Kanjoya’s analytics dashboard retrieves data from your company’s Yammer network to gauge emotions across departments like sales, marketing, and HR. Individuals won’t be tracked and profiled, as the algorithm will aggregate and compare conversations between groups.
The company claims to use machine learning processing to understand varied and complex emotional signals. Kanjoya’s founder, Armen Berjikly, told me the algorithm works with 70 percent accuracy in detecting 80 emotional responses, such as “surprised,” “frustrated,” or “happy.” He said this is on par with any group of colleagues or friends, who, like machines, often misunderstand complex emotions.
Through this partnership, Yammer users can access a word cloud to view the trending company-wide conversations.
For all you gossip-mongers, the new tool will also showcase the most praised and liked individuals in the office. Yammer’s VP of business development, An Le, told me this helps companies locate “rising stars” in departments, who are asking the tough questions and gaining both influence and authority.
Le offered an example of a junior engineer at a global research company in Perth, Australia, who challenged his company’s status quo on Yammer. “He became the second most followed person behind the CEO,” she said. “That’s the type of conversation we need to understand and track.”
Le told me that Yammer has been searching for over a year to integrate with a sentiment analysis tool to capture emotion across the enterprise.
“Kanjoya offered us a way to understand not just the number of “likes” or volume of conversations, but drive into natural language processing and 80 layers of emotion,” she explained.
Kanjoya claims to have a different approach than its major competitors, Lexalytics and Crimson Hexagon. It began as a social networking site — Experience Project — where users expressed a broad range of emotions. Berjikly leveraged this data set, primarily unstructured data like text, to build an algorithm to track both sentiment (positive, negative, and neutral signals) and emotion.
It’s more reliable than a survey, but sentiment analysis is a notoriously tough nut to crack. The most enduring problem is that it’s difficult for an algorithm to tell the difference between a positive and neutral comment. Consider this ambiguous statement: “The parking lot is located nearby.” Berjikly told me that the company grappled with this problem, and responded by discounting these types of comments.
Another obstacle is that some emotions have a stronger signal than others. It might be an over-indulgence in the TV show “The Office” on my part, but I’d be surprised if any office in America doesn’t engage in a little, healthy sarcasm. Berjikly admits this is a problem, but claims the algorithm gets better at detecting an organization’s nuances over time.
It may seem a bit invasive, but the technology has proven useful in the beta phase for specific use-cases. For instance, if an HR manager decides to switch the company’s health provider, she can view a dashboard to gauge the response and react accordingly.
The new tool is available to Yammer users today, who can try it for 30 days for free before paying an additional fee. Le would not disclose pricing but told me the fee is negotiated on a company-by-company basis.