Businesses and customers love chatbots. An automated, interactive tool that lets customers get their questions answered at any hour of the day without costing businesses an arm and a leg for 24/7 live agent support? The future is truly here! 69% of consumers prefer to use chatbots because they deliver quick answers to simple questions.

As much as customers love chatbots, companies love them too. You can use chatbots to cut costs, figure out where to upsell, and much more. 57% of executives said that chatbots bring significant ROI with minimal effort. Their speed and consistency of response, the fact that chatbots never sleep, and their ability to quickly look up information are all reasons business owners love them. 

Now that you know how helpful chatbots can be, you’re ready to take the plunge. You’re adding a chatbot to your website! But how do you create a fantastic chatbot? Follow these simple guidelines, and your chatbot will be fire. 

1. Choose what type of chatbot you’ll create

Before you do anything, think about what tasks you want your chatbot to handle. Creating an everything bot will probably be frustrating for you and your customer. It’s hard to know what everything encompasses, and it’s even harder on the customer if they are trying to guess what your chatbot can and can’t do. 

Avoid having customers run screaming into the night by choosing a chatbot type. And don’t worry, if you end up needing a few different chatbots, that’s much easier to set up than a single chatbot that has to handle everything. According to Accenture, there are four types of chatbots: 

  • Informational: This type of bot helps customers and employees with their inquiries. It researches and returns documentation, support forum entries, and programmed responses to popular support questions. If it has access to customer information, it can try to guess what products and issues the customer faces and use it to increase customer engagement or even sell another product or service the customer might need.
  • Transactional: Chatbots for transactions help customers place orders and make purchases. They might also help a customer get a refund if a delivery didn’t arrive, or they no longer want a software product, or similar. 
  • Enterprise Productivity: Are you always messing up scheduling a meeting? Hand it over to a chatbot! This use case is relatively new, but this type of bot helps internal processes. The bot can help set up your meetings or collect data you want to analyze, like a marketing or sales campaign. This type of chatbot can also help new employees onboard by answering their questions.
  • Device Control: You’ll typically find these bots in the home. When you ask Alexa to add broccoli to the shopping list, that’s an example of device control. Or say you ask Siri for directions to an office for an interview; that’s another example of device control. 

2. Give your chatbot a personality

People enjoy talking to chatbots with personalities, and this can help with branding. Everything from the words your chatbot chooses to its avatar affects how it will be perceived. While character is essential to the overall experience, don’t forget that 48% of customers prefer a chatbot that solves issues over a chatbot that has a personality. However, imagine if your chatbot has personality AND solves issues – that’s the total package! 

3. Have your chatbot introduce itself

A chatbot introduction can make it feel human and help to establish a better rapport with a customer. But don’t stop there. Explaining how it’s designed to support and acknowledging its limitations will help customers type questions that get successful answers. 

Your introduction can also include a menu of choices to give the customer an idea of what kind of help or service will be provided by a chatbot. For example, a chatbot that says, ‘I can help you find the right support documentation for any product, type something!’ makes it clear what it does best. 

The chatbot’s introduction should also include any requests for information that it might need for a personalized discussion. For example, the chatbot might say, ‘Hi, I’m Bob. I can help you with basic support questions. Before we get started, what’s your name?’ Just think in 2001 Space Odyssey how much more comforting it was when Hal 9000 said, ‘I’m afraid I can’t do that DAVE’.

4. Have the chatbot suggest options 

Whether your chatbot lets the user type anything, or provides preselected menu items, make sure you guide the customer. If a customer can type anything, the chatbot should constantly make statements that indicate what’s available to the customer at that point in time. 

Think of it like an adventure game but instead of the chatbot saying, ‘You can go east or west, and there’s a large tree here,’ it says something like ‘I can help you choose the right security camera or troubleshoot a camera you’ve already purchased.’ Then the customer knows what they can ask about from that particular prompt. 

The best way to suggest options is to create a map for yourself of the different choices and where they lead. Then you can build your script for the chatbot around the map. 

5. Make your chatbot exciting and conversational

Keep the language the chatbot uses easy to understand. A conversational style is best, and a dash of humor is almost always welcome. But be careful with humor if a customer is troubleshooting an issue, as humor when a customer is frustrated doesn’t always land as you might expect. 

6. Personalize the chatbot whenever possible

A customer name is a good start, but you can go even further by giving the chatbot access to information about the customer. If appropriate, providing the chatbot access to some account details like what the customer bought or their last support ticket can help the chatbot offer the best choices. 

