Learn about Points-of-Interest data & its applications, where to download free POI data, how to evaluate Points-of-Interest databases, and the top alternatives to the Google Places API & Factual Places API.
The definition of Point-of-Interest is a specific physical location which someone may find interesting. Restaurants, retail stores, and grocery stores are all examples of Points-of-Interest. Since the phrase is a mouthful, Point-of-Interest is often abbreviated as ‘POI’. Many companies that sell POI databases or POI data API’s name their products ‘Places Data APIs’.
Retailers & retail analytics firms use POI databases to get lists of where stores are located. Along with store location info, they use metadata about a POI such as the category (NAICS code) and foot-traffic (visitor counts) to power site-selection & trade area analysis.
Ad-tech companies create geofences around Points-of-Interest to create location-based audiences. They also join the GPS data with POI geofences for advertising measurement (online-to-offline attribution).
POI data is also used in the real estate, consulting, and financial services industries.
POI Factory is one source of free POI data, but coverage and data freshness are lacking. POIplaza is another source of free POI data but it suffers from similar problems.
In partnership with ESRI, SafeGraph has free Points-of-Interest data available on the ArcGIS Marketplace. However, this data isn’t exportable and is meant only for visualizing POI data on a map for ArcGIS Online users.
While not free, the lowest cost a-la-carte option for buying POI data is 10-cents per place from SafeGraph’s Data Portal. For bulk access to POI data, there are several paid enterprise offerings (each with pro's and con's) covered below.
The Google Places API generally has the best accuracy and one of the most comprehensive datasets of POI in the market. It is high quality and they have a fantastic team that keeps the data up-to-date.
However, one of the biggest dis-advantages with Google Places API is the licensing terms. It’s extremely hard to legally use their data unless you are just displaying it directly to users on a Google Map. For example, Apple or Uber find it tough to license Google data in a way that works well for them.
The other worry with Google Places is that they might change their (already restrictive) terms or increase their already expensive pricing. Google did that in 2018 and it hurt various geospatial and mapping companies.
The Google Places API is one of the best-known Places APIs but enterprises often choose alternatives such as Foursquare Places, Factual Places, Facebook Places, & SafeGraph Places due to factors such as price, licensing terms, & data availability.
At SafeGraph, we’ve seen companies switch from Google Places to SafeGraph Places due to SafeGraph’s broad & permissive licensing terms. Short of directly re-selling our data, almost anything goes since SafeGraph is just a data company. We encourage companies to use our data to create derivative products and applications. Our Places data is also much cheaper for enterprises looking to buy data in bulk.
Lastly, SafeGraph Places has attributes that the Google Places API does not offer. For example, when it comes to building footprint data (geofences) for a POI, Google has that data internally but it isn’t offered via API. Thus, many companies choose to augment their Google Places data with SafeGraph’s Places Polygons or to use SafeGraph Places exclusively.
Factual Places is another well-known Places API provider. They have global data on places, supporting 130 million POI in 52 countries. They also have many granular POI category specific attributes. For example, for restaurant POI, they have data on whether the restaurant accepts reservations, is cash only, is kid friendly, or has a kids menu.
For several reasons, businesses choose Google Places, Facebook Places, Foursquare Places & SafeGraph Places as alternatives to Factual Places.
Businesses chose Foursquare Places vs. Factual Places because Foursquare has POI in every single country on earth while Factual only has data on 52 countries. Due to the user-generated nature of Foursquare data, Foursquare places coverage is also higher and more accurate for POI that people tend to check into often, like restaurants & bars. But for this same reason, at less checked-in-to POI, like a doctors office or grocery store, Foursquare data can be lacking compared to Factual’s data.
Businesses also often chose as an alternative to Factual Places due to SafeGraph’s broad licensing rights, price, and high accuracy & coverage for places in the U.S. Unique to SafeGraph Places, almost every POI has data on the exact building footprint (polygon) for a place. Factual (and all other popular POI alternatives) offer only a location’s centroid.
For geofencing applications, where knowing the precise location of a place is crucial (such as for turning GPS data into store visit intelligence), SafeGraph Places polygons make SafeGraph to other Places data vendors.
Evaluating a POI database can be a tough task, so SafeGraph put together a POI data evaluation guide. Some critical questions to ask when evaluating a places database:
- What countries are covered? Is it U.S. only or international?
- Precision (how accurate is the data?) Are the places open businesses and not real consumer POI like home-business LLC’s or stores that went out of business).
- Recall (what percentage of real-life locations are actually in the given dataset)? You can measure this for both national brands (like McDonalds) as well as for long-tail POI like mom & pop stores.
- Completeness (what is the fill-rate for attributes?) POI data providers boast about all the meta-attributes for a POI, like open-hours or phone-number or category (NAICS code) but the fill-rate might be lacking. How complete is each listing on average?
- What is the granularity of a POI location? Is it just POI addresses? POI centroids? What about the exact building footprints (geofences?) for a POI?
The physical world is in constant flux. Maintaining accurate data on places is an immensely difficult engineering problem. At SafeGraph, even with $20 million in VC funding and a top machine-learning & data science team, we still find bugs and inaccuracies in our own data all the time (and openly publish our problems for transparency).
You can read more about the complexities of physical data in this blog post:
Forget ML: 4 Weird Edge Cases Which Confuse Even Humans When It Comes To Places Data
Points-of-Interest are categorized by . NAICS stands for the North American Industry Classification System. Some example NAICS categories for POI include “Restaurants & Other Eating Places” and “Grocery Stores”. Some examples of NAICS POI sub-categories are “Limited-Service Restaurants” and “Convenience Stores”. You can view SafeGraph’s full list of POI categories to understand our POI data coverage by different NAICS code.
Classifying a POI to its correct NAICS code is an incredibly challenging problem that we’ve worked heavily on at SafeGraph. We tackled this problem by using Natural-Language-Processing and human feedback to analyze the content on a POI’s website and use that to map the POI to it’s most probably category. We have a 89% fill rate for this attribute and our proud of our accuracy.
The biggest problem with POI data is that physical places info is fragmented around the web. Company store locators, governments, commercial real estate companies, & user-review sites all have some POI data but with varying attributes, freshness, and accuracy.
The tricky part is combining all the data into a unified schema and also verifying the accuracy of the different underlying datasets. You can read more about how SafeGraph creates it’s POI data.