Do you know how the raw_visitor_counts or raw_visit_counts are obtained?

Dear all, do you know how the raw_visitor_counts or raw_visit_counts are obtained? For example, some small parks are free to public and do not have guards to record your arrival. So for POIs of parks with raw_visitor_counts or raw_visit_counts in SafeGraph, do these POIs only cover parks which need tickets or memberships?

My understanding is that POIs have defined shapes, and mobile devices that enter that shape can be detected.

Hi, Dennis. If so, when the visitors do not open Bluetooth/GPS/WIFI, it seems that they will not be recorded.

They track cellphone mobility so if the phone is untrackable then I believe they are not in the universe of the dataset.

Hi, Kok. How do you know that the SafeGraph track cellphone mobility? Are there any websites about it?

Hi Zhou, funny how I did a quick scan through of the documentations and found no direct mention about the nature of the data. But anyway, here’s some words regarding mobile device tracking: “SafeGraph Patterns measures foot-traffic patterns to 3.6MM commercial points-of-interest from over 45 MM mobile devices in the United States …” in What about bias in the SafeGraph dataset?

Also if you look into the documentation, e.g. this one, you’ll see the columns like device_type which is either IOS or Android, and carrier_name which hints at mobile devices being tracked here.


Hhh, it would be better if the staff in SafeGraph can officially explain it.

Hi @Zhou_University_of_Toronto, @Dennis_Chao_Institute_for_Disease_Modeling, @Kok_Ben_Toh_University_of_Florida

Yes, the foot-traffic data is primarily derived from a panel of anonymized and aggregated mobile devices (smart phones).

We discuss some details of the data here:

These are the aggregated raw counts that we see visit the POI from our panel of mobile devices.

The blog post linked on Bias is also a good source of info about the panel and dataset.

Finally we also have this Visit Attribution Whitepaper that is free to download that discusses in detail how we match GPS mobile data with POI data to produce the visits reported in Patterns. It is not simply a point-in-polygon geospatial join.

Hope these details help

Thank you, Ryan. Thanks, Kok and Denise. Thanks for your sharing info!