Chatbot Analytics: 15 Metrics To Track for Best Chatbot Performance

Chatbots find more and more applications and evolve, so it should be no wonder that the global chatbot market size is projected to grow to $1.9 billion in 2027, while consumer spending over chatbots is projected to reach $142 billion in 2024.

To reap the sweet fruit of enhanced revenue, you need to keep track of your chatbot’s performance. Today, we are looking at both key and advanced metrics for rule-based and conversational chatbots.

If you are totally new to chatbots, read our article on what a chatbot is first.

What are chatbot analytics?

Chatbot analytics refers to the process of collecting, measuring, analyzing, and interpreting data related to the performance and interactions of a chatbot.

A successful chatbot is characterized by a low unsubscribe rate, sustainable audience growth, and high engagement. If your chatbot is not quite hitting the mark yet, you need insights, and this is exactly what you can get from looking at certain performance metrics. But there is more to it.

Let’s talk in detail about why else it is important to keep track of chatbot performance.

Why it is important to keep track of chatbot performance

The main motivation to keep track of chatbot performance is to identify whether your chatbot is getting enough traction to help you reach your business goals. But of course, there are other reasons, too. Let’s take a closer look at them.

Measuring chatbot usage and engagement

Low engagement and rare interactions often result in unsubscribes or a stale user base. That’s why you need to monitor engagement and make sure it grows or, at least, stays the same.

A chatbot should be useful — help users complete their tasks, learn more about what they find interesting, or choose your products. That will make engagement grow organically. Not only will it help you reduce the number of unsubscribes, but also will eventually provide you with sustainable revenue growth.

Identifying areas for improvement

Some metrics help detect bumps in the chatbot flow that might prevent you from getting conversions. For example, users might get confused by a complicated selection process or a request to upload personal data.

Other metrics provide data on customer satisfaction. Businesses can use it to improve the overall quality of service and, as a result, grow retention — make clients come back and buy again.

By tracking a few metrics at once, you can get a better understanding of what areas require your attention right now.

Benchmarking performance over time

Make sure to check the key metrics after marketing campaigns and user acquisition activities to measure their efficiency and repeat the best practices in the future. Comparing the metrics for different activities can also help detect seasonality and notice other patterns in chatbot user behavior.

Proving chatbot ROI

Creating and supporting a chatbot requires time and money, so a chatbot must prove its value for the business to remain a part of the sales funnel.

By looking at the number of purchases made via a chatbot and the company’s outreach, you can calculate the chatbot’s return on investment (ROI).

Chatbot metrics to track for best chatbot performance

The choice of chatbot metrics to track comes down to what type of chatbot you are using and what is your overall strategy. We will be covering metrics both for rule-based and conversational chatbots falling into three categories: users, user engagement, or bots’ performance.

User metrics

User metrics can give you an idea of how many potential clients you can get if you keep increasing engagement.

Total number of unique users

Depending on the chatbot platform you use, the total number of unique users is calculated based on unique devices, unique IP addresses, or shared contact information like phone numbers or email addresses.

Regardless of the specifics, this metric provides a close-to-reality number of individuals who use your chatbots.

Having this information, you can estimate your chatbot’s overall popularity and analyze factors influencing its growth or decline over time or within specific periods.

Knowing the total number of unique users, you can also calculate conversion.

For example, if you have a chatbot that handles pizza orders, you can calculate the conversion from a chatbot user to a customer in a given period.

Let’s say the total number of your chatbot unique users in November is 50,000, and the total number of pizza buyers is 5,000.

Divide your pizza buyers by chatbot users: 5,000 / 50,000 = 0.10. Then, multiply the result by 100: 0.10*100 = 10%. This is your conversion rate from a chatbot user to a customer.

New users

The “new users” metric refers to the people who started using your chatbot only recently. The recency will depend on the nature of your business and can be anything from “subscribed today” to “using your chatbot only for three months.”

This metric helps measure the efficiency of user acquisition activities, like social media campaigns or paid promotions, whether it is Facebook ads or partnerships with influencers.

