The best Chatbase alternative for Dialogflow ES conversation analytics

First published: Nov 2017 | Last updated: June 2022

The Chatbase service, which was recently shut down, provided analytics for Dialogflow bots.

However, it was missing an important feature: you had no way to ensure that the intent mapping was accurate. As a result, everything that was built on top of that assumption was not entirely useful.

Here is a much better alternative: Airtable

Here are the reasons why I recommend it

Maintain versions easily

You can convert your Dialogflow ES agent ZIP file to a CSV format and upload it to Airtable to maintain the version of your chatbot.

Group by sessionID to visualize conversations

You can convert your History JSON to CSV format and group by sessionID to visualize the flow of individual conversations. This is similar to the conversation view in Chatbase.

Group by intent names to see the most popular intents

You can group the History CSV by intent names to see which were mapped most often. This is not the same as the Funnels feature in Chatbase, but it is quite insightful.

Filter by fallback intents to design new intents

You can filter the History CSV by input to see what user’s say most frequently. Based on that, you can design new intents for the next version of your bot.

This is quite possibly the most important reason to use Airtable. You will find it very hard to assess the performance of your bot unless you can calculate its accuracy. Even though the process of calculating accuracy is quite tedious, you can speed it up a lot by using the filtering and conditional logic features in Airtable.

Create auto-updating summary views for your team

And you can create automatically updating summary tables for multiple versions of your bot and share the Summary table with your team members.

Note: This is my old website and is in maintenance mode. I am publishing new articles only on my new website.

If you are not sure where to start on my new website, I recommend the following article:

Is Dialogflow still relevant in the era of Large Language Models?