The story behind the success of our Chatbot : do's and don'ts

At Ideta, we had a website for quite a long time. Yet, we did not use any livechat. For those who do not know, a livechat is the small interface in which you can request assistance from a human being. A human agent will then answer your queries live. The goal is to provide the answer to your customers immediately, so that they can stay in the mood for buying. You never know if or when they will come back. But to be true, for smaller companies, it has become just another lead generation tool. There are not enough resources to answer in real-time. So, the livechat tool simply collects the customer email and it is simply back to basics.

We were not ready

‍

We were convinced that the tool was useful. Lots of studies confirmed it. It just took us some time to prioritize the issue.

Customers kept asking "Why don't you have one on your page for demo purpose?", "You should put one on your website!" etc.

We were not ready. I know the famous Reid Hoffman quote "If you're not embarrassed by the first version of your product, you've launched too late" but this was simply too much embarrassment. At the beginning, we were good at making chatbots on Facebook Messenger, Microsoft Teams, Slack or whatever channel as long as we were not responsible for the front end, that is the user facing part. At some point though, we decided to tackle the issue and decided to build our own livechat and power it with one of our chatbots. After some preparation, the tool was finally ready to be used.

There are two main differences between a livechat and a chatbot powered by Ideta :

1- Automation

The customer on your website is not waiting for you to receive the notification email, break abruptly your coffee chitchat before you jump behind your computer, unlock it, open the chat management app and start answering. They simply navigate away. Our chatbot starts collecting data and answering questions, so that you can have the time to get to your computer and intervene if necessary.

2- Availability

The chatbot is available 24/7 which is quite complex and expensive to achieve otherwise. Indeed, you would need to organize a chat center, with people answering to your customers in real time, at any time. Most of the time, call centers are not open 24/7. It becomes way too expensive.

We finally did it!

Actually, fun fact, we were quite concerned about missing an opportunity on the livechat. We thought about the way of giving the information in the most efficient way. Our chatbots can connect to any API. We organized sending a message to a Slack channel. We would get a browser or an app notification every time someone would connect to the chatbot. The issue there is that I dislike notifications. I, for example, simply cannot focus with the constant messages and noise. We decided to find a better solution. And it so happened that we had a Philips Hue lamp that we had bought for demo purposes. It was still in its original packaging when we found it in the unpacked boxes of our latest moving. After some documentation reading and a bit of tweaking on our platform, it was finally ready. The lamp would turn green if a customer required a human agent. Hence Assane above !

In the meantime, we started developing the actual experience in the chatbot. What to say? Why? In what order? On this account, we were quite aligned with Reid Hoffmann and were eager to iterate very fast. We did not know what would work best. We decided to just publish something. With our software, it is easy to iterate. No coding required anymore. Our Customer Success Manager can handle the job from now on.

The first version was not perfect. The natural language processing was undertrained. It would not understand properly customer queries. The conversation flow was inefficient and brought lots of interaction but little leads. After every iteration, the conversion rate improved and we reached the current status with a 28% conversation rate. In addition to that, we realized that our chatbot was a really efficient click bait. It was the most clicked button on our website.

What we learned

From this experience, I prepared a short list of do’s and don’ts to help you in the process :

- Don’t start with NLP (Natural Language Processing) : it is better to train a bot on actual customers requests
- Don’t wait: launch fast in order to improve on user feedbacks

- Focus on your target: lead acquisition or brand discovery ?
- Ask for advice: we are here to help

In the end, our chatbot actually transformed into business opportunities and revenue. We managed to bring an automatized salesman on our landing page.

If you want to multiply your leads and radically improve the navigation experience on your website, come and drop our bot a line at https://ideta.io/

You can also register for our webinar on the subject.