How to Support Free Users Without Going Broke

Posted by Directly

Offer a free product? The high-stakes challenge is figuring out how to provide great customer service to a huge base of free users with the smallest team possible. Here's how fast-growing companies like Pinterest are using on-demand customer service to do just that.

A new approach

In late 2014, Pinterest had tried a variety of approaches to scaling customer service and was still looking for an answer.

Screen Shot 2015-07-23 at 2.36.48 PM

"We had a very small team and a very large user base, and we were trying to support them with the highest quality possible. We looked into forums, Facebook groups, things like that. None of them really worked out," said Maggie Armato, reactive support lead at Pinterest.

So Pinterest rolled out an on-demand customer service program in which their expert users, or "pinners" answer questions for other pinners using Directly apps. Directly integrates with Pinterest's helpdesk system and routes informational questions that don't require account access -- like how do I search my own pins? -- to the experts. The experts earn rewards for resolving the questions, with one expert winning most of the reward, and other experts earning partial rewards for collaborating.

The program started out strong and grew from there. Based on a 90-day period May through July 2015, expert pinners answered tickets within 8.7 minutes average, and resolved 84 percent with an 82 percent customer satisfaction rating. Statistics that soundly beat other options, like virtual agents, and that rivaled the performance of their in-house team, but at a lower cost per ticket.

Success factors

Data from case studies like Pinterest's and from hundreds of thousands of tickets across the Directly platform provide some interesting insights about the success attributes of using on-demand customer service for free products.

1. Reward amountExperimenting with reward optimization is key to incentivizing performance and retaining affordability. Successful programs typically start by offering rewards set at half their cost-per-ticket for in-house or outsourced agents. The final amount is a matter of testing, but rewards between $1.50 and $5.00 per resolved ticket are average for free products. For a $3 reward, the winning expert earns $2.10 of the reward, and collaborators share the rest based on a point system. Collaborators can earn partial rewards for actions like providing additional answers, upvoting or downvoting other experts' answers and even for escalating questions to the helpdesk.

2. Ticket volumeIn order to provide enough ticket volume to keep top experts engaged, successful programs for free products have at least 10,000 to 15,000 total tickets per month, and are routing no less than 3,000 tickets per month to experts. This volume sustains 10 experts on average, with the top experts resolving over 2,000 tickets per month and newer experts collaborating while they get accustomed to the environment. There is a lot of flexibility in the system; when demand spikes, experts respond. So while there is a threshold for minimum volume, the distribution of the volume can vary greatly.

3. Expert performanceWhile there are certain behaviors that make for successful experts - like good listening and questioning, concise writing and an empathetic manner - there are not yet clear predictors about who makes for the best experts. In some cases, newer customers do well, in others, long-standing contributors to community forums do best. Successful programs invite experts broadly using drip recruiting, test expert skills before approving them, provide ongoing expert education and let Directly's routing engine determine whose performance merits them getting early access to new questions.

Finally, the successful programs have a lean startup mentality, no matter what their stage. Because on-demand customer service is a new model, the only way to truly find out whether it's the right fit is to run a test. For both successful programs and a couple of unsuccessful ones, within two weeks, the data points in a clear direction.

If you support a free product and would like a demo, let me know at and I'll help make it happen right away.