The Fifth Elephant 2020 edition

On data governance, engineering for data privacy and data science

Predicting Deal Closure in a Sales CRM using Email Sentiment

Submitted by Vishal Gupta (@vizgupta) on May 31, 2020

Status: Submitted


Emails are the most common form of communication in a sale and can be used to actively determine the customer’s interest in purchasing a product/service. Statistically, deals with more email replies from the customer are more likely to win. Our project, Deal sentiment at Freshworks as a part of the Freshsales CRM involves predicting sentiment from customers’ and agents’ mails and using it to estimate the probability of the deal winning.


Cleaning and Parsing emails (Data cleaning)

  • Parsing HTML mails
  • Removing Signature from emails
  • Processing Zoom invites
  • Processing calendar invites
  • Converting emails to conversations

Annotating emails/conversations (Data annotation)

  • Sentiment tagging (-2 to +2) : Need for tagging conversations and why deal outcome cannot be used.
  • Intent tagging : One or more intents tagged from a pool of 50 intents picked by consulting salespersons


  • Pipeline
    • Ingestion : Kafka consumer, followed by preprocessing and language prediction
    • Population and generation of conversations
    • Sentiment prediction : Get embedding and predict conversation sentiment
    • Use conversation sentiment scores and other features to predict deal sentiment
  • Multi-account models : Using clustering to pick models

Deploying Sentiment (Productionizing data science)

  • Scaling to multiple accounts
  • Clustering customer mails to create buckets

Speaker bio

Machine Learning Engineer, Freshworks



Login to leave a comment