7. Test the chatbot

Your chatbot won’t be perfect, so make sure you have a way to track its use. How do people engage? Where do they drop off? Are there any points where customers get mad at the chatbot? Any moments where they seem pleased with the chatbot’s help? By tracking this information, you can determine what needs to be rewritten or fine-tuned in your chatbot script. 

Your chatbot is now live, but is it helpful?

Once your chatbot works as expected, the next step is to start monitoring your chatbot. What’s the best way to track the success of your chatbot? In places where your chatbot succeeds, you’ll want to double down, and in areas where it doesn’t, you’ll want to analyze the information to figure out what needs work. 

Let’s dig into some of the best metrics to track when analyzing the success of your chatbot. Not all metrics in this guide will apply to all chatbots; pick what’s right for you. 

1. Support ticket resolution

Check which tickets the chatbot handled and closed vs. how many times a customer asked to speak with a human for help. 

2. Chatbots in different channels

A chatbot will be most helpful in channels with high traction but also the right kinds of questions. For example, if you have a Facebook channel where people only ask for product demos, that might not be the right place for a chatbot that answers support questions. You could either add a chatbot that helps schedule product demos or look for another spot to add your support chatbot. 

Matching the right chatbot to the right channel is part of ensuring chatbot success. Determine your customer’s expectations of a particular channel before moving forward with anything automated.

3. Bounce rate 

You can track how often customers walk away from a chatbot. The higher the engagement, the more relevant the chatbot’s questions were. If you continue seeing a high bounce rate for the chatbot, you should rethink the chatbot’s introduction and timing. 

Maybe the chatbot isn’t effective on certain pages. Instead, have the chatbot wait for the viewer to go to a page that’s a popular spot to ask questions. Then it seems like the chatbot anticipated the person’s needs, making it appear more useful while increasing engagement and lowering bounce rate. 

4. Conversion and retention rate

When the chatbot describes products the customer can buy, does it lead to them buying or signing up to try out the product? Check conversions and figure out how the chatbot can bring in additional high-quality conversions. I say high-quality conversions because more conversions are not necessarily better conversions if the customer quickly opts out after a trial period. 

5. Total number of users

See how many people use the chatbot. Does it increase over time? Do customers return to use the chatbot more than once? Use this information to determine the success of your chatbot.

6. Number of chats over a specified time frame

This measure can be helpful for many reasons. Not only can you see if the chatbot is popular and increases in popularity, but you can tie it to company events. Did you run a marketing campaign and then see a massive uptick in chatbot conversations about the product mentioned in the campaign? That’s great – the campaign was a success, and now the chatbot is successfully helping you take your marketing further by answering more customer questions. 

Or the chatbot answering questions might be tied to a problem with the service. If many people feel comfortable asking the chatbot for an update on a service issue, that’s a sign that the chatbot is helpful to your customers. 

7. Length of conversation

It’s essential to determine how this metric works for you. For example, if it’s a support conversation, you might want short conversations with favorable resolutions. But if the chatbot is supposed to explain products and interest new customers in getting a demo or buying something, those chats might be longer. You may still want to fine-tune the bot for the shortest conversation to schedule a demo or purchase something. 

8. Completion rate

How often do customers abandon your chatbot because they get stuck? How often does the chatbot reach the end of a script that means a successful resolution was achieved? This stat is a fundamental metric that can help you determine where your chatbot knocks it out of the park and where it could use some work. 

9. Fallback rate

A fallback rate is how often the chatbot misunderstands the customer. This experience is frustrating for the customer and not good for you. You want your chatbot to understand and help the customer or swiftly get them over to a human agent. 

10. Activation rate

An activation rate is how often a customer completes a conversation flow with a chatbot but then stays because the chatbot asks a helpful question that causes the customer to continue engagement. This moment can be great for marketing or upselling new products. 

Success is just around the corner!

There are many metrics you can use, but the above is an excellent place to start. Even using a handful of these stats can help you figure out how to improve your chatbot. 

You can gather additional feedback on customer experience with a quick survey. A simple ‘rate this session 1-10’ or an even simpler ‘was the experience good or bad?’ can provide valuable insights. A rule of thumb is to keep it as simple as possible, or customers may wander away without completing your survey. 

It also doesn’t hurt to go through your chatbot’s conversations and see what gets asked the most. Then you can spend more time enhancing the chatbot scripts for those questions. Or you can create additional support and content around those questions as part of marketing campaigns since there seems to be a lot of customer interest. 
Chatbots are an essential customer experience tool, and as always, Mitto is here to help! We can help you figure out what type of chatbot is right for you and set it up using our Conversations dashboard. You’ll be up and chatting in no time!