During the last month, your chatbot had 60 thousand interactions with users. What conclusions can you draw from this? Not many, to be honest. The only thing you can do with this piece of information is to compare it with the last month’s stats. Numbers with no context are a nightmare for any analyst. It’s like responding “I’m Iwona” to someone asking me to tell something about myself during employee integration. The only information my co-workers receive are bare facts. In this article, I’ll show you how to obtain more information from your chatbot data and learn more about your customers and business.
Monitoring user behaviour on websites and apps is a must for your business. We no longer live in a world where we have to obtain information from other people. More often than before, instead of googling, we tend to contact brands via Facebook messenger. If the answer doesn’t satisfy us, we simply turn to a different source. This example proves that if you don’t monitor customers’ interactions with your chatbot, you’ll never know what really led to the loss of a potential client. Advanced analytics, integrated both with inside stats dashboards and outside analytics software, will help you understand the needs of customers, improve your chatbot and business.
KODA Bots analytics dashboard
Basic information, like the number of sent and received messages, is displayed on almost every chatbot data dashboard. If you don’t have your own verified tools to assess the performance of your bot, you can use, for example: Yandex Metrica, Dashbot.io, or Botanalytics.co. If you notice that users are less responsive and don’t answer to your messages, you’ll be able to react quickly end determine what has led to this situation. If, however, you want to find out what conversation paths bring the best results, you (or a commissioned company) will have to build relations.
Relations are nothing more than connections between pieces of data. Most of them are generated automatically in the company’s database. How does this work? Users are assigned to a specific ID after their initial contact with the chatbot.The ID is later displayed with every message sent and received by the user, and with every action the user makes (for example, clicking on a link, blocking the chatbot). This enables you to monitor the user’s behavior not only throughout time but also amongst many types of messages and chatbots. You can also gain valuable information. For instance, that the most effective actions your chatbot can take on your brand’s fanpage is sending video messages on a Friday evening. Or that it’s short, contact-encouraging messages that convert best on your website.
As an analyst, I don’t only rely on automatic relations but also try finding them myself. This allows me to see how seemingly unrelated statistics interact with each other. It may turn out that if we increase the frequency of messages sent by the chatbot, we’ll raise the number of its users by 50%. However, don’t be surprised that one apparently insignificant change (like using a different personal pronoun) can decrease the number of users by 40%. Microsoft Power BI is an analytics software I use to create interactive reports with data visualizations. One of the major advantages of this tool is that if there’s an abnormality (for example, a sudden decrease or increase in the number of messages), it scans all the available data and finds the cause of this situation.
user analysis in Microsoft Power BI
Take a look at the above example. It shows that only one user (!) caused an 870% increase in the number of messages, and then the later decrease.
Thanks to Power BI’s fast and comprehensive analysis, you can quickly react to sudden problems or sneak a peek at conversation paths of the most engaged users.
An advanced analysis should never end with quantitative data. Numbers are just numbers, but where’s the user? The amount of qualitative data we can extract from chatbots is invaluable. It’s hard for me to find a better online tool with which I can learn about the users, their preferences, way of thinking, to understand how they make decisions and how they communicate.
When I write reports, I always play customer flows in my head. What are the most frequent conversation paths, what kinds of messages cause that users choose moderators, end their conversations, or, on the contrary – how many users feel loyal to the chatbot? For each step in the customer flow, I note phrases and sentences users write at that moment and create wordclouds. In my opinion, this is where analytics provides me with the most valuable insights. I can identify the right moment to ask the right questions and direct users to products or services they need.
user analysis in Microsoft Power BI
If you want to extract as much information from your analysis as possible, you can categorize and index messages according to the user’s sentiment. This can work in the case of chatbots where users often describe their experiences with the brand or blow off their steam (we all know that the Internet is often used for complaining, not complimenting :))
Remember that data retrieved from people, computers, chatbots, and other electronic devices is not enough. The data has to be collected, organized, and then interpreted to have some value. This is where data analytics steps in. It allows us to understand the most important information about users.
When meeting a new person, you have to decide what’s interesting for you, what you want to ask about during the next meeting, what topic to extend and what subject to avoid. The same is with data analysis. That’s why it’s worth having an analyst amongst your friends 🙂