A Closer Look: In search of the perfect search experience

Published June 10, 2015
Last modified June 10, 2015

In the summer of 2013, Zendesk sent its first official NPS survey. The answers we got immediately made one theme perfectly clear: as an agent working in Zendesk, our search sucked.

Our customers’ gripes about search weren’t limited to NPS comments, either. I’d also heard feedback in user groups, tickets, and in comments on our search documentation. People used words like “useless,” “frustrating,” and “impossible” to describe our search. It was tough to hear, but it also made sense—it wasn’t intuitive. To really dig in and find everything, an agent had to know a special filter syntax (tags:my_tag_name). In effect, our search was so not beautifully simple.

The product team and UX team (and well...everyone at Zendesk) knew search needed improvement and wanted to make it better. But first, we needed more specifics on why our customers hated it, and what they were searching for in the first place. We went back to the NPS results, and saw a lot of references to “searching for my own past tickets” or “searching for answers in old tickets.” As it turns out, most of us (including me) are not very good at articulating what we want out of search; we intuitively know what we’re searching for, but aren’t used to explaining our thought process. We figured that the best way to learn what customers wanted was to just get out and observe agents doing their day-to-day jobs.

A few of us from the UX and product teams went to three different businesses around San Francisco, observing 3-4 agents at each site. Immediately a few things became clear, the first being that manager-types have very different search needs than agent-types.

The cringe-worthy things we learned
When working on tickets, a lot of agents are searching for answers in solved tickets and then copying the responses from the old tickets into the new tickets. Often the original response was given by another agent, so it’s a bit like sanctioned (and sensible) plagiarism. I found that these agents were driven to give excellent support; they wanted proven answers. They took into account things like the customer satisfaction rating on the ticket, the expertise of the agent who provided the answer, and the age of the original ticket. Many Zendesk customers (especially those in the software industry) have a constantly-changing product, so we heard over and over that older tickets could have out-of-date answers, and therefore weren’t as relevant.

Managers (or agents with additional responsibilities), on the other hand, were more likely to search based on some specific ticket attributes rather than keywords. This group was looking for lists of similar tickets for a variety of reasons—to check quality, to find examples for training, and a lot of times just to help report on how many issues of a certain type came in last week. Managers needed a little less “magic” and a little more control over their results.

I learned some general cringe-worthy things about our search experience as well. Like the fact that hardly any of the agents using our search ever saw (let alone used) the filters on the left-hand side of our interface. As in didn’t even see them. Our layout gave our customers a blind spot right where we thought we were giving them help. We witnessed agents opening multiple tickets from search results to see whether they were relevant, because we weren’t providing enough information within the search results interface. I also saw several agents leave Zendesk and go into Gmail to search. That’s right—our search was so unreliable that switching tabs to search for email notifications was a better experience. Ouch.

What we did about it
Following the field research, we gathered a bunch of people in a room—product managers, search backend engineers, front-end engineers, and designers—to think through the problems and use cases and make our search experience better. Fortunately one of the agents and one of the managers from the field research were able to join toward the end of that session to validate our concepts and sketches. Getting immediate feedback on our rough ideas really allowed us to focus on the most impactful items.

Coming out of this session, I knew that we weren’t going to be able to provide the ideal search experience for everyone, especially given that agents need search that “just works” and managers need powerful tools. We decided to optimize the search experience for the everyday agent, while still providing tools for managers to ask more specific questions. This meant starting with the most basic features.

Over the past six months, we’ve been doing just that—overhauling the very basics of searching in Zendesk. Things like presenting results in an easily consumable format, showing additional information when searchers hover over results, and placing a relevance bias on tickets updated in the past 90 days.

I’m thrilled that we’ve finally delivered something that will change how agents interact with Zendesk in a very fundamental way. And while I know there’s more work to do (and that I likely got a few things wrong), I hope that these changes improve the experience significantly. Check it out in your Zendesk today and let us know what you think!

If you don’t already have a Zendesk account, you can start a free trial anytime.