Company
🏦 Santander
Skill Area
🌐 Organic and Paid Traffic📊 Data & Analytics
Tool Stack
PythonLooker StudioGoogle Tag ManagerGoogle Analytics
Overview
🔍 Context | Need for a solution to manage Google Ads campaigns for Santander Consumer Finance's financing products across NL and BE that could rapidly adapt to changing business priorities between volume and value optimization (e.g. AOV / number of leads).
Traditional manual bid management required weeks to stabilize performance after strategic shifts, while the complex conversion funnel from click to loan approval required advanced attribution and real-time bidding adjustments to optimize for true business outcomes rather than surface metrics. |
👨💻 Key Role | • Defined adaptive business rules to dynamically shift between AOV and volume priorities
• Identified relevant key bidding variables (keywords, time patterns, audience segments) to use
• Created comprehensive performance analytics framework for RORWA measurement |
Scope & Responsibilities
- Defined Requirements: Wrote detailed business and technical requirements for a Python-based ML algorithm that could adapt to our frequently changing business priorities
- In-house and agency coordination: Led collaborative sessions with agency partners and in-house business team to define algorithm parameters and success metrics
- Data integration: Guided the technical implementation of Google Ads API integration to connect offline conversions from KYC and signing portal that served as key conversions for the algorithm
- QA and testing: Performed testing and optimization phases, ensuring the algorithm performed against established business objectives
- Strategy: Maintained strategic oversight over the life of the model while assisting the agency partner to execute on the technical development
Technical Solution
The custom bidding algorithm's technical framework I specified included:
- Rule-based decision matrix: Algorithm incorporated business rule prioritization that could be adjusted as frequently as needed with minimal code changes
- Multi-dimensional bid modifiers: The algorithm could perform simultaneous adjustments for:
- Time-of-day/day-of-week patterns based on conversion probability
- Audience segment value scoring based on historical loan completion rates
- Keyword intent classification (research vs. purchase-ready)
- Geographic performance variations between Netherlands and Belgium markets
- Data integration pipeline: Established secure API connections between Google Ads, CRM system, and the KYC verification portal
- Python-based machine learning core: use of scikit-learn libraries to analyze historical performance data and predict optimal bid adjustments
- Performance feedback loop: Created system to automatically compare predicted vs. actual performance, enabling continuous algorithm refinement
- Custom reporting dashboard: Created a looker studio dashboard showing bid adjustment impact on key business metrics
Performance highlights
- RORWA: 45% higher RORWA (financial services equivalent metric of ROI) from paid search channel since implementation
- Lead quality: 20% increase in application-to-approval conversion rate
- Lower agency manual hours cost: Saved around €3000 reduction in manual bid management hours monthly
- Objective adaptability: Successfully shifted between volume and AOV priorities, improving ‘time to performance stability’ from few weeks to few days