5 things we learnt building chatbots in 2018

Talking about chatbots is one thing. Actually building and making them profitable is another.

So to be helpful. We have put together 5 things we learnt from building chatbots in 2018.

First up…

Keep it nice and simple

Chatbots are clever. But not that clever yet. Yes, the technology is there for them to have free-flowing conversations with your customers. But, is it actually adding value to your customer or does it just sound good in the boardroom?

Our most profitable bots this year were some of the simplest. They lead customers down the desired path in the quickest time possible.

Something to factor in is that your customers probably haven’t used a chatbot before. Yours might be their first. There’s a reason why companies slowly roll out a technology to the masses. Approach chatbots in the same way.

We’ve found its always better for chatbots to do a few things really well. Rather than a bunch of things poorly.

Have a clear goal in mind

What are you building a chatbot for? To get a promotion or to improve your customers experience online?

Once you’ve figured that out… The next thing is to have a clear goal in mind.

Is the bot to drive sales?
What product? How many units? What’s the timescale?
Is it to answer FAQ’s?
What FAQ’s will you focus on? All of them or 3-4?
Is it to help build your brand online?
Where online? How will you measure success?

To build a successful chatbot, we’ve learnt that you have to have a clear reason why and method to measuring success from the start.

This will help you focus the flow and copy of the chatbot, which is probably the most important thing…

Your tone of voice

What your chatbot says to your customers is important. But how it says it to them is even more important.

Unlike your customer service rep, a chatbot doesn’t have the use of a voice. Text on a screen can be taken in so many different ways depending on the type of person reading it.

So what is your tone of voice going to be? A good start is looking at how your customers talk to you. Do they use emoji, gifs and snappy sentences? Or do they like detailed more formal responses? You can of course test (which you should always do) but its best to have an idea before you start out.

Be where your customers are

Where do your customer hangout online? For most companies, it’s going to be messaging apps like Facebook Messenger or WhatsApp.

In our experience, Messenger is currently the best platform to host chatbots on.

To build chatbots, we ever use IBM Watson or Amazon Lex. Our preference at the moment is Lex.

If you know your customers are on Facebook Messenger (which they most likely are) put it there. If you’re worried about missing some of your customers. Then you can always integrate the bot onto your site.

You don’t have to integrate into absolutely everything

Just because you can, doesn’t mean you should.

Go through your tech stack and work out what has to be integrated and what doesn’t. A lot of the time, it doesn’t have to be integrated to start with. This can be added once you’ve proven the chatbot is doing what it should do.

The bot we created for a university made them £1.3 million in 3 days. It had no integration what so ever.

It wasn’t seamless but it made them 1.3 million in 3 days. So as you can imagine, they didn’t mind too much. Nor did the students care about that one bit. They just wanted a university place quickly. The chatbot provided the easiest solution to the students’ problem

Next steps…

So before you start spending money on your new chatbot. Get a plan together, set your goals and work out how you will measure success.

Keep the bot simple to start with. Let the chatbot prove its worth. Then you can build a really strong business case internally to scale it.

If you’re interested in implementing a chatbot, we’d love to help! Drop us a line on hello@puzzlelondon.com.

Posted by

Alex Dixon