For example, if you actively promoted your products for the whole month of April, you can compare the numbers of new users in March and April to see if there is an increase and detect whether your campaign was successful.

You can also use the acquired data for segmentation: create a segment for users who joined your chatbot during your April campaign and send them personalized deals or content to move them closer to conversion. You can also ask them about their interests and needs to go for a deeper segmentation.

Engagement metrics

Measuring chatbot engagement allows businesses to detect whether a chatbot flow is aligned with users’ needs and drives them closer to conversion.

Active users and retention rate

The “active users” metric refers to users who interacted with your chatbot in a given period, while the retention rate is defined by the fact that users keep coming back to use the bot again.

The number of active users provides you with insights into the quality of your subscriber base. By extension, you can estimate engagement and retention since you can see how many users subscribed to the chatbot and forgot about it right away vs. how many still use it.

In Selzy, you have the information about active users front and center on the page that lists all your chatbots — right next to the total of chatbot users.

The data is provided for the current month, so if you notice that the number of active users has dropped compared to previous months, you can act on it and prepare a re-engagement campaign or send a survey to figure out what can increase retention.

Selling to an existing user is cheaper than attracting a new one – retention is cheaper than acquisition. For this reason, it’s very important to stay connected with the people who have already subscribed to your chatbot or used it previously and find ways to learn how to improve their experience and drive them closer to conversion.

Total number of conversations

The total number of conversations tells you how many times users open the chatbot widget or engage with it in the messengers. All the conversations are counted, even if they only include the welcome message. This metric helps estimate the chatbot’s outreach.

Total number of engaged conversations

Unlike the previous metric, the total number of engaged conversations counts only the conversations that continue after the welcome message. By tracking both, you can estimate whether users find the chatbot helpful and understand how to interact with it beyond pressing “Start” or sending the initial request.

There is no specific rate that requires your immediate action, but you can check the difference between the two metrics for some time to get better insights.

Average conversation length

Average conversation length refers to the average number of messages a user and a chatbot send each other during one conversation.

Unfortunately, there is no set window for this parameter. It might differ dramatically from chatbot to chatbot and depend on the chatbot’s primary goal.

For instance, a chatbot with a simple goal of selling airline tickets might have a comparatively short optimal average conversation length. If conversations take too long, it might mean users don’t reach the intended point in the scenario on time, and the flow needs to be adjusted.

At the same time, a chatbot that sells airline tickets, books hotels, and recommends city attractions and restaurants might have a much longer optimal average conversation length, which, on the contrary, will indicate high engagement.

Bot metrics

The bot metrics directly measure chatbot performance.

Goal completion rate (GCR) and total leads captured

Goal completion rate (GCR) shows what percent of interactions with your chatbot ended up successfully. You can set up multiple goals, like purchases, scheduling a demo, clicking the CTA, etc., and calculate GCR for each. The higher the goal completion rate is, the better.

The total leads captured metric can be an easily measurable example of GCR.

Leads are potential clients with their contact information. Businesses need to capture that information to drive chatbot users closer to conversion via chatbot. A higher number of captured leads indicates highly effective flow and welcome copy.

Selzy collects contact information captured via chatbots, puts it in an organized list, and allows sending email campaigns right on the spot.

In the list, you can see the date of subscribing, the date of last activity, and tags providing additional information that can later be used to create and send personalized campaigns. For example, in the screenshot below, the contact is tagged “from chatbot,” which might mean that this subscriber would need an alternative welcome email designed specifically for chatbot leads.

Flow completion rate (FCR)

Flow completion rate (FCR) is the percentage of users who went through all the steps of a chatbot scenario and completed it successfully. To calculate FCR, divide the number of flow completions by the number of flow initializations.

Unlike the goal completion rate, the flow completion rate includes the conversations where your goals were not completed, but the flow still worked properly. For example, when the client checked out the new collection and the prices but then quit the chat instead of making a purchase.

Human takeover rate (HTR)

Human takeover rate (HTR) measures the number of times the chatbot forwarded a conversation to a human.

