At Pinterest, a small support team with just five staffers is responsible for helping millions of people use the popular photo-sharing social media site. They offer assistance to users — or "pinners" — who need to regain access to their accounts or report bugs, among other things.
To avoid being overwhelmed, they needed a way to identify support emails that require immediate action, and they found a tool that does just that: Zendesk's customer satisfaction prediction application, says Maggie Armato, reactive team lead for Pinner Operations. "Before, we'd manually sort through tickets and mark them as ‘high-priority.' Now Zendesk does it for us — and much more accurately," she says.
What makes the system work is data culled from Pinterest's customer support activity. Predictions about how quickly the team needs to respond to support requests are calculated based on past support history and incorporate a range of metrics, such as how long a ticket has been open, the text used within a ticket and how many times a ticket with a customer has been reopened. "These were all new data streams an agent wouldn't factor in when manually triaging," Armato says.
Now the team is able to take action much more quickly. "A faster response time doesn't make all of the on-the-fence Pinners happy with our service," Armato says, "but it has made the majority happier."