Key Takeaways
- Spatial-temporal imbalance causes simultaneous scooter surpluses and deficits across zones
- The Predict-Plan-Push model creates a closed-loop system for dynamic fleet balancing
- Dual-loop strategy: rider incentives for short moves, ops trucks for large relocations
- Machine learning forecasts demand hotspots 1-3 hours in advance
- Key KPI: Demand-Supply Mismatch Index (DSMI) measures balancing effectiveness
The Problem: Fleet Utilization Imbalance
Delivery scooter fleets experience significant variance in utilization across different operating zones. Certain zones exhibit prolonged periods of scooter idle time, while others face high rider demand leading to scooter unavailability.
Impact of Supply-Demand Mismatch
Operational Inefficiency
Resources/scooters not optimally deployed across zones
Lost Revenue
Missed deliveries due to scooter shortages in high-demand areas
Poor Rider Experience
Gig workers face frustration searching for available scooters
The Core Challenge:
A classic spatial-temporal imbalance causing simultaneous scooter surpluses in Zone A and deficits relative to rider demand in Zone B.
The Predict-Plan-Push Model
The Predict-Plan-Push model creates a closed-loop system to dynamically align scooter supply with anticipated rider demand:
PREDICT
Forecast where riders will need scooters
Anticipate rider demand across different zones for the next 1-2 hours, identifying areas likely to need more scooters soon.
PLAN
Map out the smartest way to move scooters
Determine the most efficient and cost-effective way to relocate scooters to meet predicted demand, considering battery levels, staff availability, and budget.
PUSH
Get scooters to the right places
Actively move scooters by encouraging riders with small bonuses for short relocations (< 3 km) and deploying operations trucks for larger moves.
Key Performance Metrics
Utilization & Efficiency KPIs
| KPI | Definition | Why It Matters |
|---|---|---|
| Overall Fleet Utilization Rate | Total active delivery hours / Total available scooter hours | Core asset productivity measure |
| Zonal Utilization Rate | Utilization rate calculated per operating zone | Highlights imbalances between zones |
| Average Scooter Idle Time | Average duration scooters remain idle in each zone | High idle time = missed revenue opportunities |
| Average Relocation Cost | Total ops cost / # scooters relocated | Measures cost efficiency of balancing |
Demand Matching & Rider Experience KPIs
| KPI | Definition | Why It Matters |
|---|---|---|
| Demand-Supply Mismatch Index (DSMI) | Riders seeking scooters / Available scooters per zone during peak | Core indicator of solution performance |
| Rider Wait Time | Average time searching for available scooter | Direct rider experience measure |
| Rider Satisfaction Score | Feedback scores on ease of finding scooters | Overall outcome goal |
| Lost Revenue Opportunity | # unfulfilled delivery requests due to scooter unavailability | Quantifies bottom-line impact |
Root Cause Analysis
High Demand, Low Availability (Shortages)
Peak Hour Shortages
Root Causes
- • Predictable demand surges (lunch/dinner)
- • Insufficient scooters in high-demand zones
Impact
- • Lost orders
- • Rider frustration
- • High wait times
Unpredictable Demand Spikes
Root Causes
- • Events (concerts, matches), weather changes
- • Slow operational response to relocate fleet
Impact
- • Exacerbated shortages
- • Missed peak moment opportunities
Idle Scooters, Low Demand (Surpluses)
Off-Peak Idle Time
Root Causes
- • Low order volume post-peak
- • Scooters concentrated in low off-peak demand areas
Impact
- • Low asset ROI
- • Wasted availability
- • Higher charging costs
Rider Shift End Patterns
Root Causes
- • Riders leaving scooters near homes
- • Residential areas with lower subsequent demand
Impact
- • Scooters stranded in low-demand areas
- • Requires manual relocation
Inefficient Operations & System Management
Information Asymmetry
Riders lack real-time visibility into high-demand areas. Ops team lacks predictive insight for proactive moves.
Slow Manual Relocation
Time for manual decisions, physical transport, charging, and maintenance creates delays.
Inadequate Forecasting
Lack of robust short-term demand forecasting. Fixed allocation rules not adapting to conditions.
Maintenance Bottlenecks
Manual triage for repairs, spare parts delays, lack of visibility to prioritize repairs in critical zones.
Dynamic Fleet Balancing System
A comprehensive, technology-driven solution integrating predictive analytics, real-time data monitoring, and a coordinated Dual-Loop Relocation Strategy.
Predictive Demand & Supply Engine
- • ML models analyzing historical data (trip patterns, time-of-day, day-of-week)
- • Real-time inputs: weather, ongoing orders, traffic, local events, rider search activity
- • Forecasts demand hotspots and supply gaps 1-3 hours in advance
- • Calculates predicted Demand-Supply Mismatch Index (DSMI)
Operations Command Center
- • Centralized real-time dashboard for operations teams
- • Live map with color-coded scooter status (Idle, In-Use, Charging, Maintenance, Low Battery)
- • Zone-level KPIs: Utilization %, Idle Count, Real-time & Predicted DSMI
- • AI-driven, prioritized relocation suggestions based on predicted impact
- • Alert drawer for critical imbalances
Rider App Enhancements (Rider-Driven Loop)
- • Enhanced map with "demand glow" hotspots showing high-demand areas
- • Targeted incentives: bonus credits, surge priority badges
- • Encourage riders to pick up idle scooters and relocate to deficit zones
- • Clear accept/decline flows and adjusted cross-zone fee policies
Ops Relocation Process (Ops-Driven Loop)
- • Dedicated tools and SOPs for bulk relocations using vans/trucks
- • Mobile app for field operatives with prioritized pick-up/drop-off tasks
- • Handles large-scale imbalances and pre-positioning for major demand shifts
- • Addresses situations unsuitable for rider incentives
Integrated Maintenance Optimization
- • "Maintenance Fast-Lane" prioritizes repairs for scooters in deficit zones
- • Mobile technician dispatch when maintenance queues exceed threshold
- • Coordinates battery swap routes with relocation tasks
- • Minimizes vehicle downtime and redundant trips
Why This Solution Works
Combats Reactive Nature
Predictive engine anticipates imbalances before they become critical.
Increases Efficiency
Command Center provides actionable insights, AI automates decisions.
Resolves Information Asymmetry
Both ops and riders gain real-time and predictive visibility.
Mitigates Shortages & Surpluses
Proactive moves via dual loops reduce peak shortages and idle surpluses.
Optimizes Asset Utilization
Reduces idle time, maximizes ROI for each asset.
Leverages Gig Workforce
Rider incentives provide scalable, cost-effective distributed rebalancing.
THE BOTTOM LINE
The Predict-Plan-Push model transforms reactive fleet management into a proactive, data-driven system that anticipates demand and optimizes resource allocation in real-time.
By combining predictive analytics with dual-loop relocation strategies, organizations can significantly reduce idle time, improve rider satisfaction, and capture previously lost revenue opportunities.
Kaiross Team
Operations & Strategy Insights