POI Counts Clarification

Hi, looking at this summary statistics, I’d like to know how you count the total POI and precisely how we should interpret these results.

  • Does each bar reflect the cumulative number of POI reached until the month considered?
  • How do you handle POI of which you do not have data anymore? Are they excluded from the cumulative, or are they still counted in the total number of POI reached? (for example, you have collected data from the XXX POI until mar-22, is it still in the count of Apr-22 or not?)

Thank you for your help!


@andrea.savio I believe this count represents the total number of POI in the SafeGraph Places dataset at the start of that month. That includes businesses that have closed (which should then have closed_on data in the file).

So basically it is the count of the total number of distinct Placekeys with data, meaning that you could have two unique POI/Placekeys that represent the same address is one existed previously and a new one is there today.

@evan_barry_dewey So, when a POI does not appear in the following months, could the cause be only the closure of it? Or could it also be the possibility that after following it for some months, it will not share the data at a certain point anymore?
And, if this is the case, is this POI counted in the total count of the figure or not?

@evan_barry_dewey In addition, this is what I obtain when I look at the number of unique POI monthly. Can you confirm is that correct?

If you’re looking at the Places data, the number of POI should effectively always be increasing since even when POI closes, SafeGraph continues to keep that POI in the dataset into perpetuity. They’ve only been tracking closed_on dates for the last few years.

Here is SafeGraph’s logic regarding open/close: Rich Attributes | SafeGraph Docs

Note: Just because a POI exists in the Places dataset, it doesn’t mean that it has Spend or Patterns data. Both of those datasets have data for a subset of the overall Places POIs.

I don’t know why your graph wouldn’t match the one provided by SafeGraph. How are you filtering this down from just counting the total number of rows in the data?

Here another resource I found on open/close data:

@evan_barry_dewey Ok, thank you! Yes, what I’m doing is just count the total number of rows in the SP data.