8 Chatbot Mistakes That are Hurting Your Business

{authorName}

Margo OvsiienkoFreelance Growth Marketing Strategist

Tuesday, December 21, 2021

When chatbots work, they can take your customer service to the next level. When they don't, they can frustrate users and derail your CX altogether.

Article 8 Minutes
8 Chatbot Mistakes That are Hurting Your Business

Have you ever been annoyed by a chatbot? There’s a high chance you have.

The reasons for this can include vague answers, the inability to answer a question at all, failing to connect to a customer support agent when you ask for it and much more.

While plenty of companies do it wrong when implementing chatbots, for many, chatbots have been a success helping boost sales, user engagement and reducing the number of customer support tickets.

In this article, we will explain why chatbots fail and how to build a chatbot that works. Before we dive into the topic, let’s first understand what makes a chatbot a success or failure.

Chatbots: Successes and failures

Website chatbots were all the rage in 2016 and everyone thought they were the future/ However, it didn’t take long to learn that a lot of them didn’t succeed at their job.

For example, Facebook’s M assistant failed to handle over 70% of user queries, meaning that the majority of responses were later directed back to customer support agents. It was eventually closed down later in 2018.

Various news platforms such as the Washington Post or Huffpost experimented with chatbots as well – in most cases, unsuccessfully. There have been a lot of chatbot failures, but there have been successes as well.

LEGO reported a 6x return on ad spend and a 31% lower cost per conversion after introducing the Ralph chatbot on their website.

Screenshots of Lego chatbot in action

Also, who doesn’t know about the Sephora bot?

Screenshots of conversations with the Sephora chatbot

This technology helps users to reserve a makeover at Sephora stores. Thanks to this bot, Sephora managed to increase in-store bookings through the chatbot by 11%, with customers spending over $50 for each visit.

The failure examples prove that while it’s still relatively new, chatbot technology isn’t always the right tool; it also seems that many companies do something wrong when deploying their bots to a wider audience. However, that doesn’t mean you can’t make it work for your business. Some brands have noticed a good ROI after deploying their bot. So why can’t you?

To increase your chances for success with a chatbot, you have to know what has doomed other chatbots to failure. Let’s do some reverse engineering.

1. Bad planning

It’s easy to get annoyed with a chatbot that doesn’t get a user right, repeats the wrong answer over and over again no matter what way a user phrases a question. Such situations are often caused by poor planning.

For an effective chatbot, you have to think of many scenarios for questions and responses to train your chatbot. You also have to account for the questions you might not know the answers to as well as those you simply can’t think of now – customers can ask all sorts of weird questions (as on the screenshot below).

A screenshot example of a chatbot conversation with a bot unable to answer a question properly

In such cases, you have to know the right time to connect to a live chat letting support agents handle the situation.

How to fix it: Instead of rushing to deploy your chatbot, spend a bit more time on planning. Use a pen and paper approach to visualize what questions users can have and what responses your bot can provide.

2. Not offering an option to talk to a human

There are a number of questions where chatbots capabilities are limited. While this tool can be great at resolving simple tasks, it doesn’t have unlimited capabilities and knowledge (at least for now). Not understanding the right moment to connect a user with a human being, can be harmful –  it makes users annoyed.

How to fix it: Think of the scenarios when you have to suggest a user to be directed to customer support. Often, users ask to be connected with a customer support agent by typing the sentences such as “Connect to a human”, “Could I talk to an agent?” Make sure you can identify these phrases and give users what they ask for.

Here’s how your response can look when users are asking to be connected to a customer support agent.

An example of how a bot can connect you to a human agent

3. Chatbot sounds unnatural

Chatbots are often perceived as robots without emotions, character and a unique communication style. They shouldn’t! Talking to a chatbot can be a lot of fun if they are programmed to read between the lines, fetching an emotional element of user queries. If they learn to understand the emotions, they can also respond in a way humans would do – with some sort of emotion, expression and character.

