Ben: Chatbots have quickly become one of my favorite pieces of marketing technology. I have found myself quite excited about the possibilities of chatbots, but honestly in the beginning I was not really so sure on how to exactly implement them to improve conversion rates.
While doing my own research on the best chatbots out there, I came across Continually. In my opinion, it was probably the best chatbot tool of the ones that I surveyed, so I reached out to the CEO Paul McKeever to ask if they would be open to doing an interview on my podcast.
Thankfully, Paul said yes!
Paul was very generous to provide a lengthy, in-depth interview on how to really drive conversion rates with chatbots. The recording took about two hours came to about one hour of pure tactical information. You can read the guide below.
One thing to note here is that the content written below is my summary of the discussion which appeared on the podcast, these are not direct quotations of Paul. If you want to hear what he said exactly please listen to the recordings.
The Guide on Using Chatbots for Conversion
This first section of content is based on part one of the interview. You can listen to this part of the interview below.
Why is conversion one of the strongest use cases for chatbots?
Conversion is one of the strongest use cases for chatbots. When thinking about the function of websites overall, businesses are basically trying to give people the information they need to buy things.
So if you think about the challenges that people have in conversion and how to more effectively address the interests and challenges of a buyer, chat bots are an effective way of making a better experience for people in conveying that information.
Are some businesses better for chatbots than others?
From experience, there is a lot of adoption of chatbots for B2B software vendors. Also, there are a lot of marketing agencies who are looking to keep up with more innovative techniques, who are using chatbots. If you are one of these, and not using chatbots, then you really should be.
After that, there is really a broad distribution of businesses that are getting good results. This can be things from professional services such as accountants and lawyers and consultants to offline businesses that you don’t traditionally associate with chat bots such as hairdressers and hotels.
One area where you may want to be cautious with chat bots however is in regulated industries. For example health businesses.
How do you design a chatbot to improve conversion?
There are really two parts to the design of a chatbot to improve conversion.
In part one, you can start with the LIFT model by Chris Goward. Basically, Chris has invented a conceptual model with different elements that affect conversion rates. These are clarity, relevance, distraction, urgency, and anxiety.
In order to design a chatbot, you have to think about these elements and the questions a user has in their mind that you may not have anticipated when you created the page, or, questions that a person may have when they do not entirely understand what you intended. Using a chatbot, you can allow them to ask those questions and engage with them.
In part two, think of the scope or coverage of your website in Google Analytics and ask yourself:
- Where is the traffic going?
- How much traffic are you getting from different sources?
- What campaigns are running to your website?
- At what time are people visiting your page?
With this, you can then start seeing opportunities to improve conversion by increasing relevancy, or clarity, or reducing anxiety or distraction.
Use this information then to generate ideas on where a chatbot can help
Ultimately you will end up with the idea that you need chatbots plural, not just a single chatbot.
Why build more than one chatbot, rather than trying to use a single chatbot?
If you consider your site overall, and then the user journey, you will notice that people are coming into high-traffic pages and they’re not necessarily getting the overall context of your business.
If you only design one chatbot around people coming onto your home page, you missed the opportunities to broaden your scope and serve the areas that need more time and attention. There will be new opportunities to increase relevance and clarity.
Think of it like filling up a class. You can fill up the glass by about 50% with one chatbot. But then you will need to add additional chatbots on top that are for more specific pages or traffic sources, that fill up the glass just a little bit more with each bot.
At some point, however, you will reach diminishing returns, and at that point you will just want to improve the bots that you already have.
What type of language works best for chatbots?
Ultimately you’re going to have to be appropriate and consistent.
However, generally speaking, informal language tends to be a good starting point. It is often very difficult to find a chatbot that has just too much personality.
This is the kind of thing that you will need to run an AB test on. My expectation would be that if you write like a human and add engaging elements such as emojis, GIFs, and pictures you will see more engagement with your chatbot.
Writing engaging chat bot language is a skill of copywriting, as it is for anything.
One other major mistake that I see people make often is that they ride their chat bot messages like an email. You need to be moving to more concise messages over time.
How should you start implementing a chatbot strategy for a website?
Overall, you are going to be looking to start a series of small boats targeted specifically at things people are trying to do.
You can use this three-part building questionnaire for each bot;
- Why would people want to connect with this bot?
- How will the bot build an understanding of the customer and what they need?
- What action is this bot going to help them take?
You will find that trying to do one chatbot using this framework for the whole website will be overwhelming. So you will need to narrow down to one particular customer journey and start there.
1) Why would someone engage with this bot?— Benjamin Boman (@benjaminboman_) June 6, 2022
Knowing why users would engage with the bot will help you plan what steps you can help them take, and what information you need. pic.twitter.com/uTs88qg9td
The narrower you make it, the easier it will be to design, implement and improve. Think about using the agile process. See what works and iterate from there.
When should you use live chat vs an automated chatbot discussion?
Using live chat versus an automated discussion will depend on the situation.
When people initially start with live chat, they often find it is like turning on a firehose of interruptions. This can be very disruptive unless you have a dedicated member of your team, such as a marketing manager, dedicated to it.
But you will find is there are questions that can be answered by machines and others that can be answered by humans.
For example, if you have a pricing page, those interactions will be high value. Depending on your product or service’s price point, you would probably pick up the phone to answer questions regarding this page. So these conversations would be more aligned to using live chat, provided that the user is qualified.
