FAQs - pass_by

Where can I find documentation for TransUnion's profiles, classification system, and percentage breakdowns?

You can locate TransUnion CAMEO Handbook - Profiles and TransUnion CAMEO Table - Profiles in the Links section under the About menu.

What makes Pass_by data different from other foot traffic providers?

Pass_by leverages a multi-source and model-intensive approach that goes beyond raw GPS signals. It integrates:

  • Mobile GPS data from multiple providers (~175M daily active users),
  • POI polygons from SafeGraph,
  • Sensor data from 50K+ stores, and
  • Consumer demographic and psychographic profiles from TransUnion.

Unlike competitors that rely exclusively on raw location data, Pass_by validates and calibrates its models using real-world sensor data, making their output more reliable and less susceptible to noise or fraud.

What data products does Pass_by provide on Dewey?

Pass_by publishes three primary datasets on Dewey:

  • Store Visits: Hourly, daily, and weekly foot traffic counts to 1.5M+ branded POIs in the U.S.
  • Store Visitors: Demographic and psychographic profiles of store visitors, aggregated at the store level.
  • Cross-Shopping: Patterns of visitor overlap across brands, e.g., % of McDonald’s visitors who also went to Walgreens.

How does Pass_by clean and filter low-quality location signals?

They apply a multi-step process:

  • Spatial-temporal clustering of signals to define likely "visits".
  • Polygon-based attribution using SafeGraph POIs for high spatial precision.
  • Scoring and weighting based on visit plausibility (e.g., device patterns, time of day) and data quality.
    Low-quality or suspicious signals (e.g. from spoofed apps or devices) are downweighted or discarded.

How does Pass_by model actual visit volumes?

Three main stages of modeling are applied:

  1. Observed Visits: Filters GPS signals to identify real-world visits.
  2. Shape Modeling: Uses NeuralProphet (a neural-enhanced time series model) to estimate seasonality, trends, and noise.
  3. Volume Scaling: Extrapolates visits using trade area size, population density, and POI type (e.g., fast food vs. luxury retail).

Models are calibrated and validated using real sensor data from thousands of locations.

Why is sensor data important to Pass_by’s process?

Sensor data provides the ground truth for foot traffic: physical hardware in 50K stores tracks in/out counts. Pass_by doesn’t own this data but accesses it securely via APIs. These sensor benchmarks are used to:

  • Calibrate machine learning models during training,
  • Validate predictions post-modeling,
  • Improve shape (correlation = 0.85) and volume estimates.

How does Pass_by estimate demographics and psychographics of store visitors?

  1. Home location inference: Devices are assigned a probable home based on nighttime dwell patterns.
  2. Zip+4 mapping: Home locations are linked to TransUnion ZIP+4 profiles.
  3. Attribution: Each device gets a probabilistic demographic + psychographic profile (56 personas, 20+ traits).
  4. Aggregation: Profiles are aggregated by store to ensure privacy.

This process allows rich insights while avoiding individual-level tracking.

Can I match Pass_by POIs to SafeGraph Placekeys or company tickers?

Yes:

  • While Placekeys aren’t natively used, Pass_by POI IDs can be joined 1:1 with SafeGraph identifiers.
  • For finance users, Pass_by offers support for mapping brands to ticker symbols upon request.

How accurate is the Pass_by data?

Pass_by reports the following validation metrics:

  • Shape accuracy (correlation with sensor time series):

    • Daily: 0.85
    • Hourly: ~0.83
  • Volume accuracy (ranking correlation with sensors):

    • Kendall rank: 0.5
  • Demographic representativeness:

    • Mean error from U.S. population across age, income, race: 1.35 percentile points

What is the historical coverage of the dataset?

  • Standard offering on Dewey: Rolling 5 years of data
  • Update frequency: Monthly

Each row of data represents one week of activity, updated monthly (typically the first Monday of the month).

Are hospitals or other sensitive POIs included in the dataset?

No. For privacy and ethical reasons, Pass_by excludes:

  • Hospitals and health-related POIs
  • Political buildings and other sensitive categories

This is a blanket exclusion to protect sample panel privacy. On a case-by-case basis, some exceptions may be reviewed for approved research with enhanced safeguards.

What are typical research and industry use cases for this data?

Retail:

  • Store operations (staffing, inventory)
  • Site selection
  • Competitive benchmarking
  • Financial forecasting

Commercial Real Estate:

  • Prospecting tenants for vacant retail space
  • Lease negotiation support
  • Understanding foot traffic dynamics by store or zone

Finance:

  • Foot traffic to public companies (correlate to sales or revenue)
  • Quantitative strategies using real-world consumer behavior