Understanding POIs in without Visits in Patterns Data

SafeGraph Team,

We are researchers trying to understand the Patterns data. We want to know which POIs (say, in the financial services) were open in any given week during 2018-2022. For example, in a certain location, how many POIs with a specific NAICS code were operating the first week in January 2019.

To make sure we do not count POIs that are not in existence at that point in time, we were hoping to get some clarification with these data questions.

  1. If a POI in the patterns data has missing values for the opened_on, closed_on, and tracking_closed_since fields in a particular week, should we consider that POI to be available for customers in that time period?
    a) We’ve noticed a fair number of POIs in certain NAICS codes that always have missing values for raw_visit_counts (at least in the data we are using). If the raw visit counts are always missing for a given POI and the opened_on, closed_on, and tracking_closed_since fields are missing, is it still safe to consider those POIs to be open and available for customers?
    b) Obviously, this can happen if these POIs simply receive low traffic and happen to not be visited by anyone in the Patterns data panel sample of devices. But we just wanted to confirm.
    c) By way of example, consider the following placekeys:
    i. 222-222@3bt-bys-gzf (NAICS: 4453 (Beer, Wine, and Liquor Stores))
    i. 222-222@3bt-by4-tgk (NAICS: 5412 (Accounting, Tax Preparation, Bookkeeping, and Payroll Services))
    i. 22c-222@5s5-wm4-7wk (NAICS: 6212 (Dentists))
    i. 22c-222@5s5-sb2-49z (NAICS: NULL)
    i. 22q-22f@5z6-2qd-nkf (NAICS: 6215 (Medical and Diagnostic Laboratories))
    i. zzy-222@3wy-pyg-6tv (NAICS: 7211 (Traveler Accommodation))
    d) All of these placekeys have a value of “2019-07” in the tracking_closed_since field. But, at least in the Patterns data we are looking at, the raw_visit_count fields are always missing, even after July of 2019. Note: we are currently only looking at the first week of Patterns data in each month.
  2. As far as SafeGraph is aware, these locations are open and available for business?
    a) We have reviewed the closed_on and opened_on documentation. But we are concerned that, depending on the input feed for the various POIs, the data may contain business locations in, for example, the January 2018 data but in fact were only open after that (and this would not be reflected in the opened_on field due to its conservative nature). Is this possible? And if so, how would we best identify these POIs?

I hope this makes sense. Looking forward to learning more!

Hi @michael_gropper, thanks for your question.

The SafeGraph documentation has a detailed section on the Open/Close logic, which also helps explain why 2019-07 is a common data in the data.

Hi Evan,

Thank you for your response. We understand why July 2019 is a common value in the tracking_closed_since field. But (and maybe I am missing something) I do not think the documentation referenced answers the question though.

Consider, by way of example, placekeys: 222-226@5q9-b4f-94v (a dentist office), and 223-224@5x4-8tx-h3q (an H&R Block). These two placekeys have missing values for opened_on, closed_on, and tracking_closed_since throughout the data we are looking at. And they never have any visits in the patterns data. Is it SafeGraph’s assumption that these places are open and available for business throughout the timeframe?

Michael, you’re thinking about it correctly. If the values for these fields are blank the assumption is they have been open and available during this timeframe (and today). If SafeGraph’s build pipeline were to determine a POI was closed since the release of this column, the field would be marked as closed. Of course, there isn’t a guarantee that it is 100% accurate but that’s the way to read the data, and their docs mention a 60 day margin of error on closures.

There are a number of other reason why an open POI may not have Patterns visits associated with it (low number of visits, POI-type restrictions, etc.)