AT&T logo

Roughly one out of every five status updates on Twitter mentions a brand or product, which makes it a great platform for gauging consumer sentiment about existing products, and potential launches. Often, that sentiment is favorable – but even when it’s not, companies don’t have to be afraid. As AT&T shows, brands can mine Twitter for negative sentiment, and use those insights to help solve customer service problems in “real-time.”

AT&T’s track record when it comes to customer service in the age of real-time, social communications isn’t exactly stellar. (Legal threats to disgruntled customers via email, anyone?) But the company is redeeming itself on a variety of fronts, including a new effort to monitor Twitter for real-time info on dropped calls and improve the network accordingly.

MIT Technology Review reports that AT&T has developed software that mines Twitter for complaints about dropped calls and other network problems. The program extracts details like the time the tweet was sent, the location of its user, as well as keywords like “3G” or “call dropped.”

The company then uses that information to prioritize repairs as close to “real time” as possible. In some cases, the Twitter data even allows AT&T to find and solve network issues before subscribers have to call in to report them.

Of course, AT&T has millions of dollars to devote to research & development (R&D) for projects like this. But even if your company doesn’t have the resources to undertake “Twitter mining” for customer service on a massive scale, there are some tips to be had from AT&T’s experience:

  • Identify exactly what information you need to track

“We are trying to identify three pieces of information,” Jia Wang, a
staff researcher at AT&T’s in-house research division, tells
Technology Review. “Where the
customer experienced problems, what type of problem, and when they
experienced it.”

Your company’s key information cues may vary, but since Twitter produces a never-ending stream of content, it is extremely important to filter out unnecessary data. What are the keywords you should be targeting? If you’re a UK-based company that does little to no business in the US, should you exclude US-based tweets from your search?

Figuring out what to track and what not to track in terms of complaints will be easier if you’ve already been using Twitter as a listening tool, but it’s never too early to start.

  • Use the right tool(s) to gather the data

AT&T built its own program to aggregate, segment and analyze Twitter data, and Twitter’s application programming interfaces (APIs) make that data readily accessible. But you don’t need to have a developer on staff – or even hire one – to build a custom platform. Vendors including Attensity, OpenAmplify and others offer varying degrees of social media-centric data analysis, often via the software-as-a-service (SaaS) model.

  • Analyze the data in conjunction with existing sources

Don’t leave the Twitter data in a vacuum. AT&T compares negative sentiment on Twitter with its own system logs, customer service calls, as well as data from Mark the Spot, an iPhone app that lets users submit feedback about exactly where they’re experiencing network problems. 

Your company likely has a number of its own channels for feedback – be it email, a submission form on the website, a Facebook page, or even a brick-and-mortar location. Compile data from these sources with your Twitter mining results to develop a more accurate picture of what your primary customer service issues are (and how to fix them).