Automating the GsmServer customer support
Take care of low-value queries with a high-value chatbot (the Basky bot)
GsmServer is a multi-range international online store, where you can conveniently purchase telecommunication related technological solutions, STEM toys, accessories and spare parts for cell phones, various electronic components and equipment, repair/service tools etc.
What was the problem
The customer and technical support were overloaded with repetitive queries. Some user flows on the website, like the account's password recovery, were too complicated and stopped customers from quick resolutions. Besides, the team couldn't provide a personalized customer experience because it took lots of time when manually identify the customer's email, number, and previous purchase history. On top of that, the team worked only during the weekday and till afternoon on Saturdays; on the Saturday evening and Sunday there was no one to help, and, it turned out, that it did cause problems for their customers.
We understood that we need a very specific chatbot, many will have to dive into various details. That's why we needed the chatbot company to be ready to delve into our business, to listen to what we do. And not to offer a solution they make for others – because we are not a typical e-commerce store.
Here's how we helped
A website chatbot that answers general and more specific questions about products, provides information about orders, and helps customers buy items. Also, it easily connects them with a real human if you don’t get your answer.
Platform: Website Widget.
Language: English and Spanish
– The chatbot is integrated with GsmServer CRM, which will allow matching users with their order details and provide users with the order data;
– The BotsCrew platform allows GsmServer to track, analyze conversations and session flows, create cohorts and funnels to see how customers interact with the solution;
– The chatbot user can trigger customer support or a specific agent. Meanwhile the GsmServer agent can see conversations between the user and the chatbot and interwin in a conversation at any point;
– NLP training (150 FAQ about shipment, products, payment, software downloading, etc + intents which require integration, like: “Where is my order?,” “Talk to human,” “Buy an item”)
We launched the chatbot in March 2021. We did not want to scare buyers, thus the chatbot widget is in the lower right corner, while on top, there is a button “Chat with us” to talk with an agent. Despite knowing where the chat with an agent is, regular customers click on the chatbot as well. Of course, new customers choose a chatbot because it is more visible, and people have got used to chatbots. But regular customers choose a chatbot consciously.
The results were both promising and unexpected. After six months, the chatbot automated 72.28% of all customer requests. That means the chatbot covers near 50% of all chats on the GsmServer website from June to December 2021. Now you see how this number had grown when we added the Spanish version and updated the Buy and item flow.
How did we achieve it
Discovery and planning
We identified the most popular requests based on existing conversational history analysis and discussions with the client.
The client’s feedback on the Discovery phase results:
It was very valuable for us. And an interesting place for me. I have been working for the company for 11 years, and we never-never-never knew that statistically, the number one query was "Check my order status.” It was strange for us, because our customers can log into their account on our website and see the status of orders there. It seemed to us that this need was met. But they have behaved differently than we thought.
Knowing that they already had a chatbot, we understood that starting with POC wasn’t the smartest decision. We proposed to jump to more complicated project stages and scale what was done before. That’s why the GsmServer chatbot had only two phases: MVP and post MVP. So in 6 weeks, GsmServer would have a viable solution that would automate 50% of customer requests.
The development process was simple and smooth. There were just a few integrations: their website to magnify user experience, Chatbase (upd. Now our Platform analytics) to track chatbot performance, and Livezilla to connect customers and live agents. Based on the most frequent cases and questions from previous chat history, our conversational designer prepared a diagram architecture of how users will receive the information using buttons and typing their questions. When we finalized all the content and transferred it to our Platform, we did a couple of workshops with the client to show how they can update content on their own.
Currently, we are in the post MVP stage, so we are making some minor changes to the widget that will facilitate the website experience when chatting with a chatbot. And we are working on improving the Buy an item flow and analytics of chatbot’s performance.
The chatbot's name is Basky Bot. We formed it from a simple word and item – Basket. Chat with Basky Bot here.
The client’s feedback on the chatbot results:
For us, the chatbot’s value was obvious: instead of a person sitting and copying the order status – they can call customers who have not bought from us in the last year or two, ask why, find out the reasons or organize any other marketing or sales activities. We still have enough work to do, so becoming more efficient is the number one priority.
Here's how the chatbot works
You can download the longer version of the case study below. In the document we described in detail each step of our collaboration with GsmServer: from the problem description to chatbot development.