What Emojis Can Tell You About Your Bot
When I began reading up on chatbots in late 2015, I had that same excited feeling that I had about iOS apps when the App Store first launched. Mike Nathanson, also an iOS Developer (and now my Co-Founder) and I started to brainstorm on a consumer-focused bot. We were both a little reluctant though to dive deeper into it because we anticipated that whatever we build would fail at conversation, a lot.
iOS apps were predictable. We would develop, QA, release, measure, and repeat and the exceptions would be treated as edge cases. In the conversational paradigm however, the exceptions seemed to dominate both the news and the end user’s experience. Our assumption was that if we could get better visibility into conversational problems, then we could prioritize development, AI training and customer service and the KPI’s we wanted to achieve could be attainable. In short, if we understood where our bots were failing at conversation, then we’d be in a better position to optimize them. Solving this problem motivated us to create Wordhop.
Defining and Communicating Problems
A traditional analytics UX is comprised of dashboards and logs, but that didn’t feel like the right experience for the problem we wanted to solve. What we liked about a conversational UX for conversational bottlenecks was that we wouldn’t have to obsessively check a website to find problems. Instead, a computer program could monitor for problems and then communicate just the nuggets of data that we needed when the program thought it was important, and in a way that was easy to understand . At the core of our product design was an intelligent assistant that would live within our existing work environment, Slack, and communicate problems to us conversationally. Since we new bots fail a lot, we also didn’t want to be overwhelmed with messages creating even more noise through Slack, so we spent time defining this concept of a “Fail Rate” and used these emojis to communicate a problem based on severity.
These are already familiar objects of a messaging conversation that elicit emotion ranging from “oops” to “OMG”. When we receive a message containing 😱 it’s a call to action that there is a major problem. We also created a mechanism to set the severity level and be able to tell our bot to react to just our most critical problems. In our testing we felt that when metrics become objects of a conversation like this, the data becomes more actionable.
Adding In Even More Visuals To Communicate Problems
Graphs are a nice way to represent data, but when graphs become objects of a conversation it flows more like a presentation. There is no need to fiddle with menus and dashboards to uncover bottlenecks. Wordhop can show and tell you. So, if an emoji can serve as the alert to a problem then you can open the notification and Wordhop will show and tell you more about the problem with a graph and include more data. You can also ask Wordhop for a report using natural language in much the same way you might ask for a report from a data analyst and Wordhop will even correlate conversational failure to other metrics like active users.
☞ Want to see Wordhop in action? Add Wordhop to your Slack Team
☞ Tap or click ♥ to help to promote this piece to others.
