Geocode Correction Launch
Shipt ยท Target Last Mile
- Geospatial analysis
- PoD validation
- SQL
- Launch analytics
- Target Last Mile
Correcting bad delivery pins
Target Last Mile depended on delivery addresses and geocodes passed from upstream systems. When the geocode was wrong, drivers could be sent to the wrong dropoff point even when the written address looked right.
An internal geocode correction system corrected those pins. I owned the validation: proving the corrected geocodes were trustworthy enough to launch and sizing the operational impact.
What I validated
I compared corrected geocodes against proof-of-delivery photo locations and the original source pins. The analysis focused on corrections over 100 meters, where bad coordinates could meaningfully change the driver path or the delivery geofence.
- Analyzed a multi-week sample of Target Last Mile deliveries, focusing on the subset with corrections over 100 meters.
- Showed the corrected geocodes lined up with proof-of-delivery locations far better than the original source pins.
- Sized the residual bucket of likely wrong-dropoff orders the launch still needed to handle after correction.
Why it mattered
The launch needed trust. This analysis gave product, data science, and engineering a concrete read on accuracy, return exposure, and delivery risk. It turned a geospatial data quality problem into a measurable launch case for correcting delivery pins in Target Last Mile.