What are the most frequent reasons for POIs disappearing between Core updates besides removing bugs? For example, if a POI relocates, is that one way for it to disappear?

Hi, I’m studying a subset of POIs. I found that 672 of these POIs in the March 2020 Core Places are not in the latest 2021 Core Places. Among these, 93 appear in the 09/2019 Monthly Patterns data as well. None of these 93 POIs indicate being permanently closed. What are the most frequent reasons for POIs disappearing between Core updates besides removing bugs? For example, if a POI relocates, is that one way for it to disappear?

Hi @Alex_Zentefis_Yale_University, I am sure you are familiar with the changelog, but here is a link just in case - -Changelog – this will show bug fixes/closings/etc

To your point, if a location relocates, it does in fact get a new SGPID because it is a new place (this should not be confused with the same location being assigned 2 different SafeGraph place IDs – this cannot happen). I do not have a master list of reasons the SGPID might be in 1 core and not the other, but the changelog is your best place to monitor for changes

Thanks, @Jack_Lindsay_Kraken1! I’ll check with my colleagues to see if they have any follow-up questions on this topic and get back to you if they do. Appreciate your help!

Hi @Jack_Lindsay_Kraken1. We’re trying to line up Core Places with Patterns data. We’re focusing on calendar year 2019. We’re thinking of using the most recent Core Places available and using the backfilled monthly Patterns. Related to what I asked earlier, suppose a bank branch was in business for all of calendar year 2019, but closed in January 2020, and then a smoothie shop opened in the same location in March 2020. First, will the bank and the smoothie shop have the same placekey? Second, Patterns data that is backfilled - will it reflect people in 2019 going to the bank, the smoothie shop, or both?

That’s great! Thanks @Jack_Lindsay_Kraken1! Because the Patterns backfill reflects these two distinct placekeys, then, some Patterns data will be missing for one placekey when it is present for the other and vice versa, right?

Using this example again, during the calendar year 2019, the bank’s placekey will have visits (among other Patterns data), while the smoothie shop will have no Patterns data for calendar year 2019 because it didn’t exist yet. Similarly, the bank’s placekey will have no Patterns data from January 2020 onward (because it closed), whereas the smoothie shop’s placekey will begin displaying Patterns data in March 2020 (when it opened in the bank’s former location). Am I correct here?

Hi @Alex_Zentefis_Yale_University, sorry for the delay here.

Rather than thinking of patterns missing for 1 placekey and not the other, think of it like the Patterns data released will not log that Placekey/POI at all because it doesnt exist

While this is semantics in its purest form, the distinction could be important haha. Your assertion sounds correct - we should see visits to the bank in 2019 and then a switch to visits to the smoothie shop and no visits to the bank since it will be determined as closed.

No problem! @Jack_Lindsay_Kraken1. Last couple of questions, just so that I’m 100% sure I understand how Core and Patterns relate when a business closes and opens:

  1. Suppose I use the latest Core Places and study a barber shop that opened in 2021. Will this barber shop show up during the calendar year 2019 in the Patterns back fill data?
  1. Conversely, suppose a candy store was open during the first six months of 2019 but closed in June 2019. Will this candy store be logged in the latest Core Places (from 2021) as permanently closed? In the back filled Patterns data, will the candy store’s placekey have Patterns data during the time it was open in 2019, but then stop having Patterns data (i.e., stop being logged) once it closed?

Really appreciate your help and time!

For number 1:

it will either show up with 0 visits or not show up at all – I believe it will not show up at all in 2019 patterns data

For number 2:

Yes, it should show the closed since and reflect the pattern data for the time it was opened only. However, if the candy store is, for some reason, not in the Core file used for the backfill, it will not show up in any of the backfill.

I hope this makes sense, remember POI are added and removed all the time, so the version of the core used for the backfill is extremely important. If a specific location is crucial to your analysis and you need whatever data is available for it, you can check the older files and their appropriate Core files to get that data (Note: this is only recommended if the POI you are looking for is crucial in your analysis)

Great! Thank you, @Jack_Lindsay_Kraken1.