Case Study: Amazon Digital Service Transformation

Objective: Using Proactive Anomaly Detection & AI Automation to Achieve $250M Revenue Growth & 900bps CSAT Increase

Project Details

• Project Title: Project Odyssey (Service Scheduling & Execution Transformation)
• Company Name: Amazon
• Program: Service Fulfillment Optimization / Customer Experience Enhancement
• Location: North America (Initial Scope) expanded to EU
• Role: Senior Manager for Customer Experience
• Timeframe: 3 Years (Execution) / 5 Years (Roadmap Strategy)

Challenge

• Who: Amazon’s Home Services division, involving both internal operations teams and a network of external third-party service providers.
• What: The division faced a massive efficiency bottleneck where 50% of all orders resulted in a contact for resolution to customer service. This resulted in a DPMO (Defects Per Million Opportunities) of 500,000, meaning half of all service opportunities had a process defect.
• Where: Across the entire service fulfillment lifecycle in North America, affecting both customer-facing support and backend logistics.
• When: Prior to the implementation of "Project Odyssey," during a period of rapid scaling in 2019 through 2024.
• Why: The legacy processes were entirely manual, lacking data visibility and automated accountability. This led to high operational costs (approx. $6 per contact) and a disjointed customer experience where customers were forced to reach out to resolve basic issues like rescheduling or job completion.

Advisory

The Insight: We derived our insights by triangulating data from customer sentiment, direct provider feedback, and physically "walking the store" to witness the service delivery firsthand. Over a one-year period, we analyzed and categorized hundreds of thousands of contacts into strategic buckets to understand why the process was breaking.

The Big Idea: The transformation did not start immediately with AI. It followed a maturity curve supported with additional technical enhancements as it matured:

1. The Basics: First, we ensured service execution simply met the basic promise (e.g., ensuring appointments happened when scheduled and service completed within the right scope).
2. Proactive Anomaly Detection: We introduced new scan signals into the product lifecycle to identify defects before they impacted the customer. We proactively reached out to customers or providers to create a solution prior to the customer being aware.
3. AI-Driven Automation: Once the data was clean, we deployed AI to automate resolutions via bot messaging and proactive notifications (email/text), giving customers self-service options to reschedule, cancel, or provide alternative solutions based on the use case.

Strategy

Timeline: A 5-Year Technology Enhancement Roadmap, with key deliverables executed in priority of impact.

Framework / Model: We utilized a Lean Six Sigma approach to rigorously identify defects and managed the transformation using Agile methodology to iterate quickly on solutions.

Gaming (Incentives & Behavior):

• External (Provider Scorecards): We restructured external service provider scorecards to directly tie financial incentives to customer-centric metrics rather than just volume. Providers were measured on Completion Within Time Window, reschedule rates, star rating and Claim Rates. By linking financial rewards to these compliance and quality metrics, we effectively put the guardrails for the provider network to self-correct and compete for quality service execution.
• Internal (Operational Goals): We established rigorous performance goals for internal teams focused on CS Resolution Speeds, CSAT Improvement, and First Contact Resolution (FCR). To support this, we implemented a constant feedback mechanism that allowed operations to identify roadblocks to CX improvements and prioritized them against other issues for rapid resolution.

Empathy (The Customer View): Empathy was not just a buzzword; it was a measurable metric.

• Measurement: We utilized specific quality surveys sent to customers post-interaction.
• Validation: We conducted human-led QA audits on contact interactions to score the specific "level of empathy" demonstrated by agents.
• Outcome: This data forced stakeholders to "walk the store" digitally, seeing the friction points where customers felt abandoned or CS lacked the tools to resolve the customers issue.

Management

The project was managed through a Cross-Functional Stakeholder Alignment approach supported by Lean Six Sigma rigor:

• Methodology: We applied Lean Six Sigma to analyze defect root causes across the 500K DPMO baseline. Change management was executed using Agile sprints to rapidly deploy fixes (e.g., new scan signals, tools, notifications, planning systems) with constant iteration to improve the CX. .
• Strategic Alignment: We worked across multiple business segments to break down silos by identifying synergies across partner teams.
• Data Governance: We identified critical data gaps that prevented accurate measurement and established new standards for service metrics, moving from anecdotal evidence to hard data.

IMPACT

Quantitative Results:

• Revenue Growth: The services business grew by $250M in revenue.
• Efficiency: Contact rate dropped from 50% to 6% (DPMO reduced from 500,000 to 60,000).
• Cost Avoidance: With volume growing to 4.5M orders in 2025, maintaining the old 50% contact rate would have resulted in ~2.25M contacts. By reducing this to 6% (270K contacts), we avoided ~1.98M contacts.
• Satisfaction: Overall CSAT scores improved by 900 basis points (bps).

Qualitative IMPACT:

• Attitude: We fundamentally shifted the organizational attitude by "connecting the dots" across teams. We ensured every stakeholder understood how their specific defects and actions translated directly into business success or failure.
• Core Values: We embedded Customer Obsession into decision-making, ensuring it was balanced with the reality of our Service Providers. We established that providers are also customers, and decisions must be viable for both parties with Amazon as the supportive intermediary.
• Identity: Transformed the division from a "manual operational cost center" to a "tech-enabled revenue driver."
• Trust: The 900bps improvement in CSAT signifies a restoration of trust in the delivery promise.