Upgrade to Pro — share decks privately, control downloads, hide ads and more …

Contextually Relevant Retail APIs

Contextually Relevant Retail APIs

by Jason Lobel @ APIStrat 2014 in Chicago

More Decks by API Strategy & Practice Conference

Other Decks in Technology

Transcript

  1. Jason Lobel, CEO @jasonlobel Contextually Relevant Retail APIs for Dynamic

    Insights and Consumer Experiences September 2014
  2. Contextual APIs Activate Insights and Digital Experiences From One Platform

      Critical data sources   Unify data from disparate sources   Enable data to be machine readable  Embed data into digital apps easily  Activate digital personalization efficiently via web, mobile, in-store (beacon), ads and other channels  Visually interpret data  Queries on demand  Predictive applications
  3. Adaptive Intelligence Data > Insights > API > Activation Retailer

    (Data) Unified Data / Algorithm / API Platform Point of Sale Transactions -  Data Storage -  Query Explorer -  Algorithms -  Applications (Alerts, Dashboards, etc) Data Scientists Suppliers Category Captains Product Catalog Internal (API) Media Buying/ Marketing Digital (eCom) & In-Store (BLE, NFC) Locations Promotions Internal CRM (Web/Email) Marketing Assets Suppliers (API) Public (API) 3rd Party Developers Data Scientists / Research Category Managers Media Buying (DSP) Inventory Deliveries
  4. Why APIs? http://apievangelist.com/2012/01/12/the-secret-to-amazons-success-internal-apis/ Mandate for APIs: “Anyone who doesn’t do

    this will be fired. Thank you; have a nice day!” Value of Retail APIs  Contextual Insights  Contextual Experiences  Omni-Channel Agility  Predictive Analytics  Optimize Supply Chain  Partnerships  Open API
  5. What Retail APIs are Relevant? Core Retail  Products  Orders  Prices

     Inventory  Categories  Shopping Cart  Customer History  Loyalty Marketing  Advertising Assets  Promotions  Coupons Company Information   Stores / Locator   Brand Assets   Events Contextual Retail  Item Recommendation  Affinities  Clusters  Item Tags/Facets  Product Reviews  Search Results  Queries (Top Clicked)
  6. ……Day/Week Parting 9  Orders & Stores API > Queries =

    context (user purchases “now” by “location”)
  7. ……Facets/Tags = Semantic Context 10  Products are complex to “describe”

    to a machine  Facets/Tags/Linked Data is mission critical context Source: Jay Myers (BestBuy) www.slideshare.net/jaymmyers/better-business-through-linked-data Clam Chowder   Category: soup, appetizers   Season: winter, fall   Ingredients: Crème, corn, carrot, onions   Pairs: seafood, red wine
  8. Predictive Targeting – Crawl……Walk……Run……Repeat  Most enterprises will start small with

    low sophistication targeting  The degree of individualization can vary significantly Source: Forrester Research 11
  9. Frequent Pattern Mining Product Associations: if X is bought, what

    else is likely to be bought (e.g. men that buy diapers also buy beer) Recommendation Item/Offer Recommendation: suggest products a consumer may like based on known interests Clustering Discover Customer Segments: examine purchasing habits to identify clusters of shopper segments Algorithm Type Application Applying Machine Learning to Extract Insights
  10.   Compute all permutations of behavior (e.g., basket patterns)  

    APIs facilitate three-tier access   REST API = developers   +angular = interface   +angular+d3 = visualization Algorithm API – Pattern Mining FPM Interface Visualization Layer "name":"GENOVA TUNA IN OIL", "itemsets":[ "items":[ "CDF ITALIAN BREAD", "PLNTRS LT SLT MIX” "count":8, "support":0.04, "confidence":100.0 API Output 13
  11.   Grouping “like items” (search terms, items, people, etc).  

    Dynamically, application of clusters as behavioral changes (clicks) occur Algorithm API – Clustering Visualization 14 API Sample
  12.   Recommender algorithms (user, item, anonymous)   Post algorithm Logic

    layer is very important  Add human layer  Suppress bad output Algorithm API – Item Recommendations API Sample User Matrix 15 Jason X X X Jessica X X Kin X X X Steve X X X Sarah
  13. Use Case: Interactive Visualizations 16  API + Open Source (D3)

    = interactive dashboards   Easy to interpret large data sets (~20-40 hours per application)   Enable access to decision makers faster Interactive Dashboards Open Source (D3) Libraries
  14. Use Case: Web, Email, Ad Personalization Apply Algorithms •  Train

    models •  Generate recommendation scores per user •  Output sent to web/mobile site, ESP, etc. Data Logic & Verification •  Ensure correct language •  Ensure copy exists •  Suppress previously-presented items/offers •  Suppress inappropriate items (logic-based) Data Collection •  Data storage •  Reports Hero Image Dynamic Web/Email Templates utilizes Predictive Algorithm to pull in the relevant coupons, upsells, etc Logic to determine title to display Ad Tiles or Custom Messaging Data Import  Purchase Behavior (real-time/next-day)  Web actions, reviews (real-time)  Loyalty (real-time/next-day)  Email History (one-time)  Product catalog (as changing)  CRM/Ad Segments (weekly)  Logic Exclusions (one-time) via API to Front-End Experience
  15. Engage at shelf Welcome content is pushed by Bluetooth b

    e a c o n s a t s t o r e entrance At shelf engagements are delivered through BLE, NFC and QR Beacon pulls contextual content (recipe content, real-time web trends, POS affinities, coupons) Use Case: Mobile In-Store (Beacons, NFC, QR, SMS) Personalization   APIs to deploy content to beacon/NFC partner platforms   Deliver contextually relevant experiences upon entrance or down the aisle   Trending products   Item affinities   Recommendations  Items  Coupons  Offers Source: Thinaire