Case Study

Increasing lead conversion rates and improving team efficiency through ML models deployment

How we supported our client in qualifying marketing leads with a higher probability to convert by using machine learning models for lead scoring. ​​

Reduced disqualification rates of MQLs by 14%

Improved SQL conversion rate by 5%

Hundreds of better utilised working hours

The Business Challenge

Our client had a huge database of marketing leads that was not used in the most efficient way. Sales teams felt that they spent too much time and effort on qualifying leads which turned out not to be the right ones. 

The internal teams wanted a more scientific way to qualify leads. Together we decided to score leads based on some relevant firmographic, demographic, and engagement data so that the leads that were more likely to become opportunities were given a higher score, indicating higher priority. The goal of this solution was to identify which are the leads with a higher probability of conversion, creating a focused nurturing strategy for the teams.

Increasing lead conversion rates and improving team efficiency through ML models deployment 1
Increasing lead conversion rates and improving team efficiency through ML models deployment 2

Technical Solution: Machine Learning Models deployed on Azure Cloud Environment for maximum scalability in real-time

The Solution

Amplify created and deployed machine learning model for building lead scoresMarketing leads with a probability score of more than 0.8 were shared with the sales team to follow up. The rate of 0.8 was decided together with the client, aligned with business requirements. In this specific case, the rate is high, as technical sales force with good expertise and experience is a scarce resource, so a company should make the best decision to use its sales force’s time and efforts in the most efficient way. 

The team faced challenges with accessing disparate data sources and was able to work with the client’s team to bring practical solutions and overcome problems to ensure the sales team could get value. In some cases, using REST APIs to pull data was required. Data cleaning and deriving additional features to cater business requirements and finally tuning the model in line with the capacity of the team were essential steps of the process. 

Our Approach

The methods used

Machine learning models were used to achieve the right lead scoring. 

Decay scores were created for some relevant engagement features: Email sent, Email open, Email click & Page visit. The model doesn’t only count the number of important engagement features but is assigning some scores to the time at which the actions were done. If somebody visits a page today, he/she would be given higher importance than somebody who visited the same page a month back. These decay scores are calculated using an exponential equation. 

Decay rates are different for the diverse engagement features: the decay rate for “Email open” is slower than the decay rate for “Email sent”. 

ML Model deployment

This machine learning model was deployed successfully on Azure Cloud with no bugs and is being used by the client’s sales team. 

Amplify pulled data from the client’s marketing applications such as SFDC and Pardot and used the Databricks platform, to train multiple models and MLflow, part of Databricks, to track experiments. 

The fresh scores are then pushed back to the Salesforce application in an automated fashion. 

Business impact

The solution resulted in an increase in conversion rates with the model of correctly assigning between 75% and 97% of the leads, depending on the country, based on relevant historical data of the marketing leads.

Related content

Want to know more?

Lead scoring

Lead scoring is a baseline methodology for resource and performance optimisation.