Believe it or not – chatbots that pop up on your website or mobile app — like IBM’s Watson or LiveChat – can double conversion rates. The secret is tracking and analyzing chatbot data you can act upon.
Many websites today, particularly e-commerce sites or consumer-facing platforms, have some sort of chatbot or chat service to help users if they get stuck, or if they have questions about their account or their orders. A lot of data can be captured around those user-chatbot interactions. For example, you can learn when people are launching the chatbot, what are they asking, and whether they were able to find an answer with your customer service reps (whether it’s an automated AI machine or an actual human being that is responding to them). Such information is crucial to really understand where users need more help, and potentially it can guide you in your marketing efforts, your site layout and site optimization efforts. It can affect offline customer service, too.
A lot of opportunities exist when you can connect interactions with your chat system and customer service platform with your online analytics platform. For example, when tracking in Google Analytics, all your site interactions can be really, really valuable to understand if the chat service is helping with your online platform and if your customers are finding it satisfactory and actually improving your conversions. Some example interactions and insights you can obtain around the value of chatbot include:
- What pages people are seeing
- What they’re buying
- What they’re converting on
- The interactions they’re having, alongside with how often a chatbot is open
- What people are asking and if they are clicking on the answers
HOW IT WORKS
The prerequisite is getting conversion tracking and user interaction tracking in your analytics platform.
At InfoTrust, for any kind of insights, we first use Google Analytics to set up some kind of conversions around what’s being purchased and what type of actions users are taking. We also set up some goals, separate from any kind of analysis in the chat service.
Once we have conversions in Analytics we can start to track basic things like just loads of the chat system or how often it’s being accessed, because that gives us a wealth of information even in the most basic tracking. Just having a pageview or event trigger when a chatbot is loaded gives you things like, for instance, what time of day users are asking questions.
With Analytics, you can create segments to see the user’s progression, such as where people enter into the chatbot, where they’re getting stuck from a functionality perspective, or where there’s some confusing content or page. Segments can identify how users start in one place and move on sequentially. You can build funnels as well. So with the most basic tracking implemented on your chatbot, just when it loads, either with a pageview or event, you can build all sorts of custom funnels and sequence segments to see a user’s progression, drop-offs, or maybe where they’re getting stuck before and after interacting with that chatbot or chat system.
If you have the ability to add even more tracking, such as the ability to see what people are typing or clicking on within the chatbot, or even better what they are clicking to exit the chatbot to go back to the site, (potentially to complete a purchase or update their account, etc.), there’s a lot more information you can grab as well.
One of our clients had an e-commerce platform with a chatbot implemented and they wanted to understand more about it. Is it really driving any conversions or being helpful for users, they asked.
We had the conversion tracking already in Google Analytics for completed purchases. So first we added tracking on the website itself for how people can click to get into that chatbot or load it. Any click on an FAQ button or little chat icon that would load the chat system, we would fire up in GA.
The second type of tracking began once the chatbot was actually loaded. We would track a page view within it. After so many interactions if there would be a refresh or a type of answer provided to the user we could have some event tracking included within the chatbot as well.
But the third, and arguably most important tracking was having a link provided from this artificial intelligence chatbot to take users back to the site to continue through the checkout process or direct them to the product detail page or potentially to direct them to more information – an FAQ page or their account login page. All of those links that would send users somewhere else, we were also tracking via event tracking. This could be done in the chat system as well at the point of click — that’s the ideal way an event system could be tracked. It might be difficult for some systems to add custom tracking within the actual chat system but you do need to provide a URL for users to navigate on to the site. So for our client we added some URL query parameters or extra information in those links that users can click, and on that second page load back on the site those URL query parameters would be read by Google Tag Manager or a tag management system and trigger an event to capture the answer clicked in Google Analytics with information about what that click was and the fact that it came from the chatbot or the chat system that was interacted by the user.
This allowed us to do all sorts of things in GA, and analyze using segments, sequences. We could learn, for example:
- Where people were entering
- Why they were entering from different locations on the site
- How long they were spending in the chat system based on where they’re coming from. Maybe if there was something wrong with their account settings people were entering the chatbot but really spending a lot of time within the chat system. That would mean the artificial intelligence needed to be improved for that type of information or questions
- What time of day people are entering, so potentially we can start adding some human support at higher peak times when there is a lot more activity with the chat system, or make sure we have the processing power in our backend to handle potentially more orders or more users accordingly
Also, we could know where they clicked after the chat system, enabling us to build dashboards in Data Studio or Google Analytics to see what type of answers were most effective and that led to more conversions or higher engagement rates, or returning users – all sorts of information. We could see what browsers or devices people were using when entering the chat system more frequently so, for example, if we would see more on mobile maybe there’s an issue on our mobile devices with the functionality or the type of questions that we didn’t experience with desktops quite as much. The list goes on.
Any automatic dimensions or metrics, of which there are over 200 in GA, will be tracked as long as we have that basic core tracking in the chatbot as soon as it’s launched with an event or page view.We highly recommend any e-commerce platform or any organization that has a lot of users coming to their online platforms to invest in artificial intelligence, or even a basic human intelligence chat system because it really can drive and impact conversions. From our experience we see conversion rates double from those that use a chat system versus not when the answers are tailored to exactly what they’re looking for.