We’re constantly tinkering at Instabot, looking for ways to make sure Instabot adds value to you. For many marketers starting their first chatbot, there is a temptation to add conversational elements to mimic the way you would text or chat with a real user. The assumption is that with a conversational interface, more “chatting” makes a better bot experience. We see this mistake all of the time. In reality, it’s a chatbot’s ability to elevate information to users in a concise, targeted format that makes them so powerful; they should be used to eliminate the need to hunt for answers.
If your users wanted to endlessly browse your website, scanning through content and using ineffective search bars they wouldn’t need chatbots. Bots perform best when they highlight clear decision points and quickly direct you to what you need. In our experience, extra conversational text is often more distracting than worthwhile. That’s not to say you can’t be playful or provide content when relevant, but when starting out its best to get to the point.
Learning from Experience
We’ve seen this lesson play out in our own bots. Back in September, we launched a support bot inside our portal with the goal of quickly addressing users’ questions and handling a majority of support requests. After 3 months, our initial results looked promising. We had a 40% engagement rate (very solid level from users), and more importantly we were getting a ton of feedback through the bot about what people were having issues with. It wasn’t until after launched an experiment to tune the bot via cutting extra nodes that we discovered how many users we were losing from 1 unnecessary sentence of 5 words.
Bot Results
In our original conversation, there were two prompts. The first was an introductory statement, followed by a user-centric statement & multiple choice options
“Let us know what you need and we’ll point you in the right direction or reach out directly”
“I’m looking for help with” (Getting Started, Help with bot trigger, how to deploy, schedule a call, something else)
After going in for edits we decided to experiment and remove the second prompt and simplify the bot to:
“Let us know what you need and we’ll point you in the right direction or reach out directly” (Getting Started, Help with bot trigger, how to deploy, schedule a call, something else)
After removing the prompt “I’m looking for help with” we saw the engagement increase from 42% to over 67% leading to dozens more interactions with users needing help. We collected this data over hundreds of sessions giving us confidence that it had true statistical significance. More importantly, we found that this with increased amount of responders we were also getting different results.
In the first 3 months of the original bot, 10% of users selected “Getting Started” while over 50% of users selected “Something Else”, indicating that a majority of users weren’t being helped by this bot.
In the 3 months after we trimmed the excess wording, we saw this number drop to 28% while 42% selected “Getting started!” That's a LOT more people who now knew they needed help getting started!
Why the change?
Overall, the removal of one line of text lead to two things occurring in our bot.
-
Less text for users to have to read, parse, and respond to.
-
A simplified question which lead to clearer responses from users about what they needed.
As a product organization, this was a huge win. We were able to get an increase in engagement meaning we were solving MORE questions leading to happier users. More importantly, we learned more about what users were having trouble with, allowing us to tailor our support materials and video content to the getting started use case.
In the end THIS is what bots should do; provide clarity to users about what you do or how you can help, crystallize their needs and provide you with the data to help address their needs more quickly and accurately.