In the course of some analysis I found a handful of POI with abnormally high visit counts or square footage that may be anomalous

Hi all! In the course of some analysis I found a handful of POI with abnormally high visit counts or square footage that may be anomalous, since our organization does not have access to the Geometry files at the moment I was hoping to get a second opinion

sg:a1cd41b5364e460ba6242827aa4437ff, Cafe, 9535 Forest Ln - apartment complex?

sg:f8b3ed929b4e4a459459a37c0e998f0c, Pizza, 3601 Pinnacle Point Dr - apartment complex?

sg:05fcb1814d5a48f692c5bd6d243b6dce, Galindo Tacos, 1700 W Irving Blvd - 4,000+ raw visits to taco truck by muffler shop, incorrect polygon?

sg:2b41a1e40d1f49d4a84607300d1f41c4, Scout & Cellar, 4531 Simonton Rd - 180k square feet (August release), incorrect polygon or includes parking lot?

sg:96249baec7b241bb8d1f37025792d525, Sidecar Social, 5100 Belt Line Rd Ste 401 - 23k square feet, incorrect polygon class?

There are a handful of other locations with abnormally high square footage (mostly restaurants/bars with over 50k sqft), does that imply inaccurate polygons or incorrect polygon class or parking lot flag?

Hi @Alex_Turicchi_Parkland_Center_for_Clinical_Innovation, sorry for the delay in response

If you take a look at the Geometry schema, there is a column called polygon_class it is possible that these POI are marked as shared, but I would need to look into it.

> There are 2 possible values: 1) OWNED_POLYGON: indicates that the polygon describes the shape and size of the POI itself, and only one POI maps to this distinct polygon excluding that POI’s children. However, if we do not have a POI’s co-tenant in our dataset, it is possible that a POI will have the OWNED_POLYGON value but, in reality, the polygon includes another tenant. 2) SHARED_POLYGON: indicates that at least two POIs share the same polygon, and the POIs sharing the polygon are not “parents” nor “children” of each other. See the Polygon Class section of the Places Manual for more on how we think about polygon classification.
I will try to look over these POI in the next few days and give you a more definite answer.

1 Like

@Jack_Lindsay_Kraken1 thank you for the information! I had already filtered out POI that were classed as shared polygons and those that included parking lot, so that’s why I was thinking these could be misclassified or inaccurate. For context, we are using square footage from the geometry supplement and peak visitors from visits_by_each_hour to evaluate how crowded a place gets over a week.

The locations I sent over had abnormally high square footage or number of visitors (e.g. Galindo Tacos, regularly 4,000+ weekly raw visits to a taco truck) - the first two locations looked like apartment complexes misclassified as restaurants, the third had abnormally high raw weekly visitors, and the fourth and fifth had abnormally high square footage. Thanks again for looking into this!

Hi @Alex_Turicchi_Parkland_Center_for_Clinical_Innovation, Sorry for the delay in response. For now, is it possible to continue your research by just filtering out the outliers? The SafeGraph backend team is adding these POI to the corrections to be fixed as soon as possible. Sorry for the trouble.