.svg){kind=small}
Know where riders
will be next.
Levy forecasts demand by zone and time block, then turns the forecast into rebalancing, charging, pricing, and expansion actions your operations team can use.
Tomorrow demand forecast
9 AM - 12 PM - Downtown grid
Stage
8 vehicles
Charge
5 before 8 AM
Price
+12% peak
Forecast Inputs
Not just a heatmap. A demand model tied to fleet reality.
Demand only matters if you can serve it. Levy models riders and vehicle availability together so operators can see both opportunity and constraint.
Historical rides
Completed trips grouped by H3 zone, day of week, and time bucket.
Vehicle availability
Effective supply, battery, offline state, utilization, and constrained inventory.
Weather and calendar
Weather fetches and event windows explain demand spikes before the shift starts.
Revenue outcomes
Forecasts are scored against actual rides, missed demand, and net revenue lift.
Operational Loop
Every forecast turns into a shift plan.
The model is useful only when it changes what operators do before the peak starts.
Stage vehicles before demand
Move vehicles into predicted pickup clusters before riders open the app.
Charge what will earn
Prioritize charging and swaps for vehicles closest to tomorrow's high-value zones.
Adjust price windows
Use forecast confidence to set time-of-day and zone-based dynamic pricing.
Prove accuracy
Compare predicted rides, actual rides, unmet demand, and revenue outcomes after each block.
The Product
Forecasting already lives inside the Levy analytics stack.
Operators can pull forecast points by horizon, render predicted and actual H3 cells, review unmet demand, and use AI ops jobs to keep the model fresh.
Cells
312
Hot
24
Unmet
17
Lift
+9%
Hotel Row
Stage 8 scooters by 8:45 AM
Transit Plaza
Move 5 low-battery vehicles out
Campus East
Open bonus zone for rebalancers
Move vehicles before riders ask for them.
Turn ride history, weather, and fleet constraints into a daily deployment plan.
Proof & readiness
Built on the existing Levy AI ops pipeline.
Forecasts are generated from real fleet data, mapped into H3 zones, and compared against actual rides and unmet demand after each time block.
Works with
What ships
- 1h, 4h, and 24h demand forecast API
- Predicted vs actual map layer
- Unmet-demand overlay
- Forecast accuracy dashboard
- Operator recommendation feed
Deployment timeline
- 1
Week 1 — Data readiness
Ride history, zones, weather, and availability jobs validated.
- 2
Week 2 — Model baseline
Forecasts generated for your active service areas.
- 3
Week 3 — Ops loop
Rebalancing, charging, and pricing recommendations tuned.
- 4
Week 4 — Review lift
Predicted vs actual performance reviewed with operators.
Controls included
- Forecast confidence intervals
- Actual ride comparison
- Unmet-demand tracking
- Weekly model and ops review
Keep exploring
Related products
Turn on demand forecasting
Tell us where your fleet operates and what decisions you want the forecast to drive.