When a leading bank wanted to find out how it stacked up against competitors, it assumed customers would focus on lending terms and interest rates. To the bank’s surprise, the most enthusiastic discourse on blogs and specialized financial forums related to a smartphone app a competing financial institution had just put out.
The bank had dismissed apps as a generic marketing gimmick, like the old custom of giving away a toaster for opening an account. After learning how much customers valued the app, the bank quickly created its own with the same prized features as its competitor.
“If you know where to look, you can find out a lot of information about competitors online,” says Joseph Carrabis, founder of NextStage Evolution, who did the analysis for the bank. “You can get the benefits of corporate espionage without actually doing corporate espionage.”
Customer sentiment analysis uses natural language processing, computational linguistics and text analytics to identify and extract subjective information from the web. The tools offer dashboards and other visualization techniques that can reveal the volume and quality of the sentiment.
Much of the effort in this embryonic field has focused on what companies can learn about how people feel about their own products, services, and conduct. Sleuthing out sentiment about competitors can provide additional insights.
While the same sentiment technology is applied to competitive intelligence, there are subtle differences in how to interpret and apply the information.
Opportunities in the Twitter Stream
Themos Kalafatis, a consultant specializing in predictive analytics, recently did an analysis of Facebook posts and tweets related to European telecommunications companies. “Identifying the topics that subscribers discussed in their posts repeatedly gives us an indication of where a telco should pay attention,” he says. “If we find that customers are repeatedly talking about a competitors ‘network’ with intense, negative language, we can choose the right time to air commercials implying we are constantly working to create better network coverage.”
Sentiment analysis can also provide early insight into competitors’ initiatives. “Very often companies will test market before they release a product,” Carrabis says. “And no matter what you get people to sign saying that they won’t share information, they’ll go online and talk about products they’re excited about.”
In addition, sentiment analysis can alert companies of new competitors who are bubbling to the surface. “If you are Ford, you know that Chevy is one of your competitors,” Carrabis says. “But Ford may not think of public transport being a competitor.”
However, he says, a car company which realizes that can analyze online discussions to get insight into why people are making different transportation choices and change their products or marketing to emphasize growing concerns, like ecological impact. “We have to think broader and wider than we used to,” he says.
That’s why it’s important to understand how people discuss competitors online. “When car shoppers talk online they don’t talk about ‘quality,’ ” says Susan Etlinger, an analyst with the Altimeter Group, a consultancy. “They’ll say, ‘I love the leather interior’ or ‘the cup holder fell out.’ It takes meticulous work to roll together all the indicators of quality.”
Etlinger suggests that social-media listening teams work with the teams in the organization that handle keyword search terms and search-engine optimization efforts, since they have a solid grasp on how people online actually talk about the industry and products.
Another thing to keep in mind: “At any point in time, the way people feel about a brand can be distorted online, because things like Twitter are so volatile and affected by the news of the day,” Etlinger says. “But over time, you can get directional trends—why do people love or hate you, how do they feel about your product compared to the competitor’s products.”
A Human Element to the Detective Work
Such monitoring can give a company a glimpse of a competitor’s product a year or two before they are released. “You can see people are talking about your competitor in more engaged, emotional language—they’re more juiced,” Carrabis says. “From that we can tell a health care provider that a competitor is about to clobber you.” Exactly what the competitor is about to do may be hazy, but the online chatter can alert a company to start detective work.
As all this suggests, sentiment analysis requires creativity. For example, Kalafatis’ analysis for the European telcos was particularly challenging because it focused on posts and tweets in the Serbian language, which required him to develop specialized tools to find links between words that provided insight. The Serbian word novo (“new”) received a lot of reactions along with hocu (“I want”) and dopuna (recharging credit for prepaid subscriptions).
Because of that, he could determine that the telcos’ subscribers wanted new promotions, but the companies had to determine exactly which types of promotions. “We learned that there were issues with incorrect re-charging that were creating a very negative sentiment. But then we needed to find which operator co-occurs with this sentiment,” he says.
There are limits to what competitive sentiment analysis can provide. “The challenges you might address, using your company’s own customer, product, and transactional data, are far more extensive than those you can tackle via available competitor data,” says Seth Grimes, an analyst who runs the annual Social Analysis Symposium. “For instance, you’re not going to have access to your competitors’ contact-center notes and warranty claims, or to your competitors’ customer profiles and transaction records. But with your own company’s, you can create some very rich analyses.”
One Piece of the Analytics Mosaic
For such reasons, competitive analysis is usually only one piece of the data mosaic. For example, one company noted a drop in sales of its flagship product. It analyzed online chatter and found customers talking enthusiastically about a new product a competitor had just introduced. When the company analyzed its contact-center data, it found that returns correlated to discontent about an attribute its own product lacked, but the new competing product offered. The company was able to identify the problem and tweak its own product thanks to the combination of competitive sentiment analysis, its internal data—and brainpower to put the pieces together.
As Grimes says, “Sentiment analysis can help you understand how the market perceives you and your competitors’ products and services, but keep in mind that sentiment is only an indicator, useful in measuring and projecting market impact, not a substitute for strong human judgment.”
Joe Mullich, a freelance writer based in Los Angeles, can be reached at firstname.lastname@example.org.