Directly, provider of on-demand customer service apps, is releasing today it’s latest innovation, predictive routing.
Companies like Pinterest, Nextdoor and FreedomPop use Directly to scale online support by routing customers questions to their expert users, who earn rewards by resolving the questions.
Unlike hourly call center workers, the experts know and love the products already, and they answer questions on demand, where and when they want to. This method delivers significantly faster response times, 4.5 minute average compared to hours or even days using traditional methods, high customer satisfaction ratings of 92.5% average, and better efficiency.
With Directly predictive routing, companies can get more questions into the right experts’ hands, where their highly qualified expert users can resolve them faster.
A longstanding goal in customer service is sending the right question, or ticket, to the right talent every time. Historically, this simple idea has proved difficult.
Using people to categorize and route tickets is accurate but slow and expensive, and the slow speeds can’t meet rising customer expectations for fast answers. Asking customers to categorize their own questions is instantaneous but frequently inaccurate. And building business rules based on keywords and natural language expressions has required large investments in a fixed set of rules that are complex to build and maintain.
Directly’s predictive routing technology augments existing methods by using machine learning to help business rules get smarter. Directly’s routing predicts whether experts can resolve tickets based on their similarity to previously seen tickets.
This way even tickets that business rules miss — like ones with missing or wrong metadata — routinely get identified and routed to the best experts. In a recent test, predictive routing identified 40 percent more tickets for expert routing than the business rules alone did.
A second phase of predictive routing will take the capabilities even further by identifying new categories of questions that experts may resolve based on similarity to existing categories. It then will drip-test tickets in the new category by sending a small sample to experts, and will expand the volume if the experts perform well.