Wondering why that is and what it means for the accuracy of the visit data in these type of cases?

Hi all! I am looking at a specific set of POIs around the August 2020 release and I am wondering about the effect of the naics code update on the patterns data. In particular, there is a number of locations (~1500) that are (newly) assigned the naics code 522298 (All Other Nondepository Credit Intermediation) with the 08/2020 release, and many of them have a naics code 453310 (Used Merchandise Stores) in the prior releases. With the reclassification, they disappear from the patterns data. I looked into a few examples (attaching two here) where the POI is in a multi-tenant building according to Google Street View. The change in the naics codes apparently changed the way the visitors get allocated in these cases. Wondering why that is and what it means for the accuracy of the visit data in these type of cases?

Hi @Daniel_Rettl_University_of_Georgia! You are correct that the change in NAICS codes could very well have impacted the visit attribution. The idea is that some types of places are more likely to see traffic at various times of the day than others (think a bar vs. a grocery store at 11pm). For a more in-depth discussion on visit attribution, check out the visit attribution whitepaper.

However, I would expect the backfills to fix any big changes in visits caused by newly assigned NAICS. Which Patterns data (weekly/monthly) and which release did you use?

Also, a little Slack tip: you can use shift + enter to get a new line without sending the message :slight_smile:

Thanks @Ryan_Kruse_MN_State!!! Will check out the whitepaper.

I am looking at weekly places patterns (for data from 2020-06-15 to 2020-11-30). For the two affected places in my examples, there seems to be only one data point for Quik Pawn Shop in the week ending on 2020-07-13 in the weekly places patterns backfill for Dec 2020 and onward release, no other data.

This seems consistent with the idea that the algorithm used to allocate visitors based on naics codes and times has not changed (hence those places disappear in the data release from 2020-06-15 to 2020-11-30 and also in the backfill). I am wondering whether this means that for multi-tenant buildings, POIs with low priority naics codes at a given time will systematically see (=assigned by the algorithm) fewer visits than comparable POIs that are in a stand-alone location (and fewer visits than they actually may have received)?

Also thanks for the Slack tip, works :slight_smile:

Hi @Daniel_Rettl_University_of_Georgia, I would recommend you use the “Weekly Places Patterns Backfill for Dec 2020 and Onward Release” for your analysis. This is the newest and most up-to-date version of Patterns, and it has been backfilled all the way back to January 1, 2018. It should fix the issue you are seeing from newly assigned NAICS codes.

I don’t know for certain, but I believe the answer to your question is yes. I don’t know of any work that has been done to verify, but intuitively that is what I would expect to happen. If you happen to do anything that gives more insight into this question, let me know!