High HTR may indicate low efficiency of your chatbot and signal that the users don’t understand the flow and need help. But in some cases, human takeover may be an organic part of the flow when the client is forwarded to sales, area experts, or other professionals.

One such example is Talkpush, a chatbot for recruitment automation. It allows companies seeking candidates to customize the flow and choose how much information they want to collect before scheduling an interview with a human HR specialist.

T]he chatbot collects through fill-out forms, voice messages and video presentations and then passes it to a recruiter. So, in this case, high HTR indicates positive results.

Customer satisfaction score

Customer satisfaction score is usually calculated through post-interaction surveys where users rate their experience with a chatbot. It can also be based on common customer satisfaction metrics such as CSAT and NPS.

By tracking the satisfaction score for different flows, you easily identify those that need improvements.

Average response time and missed chats

Average response time refers to the time a chatbot needs on average to process an inquiry and provide an answer. To calculate it, divide the total time for all responses by the number of inquiries.

If the chatbot does not respond at all, this is interpreted as a missed chat. This metric counts the number of times when the chatbot didn’t take the user to the next step of the scenario or failed to switch to a human (if that’s an option). This often means there is some kind of underlying technical error.

Fallback rate (FBR)

Fallback rate (FBR) measures the percentage of messages the chatbot fails to interpret, recognizes the issue, and falls back to a pre-written answer, previous response, or suggests going back in the conversation.

The stages of flow where the fallback rate is especially high detect the areas that can be improved through training the bot or adding more pre-written content.

Choosing a chatbot platform for analytics

Сhatbot builders come with different functionality around analytics. Here is a list of criteria you can use if you haven’t picked up one yet and need to make an informed choice.

Key metrics tracked

Make sure the chatbot platform you choose provides all important metrics for your business. Most metrics we covered in this article are calculated based on the number of users, messages, and clicks on the links or buttons. Make sure your chatbot platform tracks at least those.

Real-time data

Real-time analytics updates can help immediately detect when something goes wrong so that you can fix it. For example, if the “new users” metric goes down dramatically, you might want to check what’s up with the starting the bot part of the scenario or the welcome message.

Selzy updates chatbot statistics every 5 seconds, so you can get live updates on things like how many users are checking out your bot after you shared the link at the webinar. You can also track how the number of users currently interacting with your bot fluctuates depending on the time of the day.

User experience

Different aspects of user experience might be more important to you than the others. For example, if you need a few people to have access to the analytics at the same time, you need to make sure the builder platforms can provide you with extra seats.

Others might be looking for a mobile-friendly alternative to be able to check the stats at any time, wherever they are.

For rule-based chatbots, it pays off to have the key metrics, like active users and clicks, displayed inside the scenario. That way, you can monitor each step of the scenario and get valuable insights on where to improve the flow.

In Selzy chatbot builder, you can see how many users received messages and how many users clicked the buttons (if there were any) at each step of the scenario. You can also see how many people are currently interacting with the scenario and where they are. As we mentioned above, when the scenario is up and running, all the stats get updated every 5 seconds.

Support

Whether you are new to chatbots or looking for an alternative builder, there might be times when you require help with the functionality, including the analytics, so spend some time investigating the quality of the provided support.

Ask some questions to see how quickly you’d get help and how useful the provided information would be. This is also a great way to learn more about the platform and assess where you’d be comfortable using it in the long run.

Data integration

When you run multiple chatbots on different platforms, like Telegram, Facebook, and Instagram, in addition to the one on your website, it can help a lot to have the data per each channel displayed in one place.

In the screenshot below, the chatbot analytics dashboard displays the net promoter score (NPS) for Whatsapp, Facebook, and website chatbots in one list.

Data visualization and customization

Widgets and infographics can help you analyze data more quickly. Some chatbot builders have full-fledged analytics dashboards where you can have various rates calculated for you and put front and center.

If you have complex chatbots — conversational or hybrid — and want to track different metrics at different times, opt for customizable dashboards, where you can choose what data to display.