The use of chatbot is now more and more widespread, Gartner predicted that in 2020, 40% of mobile interactions were through automated conversations. Though chatbots seem to be well-received by customers, it's not all sunshine and rainbows. If the chatbot experience ends up being negative in the customer's eyes, nearly 73 percent won't use that chatbot again.
If you plan to build a chatbot or if you already have one, but it doesn’t work as much as you hope, keep an eye open for those mistakes, we will tell you how to resolve them.
1. Not integrating chatbot into your overall marketing strategy
Whether you have a SaaS company or an eCommerce store, your chatbot won’t have the same purpose. For a SaaS company, your bot could be a lead acquisition tool that asks for the prospect’s contact information, and for an eCommerce Store, your bot could help you upsell products. You can even push it further and say that a chatbot has to be different depending on which platform it's deployed.
Indeed, your users will be on different steps on their customer journey and you have to adapt your marketing strategy - and chatbot - accordingly. If your prospect is on your Facebook page, they may only look for information about your business, but if they are already on your website, they may be ready to go to the next step of the funnel.
-> Plan why you want a chatbot and what its purpose is
Is your chatbot a sales assistant? Do you want it to upsell products? Or just answer questions? Before building your chatbot you have to define a clear purpose for it.
Define what issues your clients are having at every step of the funnel and ask yourself if a chatbot is the best way to solve those issues. If the answer is yes, then go ahead and start creating your chatbot.
2. Not anticipating multiple questions or weird queries
Users are unpredictable, they will have several ways to ask the same question or they will ask several questions in one message.
This can be quite difficult for your chatbot to understand, to keep your customer experience good you should prepare for this. This can be done through NLP (Natural Language Processing), a subfield of Artificial Intelligence. Your chatbot will be able to understand what your user is asking and redirect them to the appropriate answer.
-> Rephrase user’s questions
Make your chatbot rephrase user questions and ask them to confirm. This ensures that your customer queries are answered correctly and will make your customer satisfied and cared for.
4. Making the conversation too stiff
One of the advantages of chatbots is that it makes the client’s experience more fun and engaging. For example, instead of having a long and boring FAQ page, you could make your chatbot answer those questions.
But, one question can be asked in many different ways and your chatbot should be able to understand that. For that, you could use AI-based chatbot.
-> Use Dialogflow or another NLP provider
AI-based bots can analyze users' input and respond accordingly. As we said, this is called NLP (Natural Language Processing). They are able to understand the variations of questions and improve the quality of their answers. You can easily add NLP to your chatbot with Ideta, in just a few clicks, and without any coding knowledge.
3. Not having a message when the bot doesn't understand something
As we said, users are unpredictable, they will ask questions your bot won’t understand. Don’t overestimate a chatbot's capabilities, it may have AI, but it is not an all-knowing genie straight out of a magic lamp. Users know they are talking to a robot and they don’t expect your chatbot to have answers to every question.
-> Offers them to talk to a human
When your chatbot is stuck, you can try to ask your user to reformulate their queries but there will come a time where your chatbot won’t be able to answer them, no matter what. When this is the case, offer an alternative. You can make your chatbot pass the conversation to a real human support agent. Humans can take over and answer your client directly through the same channel.
5. Making answers too long
Nobody wants to read a message worthy of a novel, especially when it comes from a robot. More often than not, people are using your chatbot on their phone, you don’t want to make them scroll to read your message. This is a poor user experience and will make your user stop using your chatbot
-> Make shorter answers and use quick answers
It is generally recommended to keep your texts under 80 characters and to intersperse them with short pauses, as we would do in everyday texting. Make use of quick answers and ask questions between texts to keep the interaction going and not bore your clients.
Even if your chatbot is not human, you should try to keep the interaction as human-like as possible. Try adding emoji, gifs, pictures and more to illustrate what you're talking about.
6. Set up your chatbot without testing
Nothing screams more unprofessional than a bad, uncompleted chatbot. You should deploy your chatbot when it’s at least usable and keep improving it along the way. It’s better to start with a simple chatbot than with a complex and unfinished one. Moreover, you can tell your chatbot user that your bot is still under improvement and ask them for their feedback.
-> Test your chatbot first
Testing your chatbot before it goes online will highlight any potential errors and allow you to correct them before your customers or prospects stumble on those errors. This will prevent you from creating disappointment or frustration and not getting the expected return on investment.
In order to track down all possible and imaginable bot errors, try to do all the scenarios for which it is normally designed. Let your friend and colleagues try it too, they may stumble on a case your didn’t think of.
7. Not analyzing data
It is impossible to know if your chatbot is efficient and profitable if you don't measure its performance. Identify the KPIs that make sense and gradually adapt them according to the results. Above all, look for sources of dissatisfaction so that you can continuously improve your virtual assistant.
The opinion of your customers can allow you to better understand their expectations and offer the most satisfactory artificial intelligence service possible. To collect the opinions of your users, there is nothing like a quick survey at the end of the conversation.
-> Integrate analytics tool and use dashboard
First, you have to identify the right KPI for your chatbot and your business. Do you want to increase customers' engagement? Then, maybe you should track how many conversations a day your chatbot has. Is your chatbot for eCommerce? Then, track how many sales it helped make
Once that’s done, you can easily add analytics tools, you can connect your chatbot to API, like Zapier, and send the information to Google Analytics or even your intern tool.
Some chatbot builders, like Ideta, even have an integrated analytics dashboard, thus in a glance, you can see how many conversations your chatbot had and how long they lasted
8. Being too pushy and not letting your user a way out
While it is important to boost customer engagement, some bots will try too hard, they will send unsolicited messages, they will be too pushy and ask for customer information when it is not necessary. This is really irritating and will push your customers aways.
-> Give the option to stop the conversation or restart
Instead, let the conversation flow naturally, ask customers for their private information only when necessary, and always let the user a way out or let them restart the conversation whenever they want to.
It’s quite impossible to fully avoid errors on your chatbot, however, you can reduce the number of errors with proper planning and testing. Choose your chatbot platform carefully, some builders have integrated tools like an Analytics Dashboard to help you track your bot performance.
Don’t be afraid to ask for your customers' opinion about your chatbot. These feedbacks are crucial to improve your chatbot and it shows that you are a proactive, pro-user business, and your audience will trust you.