Optimization
When you launch your bot, the platform tracks analytics from the usage of your bot. This enables you to make data-driven decisions to improve your bot even further. The platform tracks
- Stats from users, stories, and your bot itself
- Click Performance
- Broadcast Deliverability
Tracked data-points about your chatbot's usage with all of its users. There are 6 types of stats:
- Messages
- Sessions
- Time Spent
- Sentiment
- Users
- Fallback Rate
These stats can be seen on your bot's Dashboard, as well as on the Stories and Users pages.
Total number of incoming messages from your Users to your bot. This does not count Responses (i.e. outgoing messages), because the number of responses does not indicate usage performance of your bot.
Total Messages is visible on the Dashboard, Stories, and Users pages. Messages details are available under Messages in the Advanced View of the Inbox page.
Total number of sessions your Users have with your bot. They are directly affected by the Session Length that you set on the Personality Page. Shorter sessions lengths naturally lead to more sessions per user, so please be aware of this when changing you bot's session length.
Total Sessions is available on the Users page.
The amount of time Users have spent to interact with your bot. It is calculated from the time difference between any two messages from your user. The diagram below indicates how it is tracked
Time Spent is directly proportional to your bot's Session Length that you configure on the bot's Personality page. Longer session lengths naturally lead to more time spent, so please be aware of this when changing you bot's session length.
Average Time Spent per Session is available on the Dashboard. Total Time Spent is available on the Stories and Users pages.
Quantified emotional tone of User Messages derived from the AFINN-111 wordlist and Emoji Sentiment Ranking. There are 3 classifications of sentiments; positive, negative, and neutral. Positive sentiments have a value of > 0, negative sentiments have a value of < 0, and a neutral sentiment has a value of 0.
This is helpful for you to classify a your Users' happiness levels when using your bot.
Sentiment graph from the Dashboard page
Average Sentiment is available on the Dashboard, Stories, and Users pages.
Total number of Users using your bot. This is a good indicator to estimate your bot's following, growth, and virality. New Users is the number of new user accounts created within a time period, and Active Users is the number of user accounts that sent at least 1 message or 1 click through your bot within a time period.
New Users and Active Users are available on the Dashboard page. Total Users per Story is available on the Stories page.
A record of a link visited by a User who clicked on a URL Button in your Bot. Clicks are useful to track link engagement through your chatbot, and give you granular details about
- The identity of the user who visited a URL
- What date and time they visited the URL
- Which Text, Card, Audio, Video Response or Persisted Menu they clicked on to visit the URL
All Click records are available under the Clicks tab on the Inbox page. A summary of Clicks activity is available on the Dashboard page.
A record of a broadcast being received by a user in your Bot. When a broadcast is published to all or a targeted subset of users, it is throttled over a period of time to ensure maximum deliverability. With broadcast records, you can track
- The identity of a user who received a broadcast
- It's latency - the difference between the user's reception time and the broadcast's publish time
- Whether the broadcast was completed for a user, or whether it incurred an error
There are several reasons why a broadcast could incur an error for a user - they may have removed the chatbot from their messaging app, or your bot exceeded a messaging app's rate limits before a user received it. Errors and latencies are great for logging your broadcasts' performance.
A record of a Message that triggered an incorrect Response from the bot. When you launch your bot, Users will ask questions that are outside its scope more often than you would expect. Normally, these would trigger the fallback story in the bot
It may not only be the fallback story - your bot might display another existing story that you did not expect it to.
When this happens, as the owner of the bot you could solve it in 2 ways:
- 1.Make the bot answer similar questions correctly in the future
- 2.Ignore this user's question as it is too customised
If you choose the first solution, you may change the message's correction status from "correct" to "wrong". After that, you may use the Wrong Responses tab in the Inbox page as a Product Management tool to track all unhandled messages that led to wrong responses.
Wrong Responses tab in the Inbox page
For each unhandled message, you can create a rule to fix it, then click on Replay Message to test your bot against the same message again Now and compare it to the time it was originally sent. Finally when all is settled, you may change its status to Resolved
Replay Message in Wrong Responses
Wrong Responses is a Product Management Tool built-in into BotDistrikt to help you optimise your bot further post-launch. It's an alternative to other tools you use to manage these tasks like Trello, Asana, or JIRA.
Last modified 1yr ago