The Challenge
A national online learning platform was struggling with lead conversion efficiency. Their sales team was spending time on leads that were unlikely to convert, while high-intent prospects were falling through the cracks. The challenges:
- No data-driven prioritisation: Leads were distributed to the sales team based on simple rules (first-in, first-out) rather than likelihood to convert
- Manual processes: Business processes that could be automated were consuming engineering and operations time
- Fragmented data: Customer data was spread across the learning platform (Laravel/Lumen), CRM (Salesforce), and various analytics tools — with no unified view
- Stack complexity: The team needed a technical lead who could work across PHP (Laravel), Python, JavaScript, and Salesforce simultaneously
Our Approach
Data Unification
Before building models, we unified the fragmented data landscape. We created pipelines that brought together:
- User behaviour data from the learning platform (page views, course interactions, completion rates)
- CRM data from Salesforce (contact history, deal stages, sales interactions)
- Marketing engagement data (email opens, click-throughs, campaign attribution)
ML Pipeline for Lead Scoring
We designed and built a machine learning pipeline that:
- Feature engineering: Extracted meaningful signals from raw behaviour data — recency, frequency, and depth of engagement patterns that correlated with conversion
- Model training: Trained models using historical conversion data, with careful attention to class imbalance (most leads don’t convert) and temporal validation (training on past data, validating on future data)
- Scoring and delivery: Scored incoming leads and pushed predictions back into Salesforce via API, so the sales team could see lead quality directly in their existing workflow
Business Process Automation
Beyond lead scoring, we automated several manual business processes using Python, reducing operational overhead and freeing the team to focus on higher-value work.
Technical Leadership
As Technical Lead, we oversaw front-end and back-end developers in agile methodologies, implemented architecture decisions for new features, and managed the technology stack including framework upgrades and quality standards.
The Outcome
The ML pipeline automated lead prioritisation, replacing manual qualification with data-driven scoring integrated directly into Salesforce. The sales team could focus their time on the leads most likely to convert. Business process automation reduced manual operational work across the platform.
Technologies Used
- ML/Data: Python, scikit-learn, deep learning
- Backend: Laravel/Lumen (PHP)
- CRM: Salesforce (API integration)
- Frontend: JavaScript
- Infrastructure: Automated pipeline deployment