Transforming Luxury Automotive Service

With Machine Learning
8-Week
Engagement
30%
Capacity Increase
3-Day
Wait Reduction
35%
Backlog Decrease

THE CHALLENGE

Service advisors couldn't predict repair times until after diagnosis, creating a "chicken-and-egg" problem: they needed to diagnose to schedule, but needed to schedule to diagnose.

Problem Context

"Our service bays were only operating at 70% capacity despite having a 3-week backlog of vehicles. Without knowing exactly how long repairs would take, we couldn't effectively schedule our technicians or give customers reliable pickup times."

Patricia

Service Manager, Lucid Service Center

Unpredictable Service Timing

Inefficient Capacity Planning

Growing Vehicle Backlog

Guesswork Scheduling

Customer Frustration

Administrative Overload

The Solution

I helped orchestrate a solution that integrated the Data Science team's machine learning model with operational workflows to solve the fundamental "chicken-and-egg" problem of service scheduling. The system translated everyday customer language into accurate repair time predictions.

Solution Flow

Customer Concern

Input

ML Engine

Processing

Scheduling System

Optimization

Appointment

Output

System Components

BRAIN

ML prediction engine that analyzed customer descriptions and predicted repair requirements and service times with 85% accuracy.

HOW IT WORKS

The system uses natural language processing to convert customer concerns into service predictions:

  1. Text Processing: Customer complaints are normalized, tokenized, and analyzed for key terms.
  2. Pattern Matching: The system compares concerns against thousands of historical service records.
  3. Service Prediction: Using TF-IDF and word embeddings, the model identifies the most likely repair operations.
  4. Time Estimation: The system calculates FRT using the 85th percentile of historical completion times for similar services, ensuring realistic scheduling.
  5. Confidence Scoring: Each prediction includes a confidence rating, with low-confidence cases flagged for advisor review.

CAPACITY ENGINE

Scheduling system that optimized technician workloads and service bay utilization based on predicted service times.

HOW IT WORKS

The capacity engine applies mathematical optimization to scheduling:

  1. Dynamic Capacity Calculation: Continuously calculates available capacity using the formula: Available Capacity = (Productivity% × Tech Hours) - Internal Hours - Same-Day Hours - WIP Hours.
  2. Constraint-Based Scheduling: Ensures appointments meet multiple constraints including buffer days (minimum 3-day lead time), technician availability, and service center intake hours.
  3. WIP Management: Tracks all vehicles currently in service, their estimated completion times, and dynamically adjusts available capacity as work progresses.
  4. Configuration Controls: Allows service managers to set productivity factors (typically 70-85%), reserve blocks for internal work, and allocate same-day emergency capacity.
  5. Queueing Theory Application: Uses Kingman's formula to balance utilization and wait times, proving mathematically that more precise FRT predictions allow both higher utilization and shorter waits.

REMOTE SERVICE

On-location service identification system that allowed pre-diagnosis of issues before vehicle arrival.

HOW IT WORKS

The remote service component optimized mobile service operations:

  1. Mobile Capability Assessment: Analyzes customer concerns to determine if issues can be resolved through mobile service rather than in-center.
  2. Drive Time Optimization: Calculates travel times between service locations using geospatial algorithms, optimizing technician routes to minimize non-billable travel.
  3. Mobile Capacity Planning: Incorporates both FRT and travel time when calculating mobile technician capacity, ensuring realistic scheduling.
  4. Territory Management: Maps service territories and coverage areas, identifying scheduling opportunities in "white space" areas where mobile service can expand reach.
  5. Mobile Parts Management: Predicts parts needed for mobile appointments to ensure technicians carry appropriate inventory for scheduled work.

USER INTERFACES

Intuitive screens for service advisors and customers that simplified scheduling and provided transparency.

HOW IT WORKS

The interface layer transformed complex data into actionable information:

  1. Appointment Creation: Streamlined interface where advisors enter customer concerns and receive instant FRT predictions with confidence indicators, surfacing appropriate appointment slots.
  2. Configuration Dashboard: Management interface for setting productivity factors, buffer days, intake hours, and other parameters that define scheduling constraints.
  3. Capacity Visualization: Interactive calendar showing available capacity across days and service centers, with real-time updates as appointments are booked.
  4. Technician Workload View: Visual representation of technician schedules, skill levels, and utilization rates to identify optimization opportunities.
  5. Customer Interface: Simplified version of the scheduling system for the customer mobile app, allowing self-service appointment booking based on the same capacity engine.

Process Timeline

Week 1-2

Discovery

Conducted interviews with service advisors and analyzed 18 months of historical service data to identify patterns.

Week 3-4

Design

Developed system architecture and applied queueing theory to optimize scheduling algorithms.

Week 5-6

Prototype

Built interactive mockups and tested the ML model with real service advisors using historical data.

Week 7-8

Documentation

Created detailed product requirements document and technical specifications for implementation team.

Interactive Solution Preview

Explore this service scheduling dashboard that I created myself using Claude and Next.js. This interactive prototype allows service advisors to visualize capacity, monitor technician utilization, and manage the workflow of vehicles in the service center.

Key Insights

Integration Over Creation

"The most valuable approach was integrating with existing systems rather than replacing them, reducing resistance and accelerating adoption."

Queueing Theory Application

"Applying queueing theory principles to service bay scheduling was the mathematical breakthrough that unified previously competing priorities."

Cross-Functional Leadership

"Success required bridging departmental silos through a leadership approach that balanced technical expertise with stakeholder management."

Challenges & Solutions

Challenge

Organizational Silos

Service, sales, and parts departments operated independently with no shared data access

Solution

Created a cross-functional team with representatives from each department to guide development

Challenge

UX Resources

Limited UX resources to design interfaces for both service advisors and customers

Solution

Prioritized service advisor interface first, then created simplified customer view from the same data model

Challenge

Technical Gaps

Legacy systems had no APIs for integration and limited historical data access

Solution

Developed a data extraction layer that could work with existing systems without requiring rewrites

Challenge

Competing Priorities

Management wanted both increased capacity utilization and reduced customer wait times

Solution

Applied queueing theory to demonstrate how precision in scheduling would achieve both goals simultaneously

The Results

Business Impact

Capacity Utilization+30%
Wait Time Reduction3 Days
Backlog Reduction-35%

Operational Excellence

Service departments could now manage their resources efficiently, improving both capacity utilization and worker satisfaction.

Customer Experience

Customers received more accurate time estimates and faster service, increasing satisfaction scores by 15%.

METRICS & PERFORMANCE

Prediction Quality

  • Mean Absolute Percentage Error: <15% for routine maintenance, <25% for complex repairs
  • Override Rate: Reduced from 65% to under 20% after 3 months
  • Confidence Score Accuracy: 92% of high-confidence predictions completed within estimated time

Operational Efficiency

  • Technician Efficiency: 20% improvement in FRT-to-actual-hours ratio
  • Admin Time: 70% reduction in scheduling-related administrative tasks
  • Mobile Service Optimization: 30% increase in mobile appointments completed per day

Business Impact

  • Capacity Utilization: 30% increase in billable hours per bay
  • Appointment Wait Time: Reduced from 14 days to 3 days average
  • Aging Vehicles: 35% reduction in vehicles remaining over 10 days