TopN AnalyticsDemand-Driven Site Selection for New Charging Infrastructure
An eMSP platform / charge point operator seeking data-driven guidance on where to expand charging infrastructure.
Where things stood
Deciding where to build a new charging station is a high-stakes, capital-intensive decision that's often made on incomplete information — intuition, site availability, or general area knowledge — rather than direct evidence of where EV drivers actually need to charge but currently can't easily.
Existing infrastructure planning typically only looks at where stations already are, not where real vehicle activity is happening that isn't being served.
What we built
Vehicle telematics (PostgreSQL, ~20M rows/30 days) flows through a chunked, retry-hardened Python/Airflow extraction pipeline into BigQuery staging.
Distance-to-nearest-station is calculated per activity point, and activity more than 5km from any station is geospatially clustered (Vincenty distance, 100m radius) into candidate 'hotspot' locations.
Each cluster is scored with a composite metric weighing normalized vehicle density (60%) and charge urgency (40%), producing ranked candidate-site dashboard tables.
How it works
TopN Analytics built a pipeline that identifies 'charging deserts' directly from vehicle behavior — filtering to moving vehicles with a low state of charge, then clustering underserved activity into a ranked shortlist of candidate new-station locations.
The scoring methodology is a reusable framework applicable to any charging network with telematics data.
What we delivered and learned
This moves site-selection decision-making from 'where do we think demand is' to 'where does the data show unserved demand actually concentrating' — a materially more defensible basis for a capital-intensive infrastructure decision.
What it's built on
Want the full picture?
We're happy to walk through the details, numbers, and trade-offs directly.