How we build the demand surface
The Mobvynt demand surface is built on H3 hexagonal spatial indexing, open standard data fusion, and a demand scoring model designed for the analytic precision transit and MPO planners need.
H3 hexagonal spatial indexing
Traditional traffic analysis zones (TAZs) are arbitrary polygon boundaries drawn for administrative convenience. They don't respect the actual spatial distribution of demand and vary enormously in size and shape across a region.
Mobvynt uses Uber's H3 hexagonal grid at resolution 8, which creates a uniform coverage of cells averaging 0.74 km2 each. Every cell is the same shape, every neighbor is equidistant, and the hierarchy is consistent from the block level to the metropolitan region.
- Resolution 8: 0.74 km2 average cell area (~600m across)
- Hierarchical: resolution 7 (5.16 km2) for regional views
- Consistent neighbor distance: no edge-effect artifacts
- TAZ compatibility export available for four-step model integration
The demand index calculation
Each hex cell receives a demand index from 0 to 100 based on multimodal activity within that cell. The index normalizes raw counts to make corridors across a region comparable.
Transit boardings weight (40%)
Alighting and boarding counts from GTFS-RT stop events, normalized by cell area and averaged over the analysis period. High-frequency stops with consistent boardings score the highest.
Active mode volume weight (35%)
Bike-share trip originations and intersection pedestrian counts within each hex cell. Pedestrian activity is a strong proxy for latent transit demand in corridors without existing service.
Employment and residential density weight (25%)
LEHD employment density and ACS household density provide the structural demand context. High density without corresponding transit activity identifies candidate gap corridors.
Coverage gap threshold
A coverage gap is flagged when a hex cell has a demand index above 50 and the nearest scheduled transit service has a headway exceeding 30 minutes (default threshold, configurable by agency). Equity-weighted gap scoring multiplies the standard gap score by a factor derived from the Title VI population share in that cell.
How the model is validated
Demand surface outputs are compared against historical boarding data where agencies provide AVL records. Backcasting — applying the model to a historical period and comparing predicted gap scores against observed ridership changes — provides the primary validation signal.
Backcasting protocol
For each new agency connection, the demand surface is built for the 24 months before the current period. Routes modified during that period are identified. Demand index scores from the prior period are correlated with observed ridership changes to establish model calibration.
Limitations
The demand surface reflects revealed mobility (where people currently travel) weighted by structural demand signals. It does not model latent demand for modes not yet available in a corridor. Planning judgment remains essential in interpreting gap scores against local context.
Apply this methodology to your network
A 30-minute demo runs the demand surface model against your agency's own GTFS data and returns coverage gap scores and demand indices from your actual corridors.
Request a Demo