Bot Analytics: An Interview With Dennis Yang Of DashBot

[Yeti] What is the primary problem that DashBot solves?

[DY] We started out building bots ourselves. After building a few bots, we realized that we wanted to see and understand what was going on with our bot, and aside from digging through our logs, there was not a good solution available. So, we built Dashbot from the ground up to provide analytics and visibility for bots and any other conversational UI. After we launched Dashbot, our customers started asking for more features around understanding & improving bot engagement, as well as ways to re-engage and connect with their users. So, you’ll start to see us launch more features that help your bot get more usage, revenue and traction.

[Yeti] How is Dashbot different from other solutions like Google Analytics or Mixpanel in terms of solving this problem?

[DY] Traditional analytics solutions don’t handle conversational data streams properly. If you shoehorn the event-based paradigm into your bot, you’ll end up losing the richness of the conversation, from which insights can be learned. The words, images, GIFs, and emoticons that users send to your bot have meaning, which existing solutions do not understand. Plus, with Dashbot, you’ll be able to see the conversation as it appears to your user, which is incredibly informative. Why hamstring yourself by imagining a conversation as a series of events?

Furthermore, our integration is a lot simpler; in most cases it’s about three lines of code. Since we process the entire conversational stream, you don’t have to know ahead of time which events to track. Simply integrate with Dashbot and you’ll start seeing data appear immediately. Our full documentation for integration is here: https://www.dashbot.io/sdk

[Yeti] What are some examples of the unique bot analytics that DashBot provides? What type of learnings have you and your clients derived from the data?

[DY] Upon first glance, bot analytics may seem straightforward for the standard metrics: users, sessions, engagement, and retention. However, when you delve a little more deeply into the numbers, you quickly realize that although some of these metrics may exist in the web and mobile world, they are quite different for bots. For example, you may have multiple users chatting with your bot at the same time. This results in a single session, with multiple users–something that would break traditional analytics models.

Since we process the entire conversation, not only can you see the entire transcript of your conversations, but we also provide message-specific reports. Our “top messages in” report is a great way to understand what are the most common messages that your users send your bot. Since conversational UIs afford users much more flexibility in the UI, many times you have not anticipated everything that a user expects to be able to do with a bot. Many bots are seeing users ask them to “tell them a joke” or other off-book type behaviors. Your bot is a great way to develop and maintain a relationship with the user, so responding to such requests is a perfect, fun way to reinforce your company’s brand and personality.

A full set of sample reports are available here: https://www.dashbot.io/tour

[Yeti] What are a few areas where you see bots really taking off near-term? How about long-term?

[DY] In the near term, I think we’re going to see a huge range of different types of bots launched as everyone figures out how best to utilize this new medium. There are so many different messenger platforms now: Facebook, Slack, Kik, Line, WeChat, Amazon Alexa, Telegram, and Skype are just a few, each with their own idiosyncrasies.

Most bots today are not yet having complicated conversations with their users, but rather capitalizing on the other benefits of the platform: ease of user acquisition (as compared with mobile apps, where there is extreme app fatigue), ability to personalize every interaction, and saving state. In the case of some platforms, like Slack, the ability for the bot to add to an existing conversation with multiple users is amazingly useful. Many bots are also using lightweight UI components, like buttons, to simplify affordances with their users. Buttons are great to guide users in their interactions with your bot, but one of my favorite things about the bot experience is the ability to ask your users more open ended questions without sending them to an external survey.

As bot technologies continue to improve, the promise of a truly conversational bot gets nearer and nearer. Science fiction has given us several visions of what could be, from the companion in the movie, Her, to the computer in Star Trek. I’m not sure exactly where it will end up, but I am quite certain that information will be more accessible to more people in an easier way through these interfaces very soon.

[Yeti] What are common mistakes you've seen companies make when building bots? What are some common misconceptions you've seen that people often have when approaching the bot-building process?

[DY] One common mistake that we see with bots is that they fail to fully anticipate all of the messages users might send. There are so many acceptable ways to respond to even a simple yes/no question, so make sure that you’re covering all of the possible use cases. And, don’t forget that in Facebook, there’s a “Thumbs Up” button. When you ask for location, be sure to cover the case where the user sends you a location pin, and doesn’t just say a city name or zip code.

There seems to be a concept that “bots are the new apps.” And, while I do believe that bots will come to replace many of the interactions that we currently do in apps, thinking that bots are simply “new apps” fails to understand the nuanced differences of the medium. So, if you simply replicate your app as a bot, it most likely won’t work as well, and you’d be missing out on many of the great advantages of bots. That said, we are all still exploring what UI patterns work best for bots, and at Dashbot, our mission is to help our customers quickly find what best works for their bot.

[Yeti] What advice would you give to a company that is about to build its first bot? Also, why is it essential that they have proper analytical tracking set up for their bot?

[DY] First, as with any good product development process, take some time to really dig into the messaging platforms to understand how people (especially of all age ranges) use them. See how people integrate things like stickers and emojis into their own conversations with humans.

One thing I love about bots is the opportunity for them to have an actual personality. Even if you decide that your bot is going to have “no personality,” make that decision a considered choice. Onboarding, word choice, use of emojis, and even easter eggs are all great ways to communicate the personality of your bot, which will hopefully encourage your users to interact with your bot more frequently.

Seeing how your bot performs is critical to your ability to improve your bot with each iteration. That’s why we built Dashbot. We started building bots on our own, and quickly realized that we could not see the conversations that our bot was having. It is incredibly informative to read through your conversation transcripts.

After we deployed our initial version of Dashbot, we started hearing what types of insights would be helpful for our customers, and many customers wanted ways to act on the insights they were learning. So, we recently added the capability to take over your bot and insert a “human-in-the-loop.” We see this as a very useful tool if the conversation is going sideways, or even just to jump in and ask for quick feedback on a session.

Our vision with Dashbot is that you’ll not only be able to see what is happening with your bot, but you will also be able to act on those insights as well.

[Yeti] Thanks Dennis!