Saving ~2800 hours for Sales

Mar 11, 2024

Author Slug

Problem Statement

Simplilearn website gets around 35000+ chats every month from users across the world. Around 40% of the chats gets converted into leads. The chat dropping consumes a lot of human hours from our sales team.

We need to make the chat more robust which can filter better leads so that the business goals are met. We also need to automate a few things in the chat process so that the consumer goals are met too.


Target Audience


Design Process


Primary Research

Primary research includes the insights based on from a months chat data and also the requirements and insights from the product and sales team.

We had the chat conversation data for the month of September. There were around 36000+ different chat transcripts that were presented before us and we extracted the insights from there.



Product Data


Customer Chat Category








We read through the chats and did a health check for around 10% of chats (3716) marking the conversation quality as High and Low.
The criteria were,

  • The chat taken for insight generation were spread over the month.

  • High Quality conversation meant the users were willing to share their information to get the course details and other inputs.

  • Low quality conversation means the users dropped in between the chat.

    Here are the insights from it.



Key observations from the chat transcripts study

Customer needs

Out of these High Quality Conversations, we read through the chats and discovered the customer/learner needs. The tone of the conversation and the way Learning consultant are driving the conversation.

Learners with high quality conversation are asking for the following things to the Learner Consultant

Business Goals

Along with the Customer Goals we also have identified few business goals that will improve the overall experience of the chat and also aid the business and sales team to make the most out of the chatbot.

  • Generate better quality leads

  • Get CSAT after chat

  • A good design better customer experience

  • Media Support (Send links, pdfs, images)



Impact Assesment

Previous Chat Conversion

Key observation

  • 2.8% of total chat initiation leads to a paying user.

  • 56.79% of total chat initiations leads to non leads, this consumes a lot of man hours.

  • 21,459 non-leads led to a consumption of more than ~2800 man hours



Potential Impact on Chat Conversion

We aim to filter out all the non leads, there by saving a lot of man hours. But, this will be done in phases as we release the chatbot and get valuable insights on it


Phase 1

  • Feature rollout in some parts of India with monitoring the user.

  • Surveying users who have interacted with the chatbot to gain valuable insights,

Phase 2

  • Modifying according to the phase 1 results if necessary.

  • Rolling the features to whole country and monitoring them.

  • Adding multimedia support to the chats



Secondary Research

We did a market analysis for the competition analysis, academic publication and online resources.


Competitor landscape

Out of around 13 competitors in the ed-tech sector, we discovered 4 companies that already have a chatbot in their website. Here are their insights


Good to have features

We have identified some good to have features in the chatbot after analyzing the secondary research.


Flow and Framework

The analysis and research gave us a basic idea about the requirement for the chatbot. We need to build a framework for the chatbot and then create the flows which suits our customer and business goals.
For the flow, we first analysed our competition and plotted their chatbot flow, taking in consideration all the aspect of their platform. The we designed our flow framework to get going with the chat flow.

We did flows for 4 competitors and it gave us a idea of how to drive the conversation and get user details.
This helped us to design our framework for the chatbot and what data needs to be presented to the users before taking them to the human counterpart


Challenges and Learning

  1. Working with product teams while exploring the strategies and features really beneficial learning experience.

  2. Conversation design for chatbots is very interesting thing to learn as a good flow can create so much impact

  3. Creating conversation flows and then missing out on one thing and then re engineering the flows was a great learning experience.

  4. Reading through chats and generating insights was the real challenge as I had to figure out how to create a system that can be applied to 3700+ chats. I tried to read chats with python codes, tried google cloud NLP, tried openAI to do NLP but the raw chats were not easy to be read and made sense by the AIs so I had to filter out only 10% of them and do a manual job out of it. It was fun to play with so many tools to find out what works.

Overall, I had a lot of fun while doing this project and I hope that the rollout will create the desired impact for the users and the business.

Description