Location data engine Foursquare recently updated its Geosummarizer model to improve the location accuracy of its point of interest data. Foursquare says this will substantially improve customer experience where its data is involved. And it will improve the accuracy of its POI data by as mcuh as 20%, the company said.

The Geosummarizer model helps the data company achieve greater location accuracy of its point of interest data. So far, Foursquare has released the update in six countries. These are Canada, Switzerland, Spain, The United Kingdom, Japan, and the Netherlands.

Foursquare began in 2009 as a popular, early smartphone app on which users checked in to popular local spots. This foundation of location-data collection helped Foursquare to evolve into a location-data engine that supports major brands. Some of the brands that use Foursquare include Twitter, Snapchat, Uber, Spotify, Airbnb, Coca-Cola, JetBlue, and others.

We caught up this week with Zisis Petrou. He is Foursquare’s Senior Manager of Engineering. We asked him for some color on the update and what impact it will have on the location data ecosystem.

Here is our brief interview.

Localogy Insider: So this appears to be a product update. Is there any change in Foursquare’s commercial offering?

The commercial offering itself remains the same. But we are now able to offer more accurate lat-longs in the six countries where we have released this so far.

We found some erroneous geocodes in our regular quality checks. And we worked to systematically improve across every country, taking into account the unique geographies and city planning of various countries into our data science models.

• Before the update, there were some geocodes that were more than 50 meters away from the location. We estimate a reduction of distance to the centroid by 20%.
• Also, with this update, we expect the more than 2 million POIs that were outside the perimeter of the venue to be positively impacted and now within the perimeter.

Localogy Insider: Can we get a better sense of the scale of Foursquare’s data operation?

The new Geosummarizer model is part of Foursquare’s continuous efforts to improve the geocodes of our POIs. It is a large and important initiative driven by the Places team in coordination with the Data Platform team. It involves the work of software engineers, data scientists, and product experts, as well as a huge data collection effort undertaken by a manual annotation team.

Localogy Insider: Can you offer an illustration of how consumer experiences will be improved by this update?

Let me give you a real-life example. Say you’re headed to lunch at a restaurant on a crowded street in London. If you use one mapping app to get there, you could end up on the wrong side of the street at a gated parking garage. But if you type the very same address into a mapping app built on our Geosummarizer model, you’ll end up at the restaurant entrance. The difference between these two experiences depends on the accuracy of the geocode.

Our enhanced Geocode Summarizer improves the accuracy of points of interest by 20%. Ultimately this helps people get where they’re actually going.

Another illustration is navigating to a POI within a mall or any other large complex with multiple POIs sharing the same street address. In the past, a drop-off point would be a generic entrance or exit near the centroid of the venue. With the improved Geosummarizer model, a visitor can locate where the POI lies inside and then choose the most convenient entrance. This expedites the overall navigation experience.