Minimum Version Required: 22.01
Steps
Creation and Definition of Peer Groups and Peer Products
This document outlines the technical process for creating and defining Peer Groups and Peer Products.
The process involves several steps, including initial grouping, iterative refinement, and final assignment of peer groups and products.
Peer Group Calculation Process
Peer groups in our system are designed to group similar customers together based on their characteristics and purchasing behaviors. Here's how the algorithm works:
Initial Classification
Basic Premise Types: Customers are initially divided into three basic groups:
On-Premise Draft (bars/restaurants with draft)
On-Premise Package (bars/restaurants with bottles/cans only)
Off-Premise (retailers)
Parameter Calculation
The system calculates two important parameters:
Minimum Peer Group Size: 0.5% of total active customers (minimum of 10)
Product Sample Size: 20% of the average ranking of top products
Iterative Product-Based Clustering
The system then runs through multiple refinement cycles:
1. Product Purchase Analysis:
Examines purchase patterns of top products across customers
Identifies products that are commonly purchased by subsets of customers
2. Group Formation:
Creates new peer groups when it finds a significant number of customers who purchase the same products
Ensures each new group has at least the minimum required customers
Moves customers from their original group to the new product-based group
3. Multiple Iterations:
Performs up to 3 cycles with increasingly larger product sample sizes
Within each cycle, runs up to 100 iterations to find optimal product-based groupings
Stops when no more significant product-based clusters can be found
Geographic and Customer Type Grouping
After product-based grouping, the system creates additional peer groups based on:
Customer Type and Area
Customer Type and Location
Customer Type alone
Premise Type and Location (if multiple locations exist)
Naming and Metadata
Peer group names are generated based on distinguishing characteristics:
Premise type (On/Off Premise, Draft/Package)
Product names when there's a single dominant product
Brand, brand family, supplier, package type, or segment information
Geographic and customer-type attributes
Sales Metrics Calculation
Finally, the system calculates key metrics for each peer group's products:
Sales per customer
Average product rank within the group
Buyer percentage
Result
The end result is a set of peer groups where:
Each customer belongs to exactly one peer group
Each peer group contains similar customers based on premise type, product preferences, and geographic/customer type attributes
Each peer group has at least the minimum required number of customers
Each peer group has associated products that are relevant to that customer segment
This multi-layered approach allows for meaningful comparisons between similar customers and helps identify product selling opportunities based on what similar customers are purchasing.
Fields And Definitions
What are Peer Groups?
Peer Groups are used to group like customers together and provide a list of suggested products.
How are they grouped?
Peer Groups are first grouped by on/off premise and then looking at the product mix of the group of customers. It builds the most specific list of customers that have similar purchasing behavior of products and sales of those products.
When are they used?
Peer Groups are primarily used for recommended order processes throughout the system, including dsdlink, standard Invoice Edit, and AI Recommended Placements.