🤖Brand Matching

We have recently made updates with brand matching and the integration of machine learning. These enhancements significantly improve the precision and efficiency of our data offerings, ensuring we deliver higher-quality results to our customers.

Key Improvements and Performance Gains

• ML-Driven Matching: We've adopted an ML-based approach, allowing for more precise and flexible brand matching. This new system learns and iterates over time, providing better accuracy.

• Brand Matching Confidence Score: We are now calculating a Brand Matching Confidence Score to provide insight into the confidence level of each branded POI, helping teams better assess data reliability.

• Performances: We’ve seen a 27% improvement in precision. This leap is a game-changer for the accuracy of our data.

• Reduced False Positives: We've significantly reduced the number of false positives (incorrectly branded POIs), a major issue for geolocation data providers, further improving overall data quality.

• Configurable Thresholds: We’ve introduced configurable thresholds that allow us to fine-tune data quality controls, ensuring that we maintain the highest standards across different datasets.

Impact Across Industries

• Industry-Specific Lists: Precision has improved across all countries and all industries.

• Overall Impact on Branded POIs: While we have a slightly lower total number of branded POIs than before (since we reduced the number of false positives), the precision is much higher, leading to better data quality.

Why This is a Positive Change

• This improvement means we have greater confidence in the accuracy of the brands we deliver, offering our customers higher-quality data.

Validation Efforts

• To ensure quality, we’ve manually checked and annotated thousands of POIs, confirming the reliability of our results.