), Scout Labs (now part of Lithium Technologies Inc.), Synthesio and WiseWindow Inc.

Monitoring strategic sentiment for warning signs of negative PR

The customer intelligence provided by such services can be used to measure the success of ad campaigns while they are happening, measure customer loyalty, address complaints and improve consumers' experience, and identify product glitches and public relations problems -- ideally before they go viral or hit the mainstream press.

Canada Post Corp., for example, uses MediaMiser's hosted platform to analyze its media coverage, which last year amounted to about 12,000 articles, according to Anick Losier, director of media relations at the government-owned national mail service. This helps the organization respond more effectively to negative coverage, she said.

When several unflattering articles about Canada Post recently appeared in the press, Losier figured the likely source was "our very vocal union," which she said is a common instigator of negative coverage. After she used MediaMiser to analyze the articles, however, she discovered that it wasn't the union that had called the media this time around, but a member of Canada's parliament. She then alerted management to the need to talk with that particular MP.

Losier's group also uses the platform to check out such social media as Facebook and Twitter "every half-hour, every day," she said. This often gives them an earlier heads-up about customers' attitudes and where the media (other than traditional outlets like TV and newspapers) is going with a story, she added.

Tracking internal corporate customer data

Sentiment analysis of online media is catching on in the financial and investment sectors as well. Thomson Reuters Canada Ltd.'s Machine Readable News and Dow Jones & Company Inc.'s Lexicon services provide sentiment analysis of their own financial reporters' coverage. Deutsche Bank AG's Global Equity Quantitative Strategy Group uses the Thomson Reuters service to enhance the advice it gives institutional clients about portfolio management strategies. "Sentiment is only a tiny fraction of the data we use, which includes every trade and is on the order of gigabytes," said Rochester Cahan, vice president at the Deutsche Bank group. A recent simulation study by the group found that using the service significantly improved yearly stock portfolio returns, he added.

We sacrificed millions in revenue in order to protect and foster our reputation as a customer-centric company.

Brian Andrews, vice president, Intuit Inc.

Often, corporations get into sentiment analysis when one group of public relations, investment or brand managers signs up with a social media customer feedback service. The potential problem is that such deployments constitute information silos, with little or no sharing of insights with other organizations, said Bruce Temkin, a managing partner at the Temkin Group.

Such services, furthermore, and their software tools and the data they collect historically have remained separate from more traditional systems for collecting and analyzing customer data, such as call center platforms; customer relationship management, or CRM, software; surveys; market research; and customer emails.

For an enterprise to exploit sentiment analysis fully, "IT needs to take an enterprise view of text mining, and not treat social media content as an isolated thing but as one type of unstructured [customer feedback] data" to be collected, analyzed and disseminated to different groups according to their needs, Temkin said.

Professionals in customer intelligence need to be able to mesh customer feedback from social media with that from more traditional channels, such as call center systems, customer surveys and emails, and notes from salespeople in the field, Temkin said. Insights from sentiment analysis also should be shared by end users within and across business organizations.

In this way, technical support agents and brand managers, for example, could get fast access to the latest customer posts via the Web so they could respond to complaints quickly and fine-tune marketing campaigns. On the other hand, executives could get a high-level view of opinion trends to help them make more long-term strategic decisions.

Social Web feedback faster than market research

Businesses moving in this direction as part of VOC strategies recognize that the value of social Web customer feedback extends far beyond marketing and public relations. "If you're plant manager, you can capture negative feedback or an emerging problem that customers are having with your brand a week faster than by just using a market research service," said Gareth Herschel, a research director at Gartner Inc. Fixing the problem before another week's worth of products go out to stores "could mean tens or even hundreds of millions of dollars," he added.

In response to market demand for a more integrated and holistic approach to sentiment analysis, vendors are extending their platforms and services in several directions. Sentiment analysis and text mining software vendors like Clarabridge Inc., Attensity Corp., IBM's Cognos and SPSS groups, Autonomy Corp. PLC, RightNow Technologies Inc., and KANA Software Inc., whose platforms deal with the large masses of unstructured enterprise data, have been extending their reach aggressively into social media through partnerships and acquisitions. In addition, social media monitoring service providers are providing portals, workflow and collaborative tools that enable end users in separate departments to share insights and collaborate on responses.

Intuit, for example, uses Clarabridge for Enterprise to gather customer insights from internal and Web-based sources, Andrews said. "Tax season only lasts a hundred days, and we measure [customer attitudes to TurboTax] every day, using what we hear to drive improvements all the way to April 15. We don't have time to go through all the customer comments manually, so we need a sentiment analysis tool that we can rely on and trust."

Quality of social Web data a concern

Still, many companies have been slow to incorporate social media into such customer monitoring programs. According to a third-quarter 2010 Temkin Group survey of 114 companies with formal VOC programs, 22% of respondents are using social media, 35% are actively considering it, and the rest are not.

A major concern is with the quality of social-Web sentiment data, given the media's notoriously low signal-to-noise ratio. Many business leaders are addressing the issue by focusing on sites that provide direct feedback from customers, such as Facebook fan pages, review sites like Yelp and TripAdvisor, and online social communities they set up for their customers to interact and share tips.

Meanwhile, businesses are recognizing the value of social media, not only for monitoring customer sentiment but also for engaging with and responding to customers on a deeper and more immediate level than more traditional feedback mechanisms can provide.

For example, when a customer posts a complaint about one of iRobot Corp.'s vacuum cleaners on YouTube, RightNow's SmartSense cloud monitoring tool will automatically alert Maryellen Abreu, iRobot's global technical support team director. In addition to contacting that customer directly, iRobot is likely to post its response to the problem on YouTube, and perhaps initiate a discussion on its customer forum, she said. "We not only want customers to be able to interact with us on the channels of their choice, but to be able to respond to customers on those same channels," she added.

And ideally, respond to the negative feedback before it goes viral.

Elisabeth Horwitt is a Boston-based freelance journalist. Let us know what you think about the story; email [email protected].

Dig Deeper on Enterprise business intelligence software and big data