Company Overview
- US-based industrial parts retailer
- $16B revenue
Tech Overview
- 260+ Retail outlets
- 3M+ SKUs
- 70M+ Searches per day
Business & Technical Challenges
- Search drove 60% of overall website revenue – and was underperforming significantly by internal benchmarks
- Off-the-shelf 3rd-party search systems (Endeca) failed to surface relevant results and hurt revenue
- On-prem search infrastructure was unreliable, costly and inflexible, and struggled to handle peak demand
- Legacy tech driving nightly data feeds frequently failed, resulting in suboptimal search performance
- Unplanned outages and increasing P1 issues led to revenue loss and made planning and forecasting unreliable
- Comprehensive redesign of Search, using Elasticsearch and custom AI models built on AWS Cloud
- Event Driven Architecture using AWS SNS and managed Kafka enabled real-time data updates (no nightly data loads) and improved relevancy
- Real-time Elasticsearch index updates, and moving key functionality to AWS Lambda serverless computing improved search performance
- Built AI models that recognize industrial parts, and launched using AWS SageMaker-driven microservices to improve specific types of queries
- Automatic scaling using containerization / Kubernetes and AWS instances enabled efficient scaling to meet peak demand
- Splunk, AWS Firehose and Twilio were used to capture logs, analyze and notify in real-time
- Relevant, personalized search results drive greater revenue and customer satisfaction
- Faster provisioning of systems, from days to hours, accelerates innovation
- Higher flexibility, agility and resilience provided by theAWS cloud vs. on-prem
- 12% improvement in Search-to-Cart
- 70% reduction in tech-debt resulting in increased focus on innovation
- Streamlined operations
- Faster time to market
- Improved data quality
- Enhanced data governance
