Make Contribution to the Customer Happiness Team Productivity

Poor customer experience is the biggest reason why customers say they will not continue to use our product. Customer satisfaction (CSAT) is a measurement of how well a company’s products, services, and overall customer experience meet customers’ expectations.

So, to maintain and improve CSAT score at Qasir, we need to focus on providing consistent customer support experience. Continue to improve customer experience to win over them and give them the excellent customer satisfaction they expect from Qasir.

What is Customer Happiness

Did you know as a customer of digital products, where we should be asking if there is any trouble, or something we would like to ask regarding the details of the product? we immediately click on live chat support hanging on the screen, right.

In Qasir we have a team called Customer Happiness (CH), a team who is responsible for handling live chats in our digital product, meaning we rely on humans to reply to those live chats as written in this article https://medium.com/qasir/keluhan-kamu-di-tangani-secara-langsung-tanpa-chat-bot-1a531990cfd0

That’s why to keep our customers happy, hence the team name, we believe the key is to have a great “Response Time” to our customers. It means less response time and faster replies / feedbacks the customer get, the happier they will be.

Customer Happiness and Productivity

Below is a simple flow of our CH team when live chat conversations are going on

Receive Chat from Customer → Customer gives phone number/email/id → Get details / Take some actions → Reply back to Customer.

Within the flow above, the most crucial part of the process is the third part, which is when CH needs to gather customer details or take some quick actions and immediately replies back to the customer.

Below is the average response time in daily basis when CH needs to finish the third part process (unit is in minutes for any actions taken)

This means, if we have 200 merchants complaining, meaning there are 200 open conversations, it will consume more than 20 hours, with an average of 8.75 minutes in 1 action on the conversation, for CH to replies back to all conversations.

After the discussion with the CH team, we found 2 main reasons :

• No clear source of truth, it’s scattered in many places. It’s so hassle for CH team where they must open many internal websites within many browser tabs to get the data.
• Some actions rely on engineering in a person, while some depend on the availability of people, especially on the weekend.

Introducing Engineering Support Bot

We still figured out the right name for the bot, but meanwhile, let’s just call it an engineering support bot :)

Why chatbot? Moreover, CH commonly used live chat in daily basis and they frequently open the chat app to interact with engineering support team and get the problem solved. Chat app is becoming their primary weapon for handling complains and solving the issues. Oh, btw in Qasir we’re using Twist as main communication tool. Yups, the Engineering Support bot is a bot member in Twist.

By using this bot, CH and/or other internal stakeholders can simply just type the command to this bot and the bot will reply with the information they need.

Does this bot rely on NLP, AI, etc? Nope, it doesn’t. We just built this simple bot using the open-source library written in Golang :

• https://joe-bot.net/
• https://github.com/gofiber/fiber
• https://github.com/kukuhpro/joe-twist-adapter for the twist adapter

As all complex things are handled by Joe Bot, it only took us, more or less, 1 day to get this bot to production including the testing!

The Impact

As seen above, the action for each conversation has been decreased significantly, from an average of 8.75 minutes to 1.3 minutes for each action to complete.

Conclusion and Future Work

Without us knowing, this tool brings more impactful result for the CH team. At first, we thought the chatbot should be complex and use advanced technology, but we realized that our objective was to reduce the average time of the process, meaning we should focus more on delivering the value as fast as we can so we can iterate faster, that’s when we decided not to use sophisticated technology and rely on open source project instead to come up with the solution.

For future work, we will make the bot to understand more command to help CH solving more problems, and we’re planning to use this tool not only to help the CH team but also for other internal teams. And maybe if necessary, we can add more advanced technology to our chatbot system so it can understand natural language, like many chatbots do nowadays.