As the next version of the auto-complete (this time written with ReactJS) is around the corner, we want to celebrate our journey by sharing some thoughts and experiences from the team.
Building an auto-complete for all business data is an interesting task, as it constantly confronts our designers and developers with new UX-nuts to crack. From limited space on mobile, to result grouping there are hundreds of stories to tell.
Today I want to tackle a question that has been floating around for months:
Choosing the hashtag to trigger auto-complete
After kicking off development of our business chat in early 2014 we had around half of the team dedicated to auto-complete. As it represented our most important innovation focus, we invested a lot to get a proof-of-concept out as fast as possible.
Together with our research partners we engineered a set of natural language processing techniques (to check for questions, date, addresses, etc.) and matched them with data from Github, Google Drive and some calendar entries. We completed the stack with a JQuery Auto-complete component and put everything in a simple chat-room. Then we gave test users different tasks, iterating and testing every three weeks.
The goal: overwrite the urge to tab out of the chat-room
The first version of the auto-complete had no trigger-button:
It simply popped up when it had a match for any combination of the last three words and closed itself when the user kept writing or selected a result. We wanted auto-complete to feel like magic and were willing to invest a lot of brain cells and energy into a constantly delivering smart assistant.
To summarize the goal of our testing-run, I wrote the following on one of our company’s whiteboards:
Our auto-complete will be successful once our users don’t feel the urge anymore, to tab out of the chat and browse their data on their cloud services.
But while the results of the auto-complete got better, the amount of „tabbing out“ stayed more or less the same. We concluded, that the quality of the results couldn’t be anymore the sole barrier for adoption. Then we met Ben and his Google Glasses.
The Human Pilot overwrites the autopilot
Google is a company that is constantly rolling out new products. From robotics over self-driving cars to augmented reality, they invest a lot in elevating the status quo around the consumer market. As an avant-garde company they’ve learned a thing or two about emotional hurdles when rolling out new products. But no other recent product failed more on the socio-cultural side than the Google Glasses.
Ben from NextSociety as one of the early adopters had to constantly deal with distrust and sometimes anger towards his eye-wear when in public. People weren’t sure when/if he was recording, asking him to take them off or at least cover them. While being a remarkable piece of technology it failed on the most critical side of product experience: Control and Trust.
As with Google Glasses people weren’t really sure why and when the auto-complete popped up. When we changed our interviews accordingly, many users stated that they felt like auto-complete was doing too much. They also felt uneasy with the interpreters constantly listening to their text input. So we set out to increase their control and privacy feeling over auto-complete.
We want the auto-complete to become usable both for power- and casual users. The commonly used „/„ character was too technical and the „+“ too assertive. So we got inspired from the Hashtags our little brothers and sisters where using in almost any social interaction. All of a sudden our testers where playing around with the auto-complete, using it more as a form of expression than a simple solution. The control was given back to the user, the technology became approachable.
There is still a lot to tweak until we can replace the urge to tab out. We are in the middle of a long journey, but the users that convert to it see a measurable boost of efficiency. Our most active companies reference hundreds of objects every day. Things they previously had to search on other programs and websites can now be accessed while writing the message.
If you want your users to adopt new conventions, weigh-out magic and automatism vs. control and trust.
Cheers from London,