Chatbots Magazine Become a Writer Intro to Chatbots

Chatbot Analytics - How do you visualize user engagement?

(Aaron Lupo) #1

I am building a platform to create chatbots for mobile applications and I continue to get hung up on the analytics piece. I find it difficult to quantify users’ engagement when there are so many possible variations or responses.

I am wondering what others have thought about and found valuable when looking at engagement analytics data. More specifically - I am curious if there are graphs or visualization solutions out there that provide general engagement trend analysis.

Mostly curious what other people have seen and found useful. Look forward to hearing!

(Kubra Zengin) #2

Hi Aaron,

You can use Botanalytics for the engagement & retention measurement.

Let me know if you have any questions.


(Aaron Lupo) #3

Yea I have looked into Botanalytics a bit and definitely see value in a lot of their software. But I still think there is some data that is difficult to capture in simple graphs (line, bar and pie).

For example identifying most common trends of users at each question in the conversation - analyzing the types of users that take one path vs. another. This gets exponentially complicated as conversations grow.

I am not sure how others feel but I am not convinced the analytics solution for Bots has been outlined yet and I think that could be a major advantage for the company that does. Right now companies are scrambling to build bots but I am not sure there is enough data analysis to support continued bot development. Companies need to be able to measure value and I don’t know if that tool exists yet.

So I was more posing the question to see how others felt on the subject and maybe they know of or work on a tool that starts to fill this gap.


(Jordi Martinez) #4

Hi Aaron

I think that maybe there is not one measure of Engagement that catches all the possible use cases. For instance, a transactional bot to sell tickets may measure the conversion rate and the returning customers rate, which is easy to do because the data is available. For customer service bots, provided the service is only available to registered users, you’d probaby need to analyze the data collected from the bot and data collected from any other contact channels available to your users, which is far more tough, but still possible.

A/B, or multi variant, tests might also be interesting entry data to measure models which likely must be bespoke in order to provide useful insights. But, again, this makes a broad definition of Engagement harder to set.

(daiangan) #5

I use Dashbot, it is very good.

(Kubra Zengin) #6

Hi Aaron,

Thank you for your feedback. We’re improving our product of course. Hope to connect sometime. Would love to.


(arpit1) #7

@aaron.lupo I had to do the same for my project - Bottr. It is a platform where many people have created their chatbot and they wanted to see reports or data on how people were interacting with their bot. I started including basic key metrics (that I felt were relevant for user engagement) such as views, messages, shares and reactions to responses. PFA the screenshot. Later we also started sending “most asked question” and “best responses” via email.