OperationsStrategyDecember 10, 202512 min read

Fleet Optimization Strategy: The Predict-Plan-Push Model

A comprehensive framework for optimizing delivery fleet utilization using predictive analytics, dynamic rebalancing, and a dual-loop relocation strategy.

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:

1

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.

2

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.

3

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

KPIDefinitionWhy It Matters
Overall Fleet Utilization RateTotal active delivery hours / Total available scooter hoursCore asset productivity measure
Zonal Utilization RateUtilization rate calculated per operating zoneHighlights imbalances between zones
Average Scooter Idle TimeAverage duration scooters remain idle in each zoneHigh idle time = missed revenue opportunities
Average Relocation CostTotal ops cost / # scooters relocatedMeasures cost efficiency of balancing

Demand Matching & Rider Experience KPIs

KPIDefinitionWhy It Matters
Demand-Supply Mismatch Index (DSMI)Riders seeking scooters / Available scooters per zone during peakCore indicator of solution performance
Rider Wait TimeAverage time searching for available scooterDirect rider experience measure
Rider Satisfaction ScoreFeedback scores on ease of finding scootersOverall outcome goal
Lost Revenue Opportunity# unfulfilled delivery requests due to scooter unavailabilityQuantifies 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.

K

Kaiross Team

Operations & Strategy Insights

MORE INSIGHTS

Explore more articles on operations strategy and technology solutions.

VIEW ALL POSTS