Discussions

We would appreciate clarification on the definitions of key variables (https://docs.deweydata.io/docs/safegraph-spend).

1.1. Identifying unique customers: When aggregating transactions to the POI–month level, is the “customer” count based on unique card identifiers or on inferred unique individuals? In other words, if a customer used three different cards (associated with two different banks) for a total of four times at a POI, do you count this as one, two, three, or four unique customers?
1.2. Coverage of payment methods. Does the dataset capture only debit and credit card transactions at each POI, or does it also include other payment methods such as cash, prepaid cards, or bank transfers at each POI? In particular, are unbanked consumers who transact exclusively in cash or PayPal observable in any way in the Spend Pattern Data?
1.3. POI Coverage
1.3.1. How are POIs selected for inclusion in the Spend Pattern Data? Is this a random sample?
1.3.2. What constitutes “sufficient transaction volume” or “sufficient data availability” for a POI to be included?
1.4. RELATED_PAYMENT_PLATFORM_PCT, which is described as the “percent of customers that also used specific payment platforms elsewhere during the month.”
1.4.1. How is “related” defined and constructed?
1.4.2. What exactly does “elsewhere” refer to? Does it include any transaction outside the focal POI (e.g., transactions at other physical POIs and online transactions)?
1.4.3. How is it collected? For example, if a customers use a credit card to pay at a restaurant, she is recorded as a restaurant customer. A few days later (within the same month), if this user used Venmo to transfer money to her friend via (1) a linked bank account, or (2) a Venmo account balance, can both of these transactions be captured by this variable?
1.5. Can we distinguish the percentage of customers from Manhattan and Brooklyn, separately, who have “also used specific payment platforms elsewhere during the month”?

Thank you for your time and assistance!