Data Sheets

Models Overview

Issue link: https://resources.custora.com/i/1079913

Contents of this Issue

Navigation

Page 0 of 0

Q U I C K G U I D E Predictive Modeling + Segmentation Applications Create rich customer profiles Core Predictive Segmentation Customer Lifetime Value Build segments for communications Anticipate behavior + actions Stated simply, predictive modeling (synonymous with "machine learning") uses statistics to predict outcomes. Custora is a customer intelligence platform built upon a foundation of using predictive modeling for advanced segmentation. Our ever-expanding suite of statistical models and segmentation techniques are designed to better understand and anticipate individual shopper behavior. Custora collects disparate customer data (transactions, CRM, web analytics, etc.), transforms this data into user profiles, and enriches each profile with predictive insights. These insights can then be accessed through our web application to push audiences into marketing channels like email and Facebook, or exported as a data feed via our Switchboard product. Lifecycle Status Persona Pricing + Discount
 Preference Product A!nity Email Engagement CLV: $289 Status: Active Persona "Sports"
 Preferences: Outputs Statistical Model Description A customer's predicted spend (revenue/profit), over a variety of time horizons, inclusive of all lifetime behavior or purely forward-looking. Predicted lifetime value Predicted order rate Predicted order size Pareto-NBD Predicted lifecycle stage Churn propensity Rate at which a customer is predicted to place future orders (e.g., 1.2/year) Predicted average value (revenue/profit) of a customer's future orders (e.g., $212) Preferred price point Preferred list price Discount sensitivity A customer's predicted optimal price point rating (high, medium, low) based on effective price paid Predicted email engagement freq. A customer's predicted optimal price point (based on MSRP or "list price") A customer's predicted discount preference based on spending behavior A customer's predicted email engagement frequency while he or she is actively engaged Predicted preferences for specific products / categories User-level preferences for specific products or categories. Ideal for product-driven messaging (e.g., "Denim Clearance") Five distinct groups of shoppers that tend to purchase the same types of items. Ideal for versioning / content curation. Optimize promotions At-Risk of Churn Dirichlet Multinomial Mixture K-Means Clustering Predicted product / category clusters Determine the likelihood that a customer is active/inactive at any given point in time based on individual purchase patterns Identify a customer's likelihood of making a transaction in the future ("cooling off", "at-risk of churn") Empirical Bayes Alternating Least Squares 20% ! Predicted a!nity for denim Add'l details re: statistical methodologies available in "Models Deep Dive" Presentation Expected future interactions w/ email (Active / Inactive) Predicted email engagement status Pareto-NBD

Articles in this issue

view archives of Data Sheets - Models Overview