The Adaptive Commerce Experience With Machine Learning APIs

By Jason Lobel @ API Strategy & Practice Conference
San Francisco, October 23-24-25, 2013

Transcript

1. The Adaptive Commerce Experience With Machine Learning APIs Jason Lobel CEO @jasonlobel
2. About Me
3. About Us Swift Access   Data backend-as-a-service   Import, store, access and analyze data on-demand   Power digital experiences and intelligence Swift Predictions   Machine-learning-as-a-service   Harness a catalog of adaptive algorithms   Embed predictive APIs to make your apps smarter
4. We Now Live In The Age of The Consumer
5. What Do Consumers Want?
6. Why Do Marketers Generally Fail to Satisfy Us? Disparate Data Silos Conflict in Team Goals Lack of Machine Readable Data (APIs)
7. Why Machine Learning? {Mathematical} Relevance
8. <2000 2000’s 2011 2012 2013 & Beyond Shopping Touchpoint In-store Online Mobile Social Visual discovery Tablets & Mobile POS 50 Billion Internet- Connected Devices Commerce Is Complex and Fast Changing
9. Internet-Connected Devices are Exploding 50B 12.5B
10. Individualization vs. Technology Sophistication
11. Why ML? Optimize Relevance Continuously Analyze Massive Data Volume, Velocity, Variety Models are Not Static – User Purchase Habits Change Over Time Testing! Variables (e.g., Weather) Affect Businesses Differently Infrastructure for Large-Scale Distributed Computing is Available Open Source and Paid ML Algorithms Are Available Software Can Solve These Challenges Machine Learning Benefits Why Now
12. Why APIs for Machine Learning Algorithms? Hard Eas{ier} Human   Finding a data scientist Technical   Database selection   Algorithm(s) selection   Model training & iteration   Embedding predictions into applications   Security   Query speed / caching   Scaling   On-Demand Access Human   Finding an engineer that can use an API   Training (if needed) Technical
13. Why APIs? Omni-Channel Consistency
14. Why APIs? Interactive Big Data Visualization D3.js (d3js.org)   JavaScript library for manipulating documents using HTML, SVG and CSS
15. Useful Machine Learning Concepts Recommendation:   Analyzes users' preferences and finds items users might like Frequent Pattern Mining:   Discovers unique frequently co-occurring items in a transaction list Classification:   Learns from existing categorized data and assigns a category to uncategorized data Clustering:   Organizes items from a large volume of data into groups of similar items and features
16. Common ML Applications for Commerce   Item Recommendation: observes what the user likes and finds similar items (“I like the Chicago Bulls, I may like the Chicago Bears”)   User Recommendation: recommend items finding similar users and sees what they like (e.g., Kin and I are friends. He likes IPAs. I may like IPAs)   Product Affinity: if X user wants X, what else is Y user likely to want based on the relationship between X and Y (men who buy diapers, also tend to buy beer)   Predict Inventory: based on history, predict future sales (next 7, 30 days, etc.)   Discover Customer Segments: examine purchasing habits to identify clusters of shopper segments   Prevent Fraud: identify anomalies in cashier activity, such as voids (is this likely fraud? yes/no)
17. Where Do I Collect Data?   Web: JavaScript Tag   Product data: export catalog, affiliate feed, etc   CRM: export   Segmentation: third-party sources (e.g., Acxiom) Web Tag CRM Web Tag / Catalog Segmentation
18. Item Recommendation Generate Model {Inputs} 1.  Gather data   Order history (user ID, past purchases)   Web behavior (page views, adds to cart)   Segmentation (age, income)   Search path (keywords, items bought) 2.  Run algorithm 3.  Query User ID Response   Returns a score for items that are deemed relevant based on item/score pair
19. Recommendations In Use (Amazon)   Amazon’s product pages includes many behavioral-based suggestions
20. No Single Algorithm is a Magic Bullet ? 1 2 3 4 5 6 7 8
21. Frequent Pattern Mining Model Inputs   Order history (POS data)   1 = Yes   0 = No   Understand affinity between milk, bread, butter & beer   Understand affinity between products (e.g., grocery localization)
22. Frequent Pattern Mining API Response Pattern Count   [milk, bread],1   [butter],1   [beer],1   [milk, bread, butter],1   [milk, bread, beer],1   [milk, bread],1 Frequency List :[(bread,4), (milk,4), (beer,2), (butter,2)]   Number of unique items 4 Patterns found with… •  Butter: 2 Patterns, butter:[butter]:2 [bread,butter,milk]:1   Bread: 3 Patterns, bread:[bread,milk]:4 [bread,butter,milk]:1 [beer,bread,milk]:1
23. Classification   Predict inventory (continuous variables)   Match search term to a SKU (not continuous) Model Training {Inputs}   Inventory: date, purchase history, other (offers)   Evaluation: # instances, mean squared error   Query a SKU   Search: keywords, adds to cart, orders   Evaluation: # instances, accuracy score (0.80)   Query a keyword or feature (e.g., area rugs) Response 6.686672
24. Clustering: Discovering Shopper Segments Inputs   Order history(average basket, purchases)   Set # of clusters?   Set # of iterations? API Response   Links to a file with evaluation tools (next page)
25. API Responses Can Get Messy
26. Clustering: Raw Evaluation Tools
27. Clustering: Visualize to Understand Quickly
28. How Would a Machine Classify These Guys? Differences Similarities   Visual appearance: Casual   Job: Entrepreneur   IP Address: San Francisco   Buys: T-Shirts & Jeans   Name   Male   Frequent traveler   Love APIs   Love IPAs   Browses: tech sites   Visual appearance: Formal   Job: Public employee   IP Address: Washington DC   Buys: Suits & Tie   Browses: Gov websites
29. Compete With Adaptive Intelligence Commerce must be adaptive Commerce must be relevant always in the now Adaptive Intelligence v1.0 = machine readable data (APIs) / collaboration v2.0 = automate decisions with machine learning