stitchfix Blog Post - This post provides Data Strategy for Data Science

This post  provides insights into Data Science Strategy in stitchfix

Problem Solving Approach (Use Cases - Data - Models)

• Step #1 - Business Use Cases -> Finding Relevant Data -> Providing Data with ETL 
• Step #2 - Data - Multiple Models
• Step #3 - API to consume results and use data for decision making

Key Lessons

• Availability of Raw Data
• Building ETL for data updates
• Data Pipelines for Feature Engineering
• Different Data Science Algos for Algorithms
• Data Science uses cases driven from the business context

Data Demands

• Raw Data Access (Pull Everything to a Data lake)
• Data updates / Deletes (Data lake updates with events)
• Feature variables (Custom ETL to select, transform data from raw data)

Experimentation

• A / B Testing
• Validating with real-time results
• Ongoing correction of models

Connecting Data and Science

• Overlapping functions with Domain, Data and Data Science Knowledge
• A lot of Experimentation

Algorithms (Data Science use cases)

• Style Recommenders (Recombining Attributes from existing styles adding feedback), Developing Design with a certain set of attributes
• Warehouse Assignment (Shipping cost, shipping time, inventory match)
• Inventory Forecast (Demand, Unit Price, Total Cost, Ordering Cost, Carrying cost, Season, Recently emailed etc)
• Fashion Design Algorithms
• Buying Algorithms
• Engagement Algorithms
• Messaging Algorithms
• Capacity Optimization
• Assignment Optimization
• Network Optimization
• Visitor Qual Algorithms
• Latent Size Algorithms
• Latent Fit Algorithms
• Batch Picking Algorithm
• Global Optimizations
• Pick Path Algorithm
• Virtual Warehouses
• Sizebreak Algorithms
• Planning Algorithms
• Assortment Algorithms
• Replenishment Algorithms

Use Case Categorization

• Customer Context - Style Recommenders, Fashion Design Algorithms, Latent Size Algorithms, Latent Fit Algorithms
• Retailer Context - Business Use Cases (Inventory Forecast, Replenishment Algorithms)
• Warehouses Use Cases - Assignment Optimization, Allocation 
• Clients Use Cases - Style recommendations, Demand Predictions
• Optimize Supply Chain - Warehouse Assignment, Pick Path Algorithm

Data Science - Algorithm Demands

• Assortment Algorithms - Apriori / Market Basket Analysis
• Targeting Algorithms - Recommendations 
• Replenishment Algorithms - Forecasting
• Allocation Algorithms - Resource Allocation
• Virtualized Warehouses - Demand Forecasting

Key Lessons

• Data Science Use cases in Retail Space
• Data Science Use cases in Supply Chain
• Data Science Use cases in Fashion, Ecommerce Segments
• Data Lake Strategy for Data Science
• Bird's Eye view for picking right use cases

Keep Thinking!!!