We all know this title is a common title around the web nowadays. Nevertheless, we’ll get this straight and give you some tips to enhance your conversation flows and engagement metrics.
- Have a glance at your daily metrics
When you have a bot with lots of data (i.e. more than 1,000 DAU), you’ll need to check the metrics of your bot daily. These metrics are total message change, total conversations change, average conversation/user, average conversation steps/user, total user, total sent messages, total received messages, average session length, most common phrases, and most active hours.
These are your dashboard metrics that should be tracked routinely. Be on top of these specific ones much more: Average conversation/user, average steps/user, average session length. These 3 are associated more with engagement rather than the others.
Let’s say you have an entertainment bot. If your average session length is increasing, this is good news. On the other hand, if you have an on-demand bot, this counts bad for your bot. As you may imagine, users should be fulfilling their needs asap such as calling a cab on Uber bot or ordering a pizza on Domino’s bot. As a result, average session length should be considered on the nature of the bot.
Conversation step that we provide as a metric means that if the user says ‘’hi’’ and the bot answers back with “hi.’’ This counts as one conversation step. Calculate these steps to track your stickiness with your users. Average conversation steps per user is a highly critical metric to see how your engagement rates are decreasing and increasing depending on your bot’s nature, of course. Average conversation per user is a general metric considering engagement. If your users are increasing and in the meantime, your conversations are not, your average conversation per user will be decreasing as it looks like you’re in trouble. So, you should track these 3 metrics on a daily basis to iterate your bot and increase your engagement rates.
Also, most active hours can be useful to send push notifications to your users in order to increase engagement rates. However, this content has to be relevant and include a call-to-action to get the user to write back and engage with your bot. Unless it’s done in this manner, it doesn’t make any sense to push notifications because you’ll be spamming your users.
2. Filter your conversations based on conversation steps (one of the most important engagement metrics)
So what’s the most important metric to look at? Conversations of course. Aggregate data measurement is important too, of course, but conversation details answer many things for your pain points. Until we have bots speaking with our users without any troubles, we’ll dig into conversation details to find bottlenecks.
First of all, you’ll need to filter your conversations based on conversation steps within a time frame. Let’s say you filter the conversations which have at least 3 step conversations and 10 conversations in total, and also filter 5 step conversations and 10 conversations in total, you’ll see the difference in the conversation details where your users are having problems. You can extract a pattern to see how your bot should continue talking with your user.
3. Track correlation data graphs to see the whole picture
It’s important to track single metrics of course, but it’s more important to track combined metrics because correlation data analysis helps you easily understand if there’s a problem on your bot.
For instance, if your number of users is increasing, the number of conversations is increasing as well. If the conversations steps/user is decreasing — depending on the nature of the bot — there can be a problem. When you think an entertainment bot, conversation steps/user should be increasing as much as user and conversations metrics accordingly.
4. Follow retention rates and compare to analyze which conversations are less likely to churn
We’re familiar with retention graphs to track our marketing efforts, already. When we consider this on bots, it makes more sense to see the stickiness of the users. From day 1, how many users do your bot talk to — how much percentages? If we say on day 1, the users are retained on your bot in 70%. On day 2, this can be decreasing to 50%. On day 3, 30%. So you will see your retention rates on a 7-day period basically. Also, when you filter these retentions rates and compare in a way that average step conversations are considered.
If you have 5 conversation steps on average, you can filter your retentions rates more than that average and less than average. In this way, you can see which retention rates are performing better in the aspect of conversation steps. Also, you will be able to analyze which conversations are likely to churn. You can decide to design your bot accordingly to respond the user by creating a step or removing one. Comparison between those retention rates based on filtering the conversation steps helps you to increase your retention rates on your bot.
5. Take over the conversation when needed
Sometimes bots can be stuck at a point, and when this point is identified, the conversation can be taken over by a human to improve human-bot communication. In order to do that, switch the live button on to see the live conversations and dig into the conversation details.
As you can see above, on Slack bots, even with teams, (also on every platform including Messenger — you can change your bot with the multiple bots on one account), you can stop the bot and take over the conversation, talk to your user. It’s that easy. In this way, you are able to avoid decreasing your engagement rates as well.
We have covered the fundamental analysis of your bot’s engagement and retention rates. In our next article, we’ll cover the titles below:
6. Set events to see the repetition of a specific happening — transaction, payment, submission.
7. Set goals with funnels and analyze funnels to track completion rates
8. Broadcast messaging to send notifications to increase your engagement rates