Does anyone know why the “weekly visit patterns” data is not symmetric WRT the “related_same_day_brand”?
For example: if in a row where the poi is of let’s say “Exxon Mobil” brand and has 5 same day visits to “ConocoPhillips”, then there should be another row where poi is from “ConocoPhillips” brand and has “Exxon Mobil” as one of the “related_same_day_brand”.
I checked a couple of weeks, but I only observe one sided relationship.
Does anyone know why the "weekly visit patterns" data is not symmetric WRT the "related_same__day_brand”?
Thanks for asking this question, Moshen. We won’t include data for this column if the median daily visitors is less than 5
They are not in units of “visits” so it doesn’t need to be symmetrical. I have not thought of a good way to say what the units are in plain English (excess visit proportion wrt population average, as a %?), so here’s the stated definition: “The value shown for each brand is a percentage representing the median of the following calculation for each day in the month: (same-day visitors to both the brand and the POI / total daily visitors to the POI) - (daily visitors to the brand / all visitors in SafeGraph panel).”
Thanks @Dennis_Chao_Institute_for_Disease_Modeling and @Francisco_Utrera
So the questions are:
1- how does that formula take co-visits into account? there is no information about the second brand visits.
2- why such formulation has been used, while sharing actual co-visit counts wouldn’t harm and is more straightforward?
3- is there any way to get such co-visit information (actual same day co-visit counts)?
can you expand on 1? not sure I understand your question
I believe that differential privacy was the motivation for not sharing actual counts. However, now that I look at it, I don’t see why you couldn’t derive the same-day visitors to both brands by plugging in all the other parts of the formula. I’ll check internally and get back to you by EOD on Mon
then in the formula is “same-day visitors to both the brand and the POI” the actual co-visit counts?
yes, co-visits between that POI and the related brand. just keep in mind that it’s the value that corresponds to the median index
the question is:
is there any easy way to get such co-visit information?
I don’t know of any easy ways, but you’re not the only one asking for this. let me ask internally if we have plans to make it easy for users to get this info
yes, please, that would be great!
and one other question:
“all visitors in SafeGraph panel” is it the visitors in that particular day or in general the total number of active Safegraph users?
all unique visitors for that particular day
what are you trying to do? this is good context as we think of prioritization and common use-cases
I am trying to get the actual same co-visits counts among brands
ok, and what is your broader/end goal?
This is an important topic to study
to see how different brands get disconnected from others
I see! so you really don’t care so much about data at the poi level, right?
having poi level data would be great, but just brands and categories would also work
this is great context! yeah, other customers have mentioned something similar.
one other thing, what is the highest level of granularity for time that would work for you?
would quarterly work for you?
of course always higher resolution is more preferred, but researchers should be realistic not always data is as we want to be, so if quarterly is what we can get, yes, it works!