On the other hand, if it is a question about a repetitive business process and expensive to answer over and over again, then it would make sense to automate this. For example, questions about delivery timelines.
This next part of the guide is based on episode 2. If you would like to listen to Paul’s direct quotes and answers in audio form please click the play button below.
Example Case Study: Low Price B2B SaaS
Ben: So let’s apply this information in a step-by-step form. I have suggested to pull to use a relatively low price point B2B SaaS vendor as an example.
1. Define Your Objective
So the first step here is to pick your objective. We are talking about conversion here so that will be the objective we use.
2. List your Top 5 Pages
Now, make a list of the top five pages on your website in terms of traffic. For example, this will likely be your home page, our pricing page, a pillar page of some kind, and high traffic blog posts.
3. Brainstorm Opportunities for Chatbots
Now for each page go through the three-step questionnaire described earlier. Think about the visitor’s intent, what you can do to better understand what the user wants, and what is the next action you would like them to take.
4. Design your Bots
With this information, you can now design your bots thinking through how you can use them to make the page information more relevant or clear, reduce distractions anxieties, etc.
For engagement, you may consider using a quick reply format, where a user can simply select one of many options my buttons.
For discovering user needs, you may find that users have three to four common flow paths. A lot of businesses will find that there are some very common business problems that resonate with their customers. Because you are looking to drive conversions here, you need to stay focused on discovering buyer intent with 2-3 questions.
Use GIFs, emojis, and videos if appropriate.
For driving action, think about the next thing they can do to move them forward in the research process. It may be that you want to direct them to gated useful content or route them to a salesperson depending on the value of your product or service.
How should you order the questions on your chatbot?
Generally, these are based on priority for the visitor.
You may find that most people don’t actually scroll full page. So it is common, in my experience, that people ask questions that are already answered on the page. This is for various reasons. Such as, the information was not seen, it didn’t match their expectations, or it wasn’t clear enough.
Don’t worry about content duplication by providing the answer by the chatbot. Ultimately it is a reinforcement of that information which is a good thing.
How do you come up with these questions?
Most businesses have a good understanding of the questions that their customers often ask. You kind of build an intuitive model of what the customers are interested in. This is a good starting point.
When do you use a chatbot instead of a lead magnet?
Lead magnets work well when there is a common frictional pain point for people who will see the offer. It is a low-cost way of pulling people one step further into the research final because it is digital and quick to distribute.
This tends to work well early on in their research and decision-making process.
As someone moves forward in their decision-making, those general references tend to be less persuasive.
I wouldn’t suggest using a chatbot CTA for people who are not qualified and aren’t necessarily ready to buy but you will find there is a certain point of the website, such as a pricing page, or a page that shows they are qualified and motivated. Newlight
At that point, those visitors are more valuable but also qualified, and a chatbot provides a high-impact lead deployment opportunity for people who are further in their buying journey.
What integrations should you be looking for in a chatbot?
There are a couple of key integrations you should be looking for.
A calendar and scheduling assistance are useful for collecting bookings. Also, if you are going to have people interacting with your chatbots, then it is a no-brainer to be able to add them to your email list via something like MailChimp or ActiveCampaign.
Another key feature would be the ability to tag your users which will enable you to run custom automations based on user behavior and interactions on your website.
For understanding user behavior and interactions, you can use something like Full Story and Google Analytics to get the event data on overall interactions with the chatbot.
How do you analyze and measure the performance of your bot?
here are a few points to analyze to measure the performance of your chatbot.
Firstly you are gonna want to understand the engagement of your chatbot. For example look at the engagement of your start buttons, question callouts, or quick reply format responses.
If people don’t get to connect to your chatbot, they won’t see the great flow that you have built out for the engagement.
Things you could improve here would be the copy, the topic or the style of callout used.
I would look at the engagement of the flow generally. Look to see whether you are collecting your number one measure of engagement, such as phone number, or email.
Look to see how many interactions became known prospects. This is a good test to show that your bot is helping people move forward in their decision-making.
With these two points, you can tell that you may be having a connection problem, such that your bot is unhelpful, or your user flow isn’t helping them.
What are some of the biggest mistakes people make with chatbots?
The biggest mistake people make with chat bots is trying to build one chatbot to serve the entire website. This is very hard to maintain.
The other thing is that it is often hard to be disciplined with your chat box, as it is with all marketing. You can always be running more AB tests to be getting ideas on how to improve your performance. My suggestion is to have an operating frequency to analyze and improve your bots over time or else you will lose those gaps.
How much resources should people dedicate to managing their bots?
Dedicating resources to improving the performance of your chatbots is generally a function of team size and structure. Bigger teams can sometimes have a person dedicated to running their chatbots and improving them, making very many changes per week.
Generally, it is better to pick something you can sustain over time. A once-a-month review is something sustainable and practical for many people.
Ben: I hope you found the discussion with Paul interesting and practical. I thought it was a very tactical discussion so I was super happy for Paul to have generously shared, basically, an entire guide to setting up and running a chatbot strategy for a website.
My suggested next step would be to pick one user journey based on the discussion above and test a chatbot. Use his three-part questionnaire to design the bot for engagement, user flow, and action. I’ve included an example screenshot of one of mine that I made with Continually below.
Here are some other resources you may want to cheque out as well.