AI-powered chat widgets offer a capability to learn from previous interactions with customers and identify emotions faster. They can also be trained to acquire some character and communication style.

How to fix it: Give your chatbot a name, create their character that will be reflected in the communication style and make it conversational.

4. Not setting KPIs

For many companies, chatbots are just some nice add-on on their webpage that should offer a customer another way to connect with the company. However, chatbots can make a huge difference for your company’s bottom line if you approach this technology with some strategy in mind.

Introducing a chatbot should be done the way a marketer runs a digital campaign – you need to define your KPIs first and only then can you proceed with deploying a campaign. Developing an app costs – no matter if you develop your own or use an off-the-shelf solution. Without setting KPIs, you won’t know if a chatbot improves your business health or remains an expense.

How to fix it: Think of what a chatbot can help you achieve – whether you expect it to increase your page traffic, revenue or deliver a better user experience to customers.

5. Providing a long explanation

Users want responses right away when they decide to use a chatbot. If they had some time to look for a solution on their own, they would spend it reading your knowledge base or a FAQ page.

That’s why, when creating scenarios for chatbot responses, you shouldn’t copy the responses from your knowledge base. Write concise sentences that give a solution to a user on a silver platter.

How to fix it: Don’t make your clients bored with long explanations and dozens of sentences, as with the example below.

An example of a poor chatbot response that uses boring and overly long sentences

Keep your sentences to the point, taking not more than 30% of the chat window. Don’t get clients to scroll a chatbot’s message back and forth.

6. Not improving your chatbot

Over time, as you collect more information about interactions with customers, your chatbot has to offer more targeted responses that help resolve user queries faster. You have to constantly iterate on your chatbot response logic and improve it. Remember that as with any technology errors happen with chatbots too. If you manage to identify them as quickly as they appear, you’ll still be able to delight most of your chatbot users.

How to fix it: Run retrospective sessions and think of ways how your chatbot can work better. Test it on a regular basis and collect feedback from customers on things to fix.

7. Not using visual elements

When you talk to a human being, gestures and facial expressions aid understanding. When typing a message online, you don’t get those visual clues, so talking to a chatbot often feels impersonal and even boring. Just think how you communicate online using text – we would normally use some emojis or GIFs to show a recipient our emotions where we can’t do it using gestures.

Why not get your chatbot to use the same approach?

How to fix it: Add emojis, pictures or GIFs to chatbot responses that fit the context well. Aim at making responses more conversational, cheerful and positive. Here is how Jamie Oliver’s chatbot is doing it.

Jamie Oliver's chatbot using visual elements in its messages

8. Deprioritizing sales

Website chatbots are a great tool to help customers get answers to their questions and find solutions to problems. However, they are not only made for this. When done the right way, chatbots can also become a powerful sales tool. Remember we mentioned Sephora bot in the beginning? This brand is made to boost sales while improving user engagement with a brand.

While chatbots can help customer support, improving customer care shouldn’t be the only reason to implement this technology. If you’re not using it for sales, you’re missing out on many business opportunities.

How to fix it: Think of the best moment when you can offer a customer a complimentary product. If your customers are asking about a delivery date for the shoes they recently bought, you should offer the products that would go well with the previously ordered shoes (but only once a customer receives the answer to the questions about the delivery).

Wrap up

Chatbots can help improve your business health, reduce the number of support tickets by directing clients to the right resources, boost sales by suggesting certain products suited to a customer profile and improve user experience with your brand. However, there are many common pitfalls that companies usually make with this technology. Hopefully, by using the above tips, you’ll be able to avoid them and make chatbots work for your brand as well.

Margo Ovsiienko

Margo is a Freelance Growth Marketing Strategist. She creates content that converts website visitors into paying customers for SaaS companies and tech agencies by building sales funnels. You can read her posts on the blog margoleads.com

Comments

Join the conversation...