Skip to content

Replica Workspace Context

Replica provides data about the built environment and how people interact with it. It transforms diverse datasets (mobile location data, consumer/resident data, built environment data, economic activity data, and ground truth counts) into a holistic picture of mobility, land use, people, and economic activity. The core methodology is an activity-based travel demand model using machine learning to create a synthetic population and agent-based simulation to model movement.

Data Products

Product Scope Description
Places Seasonal, by megaregion Simulates complete activities/movements of residents, visitors, and commercial vehicles on a typical day
Trends Weekly, nationwide Census-tract-level mobility, economic activity, and land use updated weekly
Land Use Nationwide Parcel-level land use data (parcel area, building area)
AADT Nationwide Annual Average Daily Traffic
Free-Flow Speeds Nationwide Annual free-flow speed data
Quarter-Hourly Speed Profiles Nationwide Speed profiles at 15-minute intervals
TMC Nationwide Turning Movement Counts

Megaregions (12 total)

Alaska, Cal-Nev, Great Lakes, Hawaii, Mid-Atlantic, North Atlantic, North Central, Northeast, Northwest, South Atlantic, South Central, Southwest

Seasons & Temporal Coverage

  • Season definition: Spring = Mar/Apr/May; Fall = Sep/Oct/Nov. Each models a typical Thursday (weekday) or Saturday (weekend) over a 13-week period.
  • Available Places seasons: Fall 2019, Fall 2021, Fall 2022, Spring 2023, Fall 2023, Spring 2024, Fall 2024
  • No 2020 data (skipped due to pandemic; 2019 serves as pre-pandemic baseline)
  • No pre-2019 data (Fall 2019 is the earliest)
  • Trends mobility: January 2019 to present
  • Trends spend: January 2019 through March 2025 (discontinued)

Access

  • No API — BigQuery is the primary programmatic access method, available by subscription tier
  • Custom geographies can be uploaded (zipped shapefiles, KML, GeoJSON)

BigQuery Project

  • Primary project: model-159019
  • INRIX data: model-inrix (speed profiles, volume profiles, OSM conflation)
  • Common datasets: landuse, populationgen, geos, places, mobility, mini_kc (test region)

Key Terminology

  • Season: A quarterly time period (e.g., 2024_Q4). Spring = Mar/Apr/May, Fall = Sep/Oct/Nov.
  • Place: A physical location (POINT geometry) modeling "all locations a person can visit" — extends POIs to include residential locations.
  • POI (Point of Interest): SafeGraph-sourced commercial locations + Replica-added block centroids for residential.
  • Link: A road segment in the transportation network, identified by stable edge ID or OSM way ID.
  • Synthetic population: Statistically representative population built from Census microdata (ACS, PUMS, LEHD, CTPP), not real individuals.

Critical Gotchas

  • Visitors have no demographics — all person columns except id and residentType are NULL for visitors.
  • Spend data discontinued Dec 28, 2024 due to upstream source reliability issues.
  • Every trip query needs: WHERE type = 'TRAVEL' AND optionIndex = 'FIRST_OPTION'
  • Replica models typical days, not specific dates — don't compare directly to AADT or single-day counts.
  • Larger aggregations = higher confidence — filtering to very small populations or geographies increases noise.