Building a chatbot

How I increased Ecommerce revenue by building Chatbot

I was recently employed as a Business Development Manager for a Sustainable Building Supplies E-Commerce Company. Basically the business had up until then been run by one employee. The product range is actually quite broad, but 95% of the sales did come from 4 or 5 products with 2 or 3 variations each. The owners of the business literally showed me to my newly built desk area, gave me a 1 hour crash course on the current business structure and account details, then left me to observe the business for a week.

I thought this was quite a fair bit of trust in me and my value from the very start so I was actually more motivated than ever to make a positive impact on this business. One of my first BDM ideas was to build a chatbot.

Same phone call twice a day

After quietly observing the operations for a day or two, one of the first things I noticed was the business received quite a few calls and Emails with very similar enquiries. My first thought was the UI, or Website user experience needed improvement. But as I dug into this a little bit, I realized all of the information these people were after, was exactly where it needed to be. The people were just too lazy to find it. Having noticed these calls coming in during our 8am-4pm Office open hours, I started wondering how many enquiries we were missing from not being able to answer after hours.

This also reminded me of my previous job. I was working as a booking agent and sometimes we had a selection of 3 or more similar suppliers to book through. With experience, you begin to gravitate towards the suppliers that you know will give you the answers that you need, the moment you need them. As it was in my best interest to get the booking organised as soon as possible, there is no reason for supplier loyalty. If they didn’t answer their phone or have the availabilities online, sorry! I’m on to the next guy..

Building a Chatbot

So I made my first business optimization decision. It was time to build a chatbot! As the business is hosted on Shopify, there are a multitude of Chatbot options on the App store. But ultimately I chose Tidio because of the visual building process, the ability to also have customer service reps take over the conversation, the quality customer support and the supplementary mobile App you can also use.

Initially it took me around 2 or 3 hours to get the chatbot set up on the site and a basic conversation structure. To simplify the idea of a chatbot, I imagine it as an easier to navigate Frequently Asked Questions section (when there are more than 20 FAQ’s) The frequent phone call questions included: “Do you have an installer that can come and quote my job in…?” “Do you X timber available in X length?” and other similar shipping or custom product questions. I knew it would be overwhelming to try to build a super complex Chatbot that could answer all questions from the very start. So I set myself the goal of building it to answer the most common questions whilst also having answers that might encourage the customer to leave their email for any question the bot is not trained to answer. The easiest way to achieve this was to have clickable options for each question to have the customer click through the flow to get to their final destination/answer..

*Photo of chatbot flow*

Trial and Error

The first two weeks, the bot had 10-15 conversations that were pretty poor, but as the chat transcripts are all saved, I could use the real world data to start to go in and make minor improvements. After a month, the depth of conversations it was having were already very encouraging. One thing I refused to do from the start was have the “Customer must leave Email before they can chat” option selected. I think this would have dramatically reduced the Chatbot interactions as most people are unwilling to provide their Email when they are only after a quick answer. 

As I write this, I am still regularly checking over the chatbot transcripts, and I even have a system with the full time employee where he will go in and tag any conversations that he believes can be optimized. I then either add additional options or adjust the language or flow of the chatbot options. I have also been working on trying to capture the contacts details for a chance to increase our conversion rate.

*Another photo of Chatbot flow but more complex*

Website improvements

The other benefit of the Chatbot is the fact that it gives us much clearer suggestions on website improvements. As you can see the transcript in english, it becomes very clear where the pain points are on our website and product pages. So we have also been making small improvements to individual pages. I created a much easier way to have the customers calculate their own shipping prior to checkout and we condensed and filed the product information into tabs. 

This alone must have saved many hours of those similar phone calls and Emails, but the Chatbot is still busy getting more complex conversations, and I am totally fine with that!

The next step

Next I am looking into integrating the Chatbot with a new CRM software. As the chatbot conversations grow, some of the enquiries have been slipping through the cracks. I am looking to create a system where all different types of leads come together on one CRM. I have also started playing with the AI auto response settings, but this seems a tad difficult to master!

*Photo of Chatbot AI section*

The numbers don’t lie

As I initially suspected, there are a lot of chatbot interactions being taken outside of office hours. The website user data is obviously shown on Google Analytics, but I believe the Chat Analytics is a better gauge of when customers are actively trying to find the answers to help decide on a purchase as opposed to aimlessly scrolling through Instagram and then taking a quick look on our site from a linked instagram post.

*Chatbot conversation times photo*

Data to find list: 

  • Average Website Traffic
  • Total Chatbot Conversations
  • Average Chatbot conversations
  • Average Chatbot Conversations Per month/Cost = Cost per convo
  • Chatbots that have left their contact details/Total

Time is money

I would love to be able to put a figure on Chatbot Return on Investment, but I haven’t quite gone down the path of figuring out how to tag Chatbot Enquiries and correlate them with each sale. Something I will certainly look into in the future. 

Having crunched most of the numbers, I would still say that the time saved answering the phones and emails has been highly beneficial and it is for that reason, I highly recommend any similar businesses do likewise.

There is a bit of a learning curve when initially building the chatbot from a clean sheet, however once it begins to have conversations, small improvements become easy and quite fun actually!

If you are interested in building one for your site, but don’t have the time to learn and build the initial conversation flow, feel free to reach out and we will see what I can offer!


Automate Tedious Tasks is a Blog and Business Development Service for those looking to take more control of the most precious asset we have; time.

Latest Posts