What Makes a Good Chatbot Service?
This new wave of chatbots and artificial intelligence has been going for what, a good 18 months now? Why don't we take a critical look at some of the positives, a few case studies, results and examples of good chatbot service.
You can count the chatbots that have had any form of traction on one hand, and by traction, I'm talking more than 1 million users.
You've got CNN and their 4.7 million Line messaging app users, Poncho and their weather updates, the KLM customer service chatbot and Swelly with their 'this or that' chatbot.
Really, with public data, that's about it. I'm sure there are many hundreds of more chatbots that see huge numbers (after all, we've delivered a few of them ;) ) but this is all we have to go on right now.
What can we draw out of this public information? (I'll be giving some insight into our non-public data later...)
The best chatbot services are niche
And, by that I mean, really super-targeted.
It shows the importance of getting a chatbot's one true goal nailed before even starting to build anything (which btw, we talk about when working with a chatbot development company). The very best chatbots not only delight, but they actually do what they're supposed to do and do it better than any other channel can.
The CNN Line chatbot delivers videos, KLM does customer service, Poncho tells you about the weather, and Swelly makes you choose between two things. Simple, functional and better (and faster) than an app or using a search engine.
So here we are, lesson 1. Make sure your chatbot does what it is supposed to do better than any other way of doing it. Period. Need help for that? We wrote on how to benchmark your chatbot's performance.
Make it fun
If doing something great is not an easy option (perhaps you're up against a high-traction mobile app or a popular website), then you better make sure you do something fun and use artificial intelligence well.
As an example, Unilever wanted to raise 1 million laughs for their PG tips' Monkey contribution to Red Nose Day 2017. How did they make this chatbot service successful (other than having us build it)? They made it fun and a memorable experience. Monkey talks to you, he tells you jokes and makes you laugh. He doesn't rreeaaallllyyyy, 'do' anything, but lots of people engage with him every day. According to a world leading marketing and ecommerce website, it was a resounding success and one of the best chatbots they've seen.
So, lesson 2. If your chatbot service can't be better than any other way of doing the same thing, make it a much more fun way of doing it.
Timely content
When we look at how chatbots are delivered, as in, the messaging applications, there are some valuable insights.
Johnson & Johnson scored a massive 84% read rate and 53% average click rate via their Facebook Messenger chatbot. Compare this to what, a ~20% read and ~2% click rate for email? Winner.
How was this achieved?
Well, it had a little to do with the whole niche thing we talked about earlier. But, it was more than that. It was also down to timing and content delivery.
It's not just Johnson & Johnson, AdmitHub just raised 2.95 million to help students get through college and Universal struck a deal with the ReplyYes chatbot to sell the correct music at the correct time.
You see, a chatbot delivered via messaging app has a huge advantage over mobile apps and websites. They have direct access to time.
I'm not going all Doctor Who on you, I mean, time, as in, knowing what a person is doing at a point in time, and having the ability to send a message at the exact time it is more likely to be read. I know, I know, the purists will say a mobile app can do that too, and it can. It's just that, compared to messaging apps, no one is downloading or using them anymore.
So, the final lesson, lesson 3. When you make your chatbot service the best way to deliver information, and you make it fun, you need to make sure it delivers the right message at the right time.
Nail these three lessons, and I'll soon be writing a blog post all about your chatbot service.
