Weekly Patterns Plus

Overview

Weekly Patterns Plus data provides the foot traffic data insights on a weekly basis, tracking data from Monday to the end of day on Sunday each week. Weekly Patterns+ are available starting from January 1st, 2017 and includes coverage in the US and Canada.

Differences between Patterns and Patterns Plus

Methodology Differences

• Advan Research uses polygons that are developed and maintained internally for visit attribution in Patterns Plus. Patterns relied on SafeGraph polygons.
• Every field in Patterns Plus is normalized for the panel.
• The concept of shared polygons does not apply to Patterns Plus. All POIs have their own unique hand drawn geofence.
• Placekeys are not currently available for POIs on Patterns Plus. See id_store and persistent_id_store for persistent place identifiers.

Notable Schema Changes

• id_store is a unique ID tied to this point of interest (“POI”). This ID is guaranteed to be persistent across the lifespan of the company, excluding M&A events. In M&A situations, a different id_store is created and the pre/post merger location is tracked separately. See field persistent_id_store below for how to handle M&A cases.
• persistent_id_store is a persistent ID tied to this point of interest (“POI”). This ID is guaranteed to be persistent across the lifespan of the location, including M&A events.
• ticker is now included for the location if available. Custom ticker is applied for private companies and subsidiaries.
• persistent_id represents a unique, human-readable identifier that is guaranteed to be persistent even across M&A events that represents this specific company.

Key Concepts

Underlying Places Data

For Weekly Patterns+ Advan Research does not use SafeGraph Places for underlying geographies and polygons; instead, they develop and maintain polygons internally. These internal polygons are used for visit attribution and are backfilled back to 2017. You can review a complete list of places on Dewey here (these are pre-joined to the foot traffic dataset). Additionally, the concept of shared polygons does not apply to Patterns+. All POIs have their own unique hand drawn geofence.

Places associated with the following NAICS codes are removed:

62%, 9221%, 9231%, 6111%, 813930, 8131% and 8133%

This is consistent with other mobility providers on the platform.

Visit Attribution

Advan Research computes the visits/visitors and other metrics inside a POI using the POI’s geometry. They do not apply any dwell time or any concept of “stops”; they rely on the polygon for accuracy. Advan Research has tested their data on 1,500 publicly traded tickers versus (a) top line revenue as reported from the companies and (b) credit card transaction counts on physical locations, and they have determined consistently that in the vast majority of cases filtering for dwell time reduces the signal and makes the correlation/forecasting worse.

Determining Home Location:

Advan Research computes a device’s home/work (night/day) location by computing the time a device spent in each building in the country; then taking the most frequented building.

Backfills

Backfill is when we take our most recent locations (i.e., addresses + geofences) and run our visit attribution algorithm backward in time to generate a new history of “backfilled” Patterns+. Backfills are typically generated every time new Advan POIs are added (typically monthly, with the exception of August and December).

Updates to Underlying POIs

Advan Research's list of POIs are updated monthly. When an updated POI list is used for generating the Weekly Patterns+, the traffic is being computed for the full history of any new or updated POIs (a “backfill”, or a “release”). The majority of the changes pertain to new POIs, but there are a number of POIs (on the order of 1% of the total) that are updated on each release; such updates typically modify the date the POI may have closed, i.e., Advan Research marks any store location that closed since the prior POI update as closed in the latest POI version, or improve the accuracy of the polygons delineating the POIs. If you require a very high level of accuracy for your historical visits comparisons, you may want to load the historical back-filled data on each release; for most use cases however, you do not need to reload the historical data. You can also employ an intermediate approach where you reload the history only on quarterly, semi-annual, or annual basis.

Additional Column Information

street_address

• Advan Research implements a number of steps to clean, validate, and standardize street addresses.
• Street addresses are expected to be title-cased, consistent, and optimized for human readability. Feedback is welcomed if otherwise observed.

city

• City names are derived from normalized address strings obtained from POI sources.

region

• When iso_country_code == US, this represents the U.S. state or territory.
• When iso_country_code == CA, this represents the Canadian province or territory.
• When iso_country_code == GB, this represents the United Kingdom county.

postal_code

• For iso_country_code == US, this is the 5-digit U.S. ZIP code.
• For iso_country_code == CA, this is the Canadian postal code in the format of a 3-digit Forward Sortation Area (FSA), a space, and a 3-digit Local Delivery Unit (LDU).
• For iso_country_code == GB, this is the British postal code. Additional information on Great Britain postal code precision can be found through Advan Research’s documentation.

visit_counts

• Represents the number of estimated visits to the POI during the specified date range (the sum of each day’s unique visitors).
• Estimates are scaled using Advan Research’s best current methodology for calculating actual visits.

visits_by_day

• An array of visits for each day of the week, Monday through Sunday.
• Days are segmented using local time.

visits_by_each_hour

• An array of visits for each hour within the week.
• Only the start of a visit is counted, so multi-hour visits are only recorded once unless the visit crosses a UTC day boundary.
• Days are segmented using local time.

visitor_home_cbgs

• Represents the home census block groups (U.S.) or dissemination areas (Canada) of visitors to the POI.
• Each entry shows the number of associated visitors (not the number of visits).
• If visits by home CBG are desired, the visitor count can be multiplied by the average visits per visitor (i.e., raw_visit_counts / raw_visitor_counts) as an approximation.
• Not every visitor has a determined home census block group, and not all visitors originate from the U.S. The column visitor_country_of_origin represents the total visitors identified from the U.S. versus Canada.

visitor_home_aggregation

• Similar to visitor_home_cbgs, but represents home census tracts (U.S.) or aggregate dissemination areas (Canada).
• Recommended for analyses where fine-level detail is not required.
  • CBGs: 600–3,000 people.
  • DAs: 400–700 people.
  • CTs: 2,500–8,000 people.
  • ADAs: 5,000–15,000 people.

visitor_daytime_cbgs

• Represents the daytime census block groups of visitors, defined as the most frequent location between 8 a.m.–6 p.m., Monday–Friday, during the calendar month.
• Values represent the number of associated visitors (not visits).

visitor_country_of_origin

• Lists the countries of origin for visitors to the POI.

distance_from_home

• Represents the median distance from home to the POI in meters, based on visitors with identified home locations.
• Calculated as the haversine distance between a visitor’s home geohash-7 and the POI location for each visit. The median value is then used.
• If a POI has fewer than 5 visitors, the value is null.
• Each visitor is counted equally; visits are not weighted.

median_dwell

• Represents the median of the minimum dwell times calculated for visits to the POI.
• Minimum dwell time is defined as the interval between the first and last device ping observed during a visit.
• Since it is based on pings, this may underestimate actual dwell time.
• A dwell time of 0 is possible if only one ping was recorded and the visit was inferred from signals (e.g., Wi-Fi).

bucketed_dwell_times

• A dictionary mapping visit counts to dwell-time ranges (in minutes).
• Includes the following bins:
  • <5
  • 5–10
  • 11–20
  • 21–60
  • 61–120
  • 121–240
  • >240

related_same_day_brand

• Represents the brands that visitors to this POI also visited on the same day.
• The value for each brand indicates the percentage of POI visitors who also visited that brand on the same day.
• Only the top 20 brands are returned.

related_same_week_brand

• Represents the brands that visitors to this POI visited over the course of the same week.
• Calculated in the same way as related_same_day_brand.

carrier_name

• A column that maps visitors’ wireless carriers.