Introduction
In this article, we present some chatbot KPIs you can use to measure how well your chatbot actually performs. For traditional services such as phone and email, performance measurement is already standard. With the rise of AI-based tools, measuring the performance of chatbots, for example, has become essential too.
In the following lines, you'll learn which chatbot KPIs to keep an eye on.
First of all, you should know that chatbot KPIs can be divided into categories. Let's start with the general ones. Then we'll move on to engagement and revenue.
General chatbot KPIs
1. Number of users
It seems like an obvious metric, but its obviousness is exactly what makes it meaningful. It shows you how popular your chatbot is with your website's users.
2. User satisfaction
This chatbot KPI shows you what users think of your chatbot. Useful? Helpful? Engaging? The best way to find out is with feedback buttons or surveys at the end of the conversation. But be careful! You shouldn't overwhelm your customers, so before adding such buttons or a survey, check when the right moment is and which triggers should prompt a rating.
3. Chatbot goal completion rate
If you've opted for an AI- or NLP-powered chatbot, you should also measure its goal completion rate. This is where the fallback rate comes in, i.e. how often your chatbot was not able to handle the request. This in turn creates another variable: the human handover rate. We'll talk about that in a moment.
In other words: with this chatbot KPI, you determine the success rate of the actions carried out by the chatbot.
This chatbot KPI is easy to use and only requires basic math. Subtract the failed messages from the total number of messages. Then divide the remaining number by the total number of messages and multiply the result by 100. This gives you two parameters: how often your chatbot was successful and how often it was unsuccessful.
.webp)
Sounds hard? It isn't!
Assuming your chatbot received 1,600 messages this month and 125 of them were fallback messages, you do the following calculation:
- 1,600 – 125 = 1,475
- 1,475 : 1,600 = 0.92
- 0.92 × 100 = 92
That's it! Your chatbot answered 92% of requests correctly and 8% incorrectly.
This percentage can in turn be analyzed further: in which cases did a human handover occur? In other words, how often did employees have to take over the conversation? If this rate is too high, it means your chatbot has not been trained enough – it may be missing a lot of information and therefore not fulfilling its function. We've already discussed human handover. If you're interested, just click here.
Engagement chatbot KPIs
Measuring your chatbot's potential and effectiveness is a decisive criterion. Here you're dealing with the last link in the chain: the customers. However, a lot depends on engagement. Let's look at which chatbot KPIs play a part.
4. Number of active users
This measurement shows you how many users sent a message within a certain period. For this, you need the total number of visitors to your website. With this number and the active users among them, you get a clear picture of how people interact with your chatbot.
5. New users
How many new users interact with your chatbot? This KPI in turn helps measure your marketing performance and once again shows what your customers prefer.
6. Retention rate
This KPI shows you how many users come back to your website to chat with your chatbot again. That means they found the chatbot engaging and useful.
7. Conversation length
By measuring conversation length, you can understand the average interaction time between your chatbot and your customers. This metric shows the total duration of all conversations in seconds during a certain time frame, divided by the total number of conversations in the same time frame. The measurement takes place at the end of the conversation and counts the time from the start to the end of the conversation – in other words, from the first to the last message. However, don't be fooled: this chatbot KPI is a fairly superficial parameter and not meaningful enough on its own. In fact, conversation length should be used together with many other chatbot KPIs.
The total length of your chatbot's conversations doesn't necessarily say anything about the time it took to resolve the request. Conversations may be short because customers give up in frustration. But if they simply found the information quickly, resolving requests fast is of course a very good thing.
8. Bounce rate
The bounce rate is a very useful chatbot KPI for measuring the number of user sessions that don't lead to a real consultation of the chatbot. In other words, it's the number of sessions in which the bot was opened but not used. A high bounce rate is often due to insufficiently updated content or indicates the wrong placement in the customer journey.
9. Flow completion rate
The flow completion rate shows the percentage of users who started the conversation flow and completed all steps – up to receiving the last message. This chatbot KPI can be calculated by dividing completed flows by the number of flows started. You can also see at this point where users lose interest and therefore where the chatbot needs improving.
A chat flow might look like this:
.webp)
Revenue chatbot KPIs
10. Revenue generated
If your chatbot's main goal is selling certain products, it can bring revenue to your company. This chatbot KPI measures how much revenue the chatbot has generated. The chatbot can make sales right in the chat (which is quite often the case with e-commerce chatbots) or forward users to employees who close the deals. Either way, the chatbot has helped generate revenue.
11. Return on investment
This metric captures the bot's monetary value (e.g. time saved by customers and support staff, or customers won through the chatbot). It also tells you how efficient the chatbot is and whether you should invest in further optimization.
Chatbot KPIs – conclusion
You should now have a good overview of various chatbot KPIs you can use to measure your chatbot's performance and success. Dividing the KPIs into general, engagement and revenue chatbot KPIs helps you keep track. If you'd like explanations of other KPIs or general answers and guides on customer service in e-commerce, take a look at our e-commerce customer service academy.
Because it's not just about the numbers but also about what the numbers mean for your company. It's important to know how many users land on the website and how many stay, but for marketing purposes we also need to understand how the chatbot helps us – whether it's actually used by customers, whether it keeps them on the website and whether it can satisfy them.
We hope we've made clear how you can measure your chatbot's performance, and we hope you'll use these criteria – and think of us while you do!
And of course, the whole melibo team is always there for you if you need us. ✨
