←
Go back to case studies
Tyler Nagano
Data Engineer Consultant
February 15, 2024
By the Compass Team
Services
Data Science
Analytics & BI
industry
Consumer Packaged Goods (CPG)
Stack
Alteryx
Tableau
We leveraged a low-code, no-code solution in Alteryx to create workflows that normalized all data inputs into a defined data model, performed feature engineering, and applied a clustering model to identify similar stores and assigned product assortment. The model was executed using Alteryx Analytical Apps, providing a simple point-and-click user interface so that the client did not require coding or technical knowledge. This intuitive solution allowed business users to build clusters and adjust their desired scope for each market.
The Category Management team at a multinational beverage company sought to optimize their product mix across their stores using machine learning. The client needed our firm’s data science and data engineering expertise to build a robust solution and interpret the results.
The multinational beverage firm faced inefficiencies in providing the right product assortment to stores due to their rapid growth. Previously, they relied on expert intuition to set arbitrary thresholds for product assortment, leading to overstock in some stores and understock in others, resulting in lost market share, cost overages, and revenue loss.
To improve product assortment nationally, they sought a powerful yet explainable modeling approach that non-technical stakeholders could understand. The model aimed to determine the optimal product mix for each store based on a myriad of inputs.
Our client needed to identify stores with similar product distribution patterns. Earlier, segmentation was based on arbitrary thresholds, lacking data backing and relying on expert intuition, which was unsustainable if key personnel left.
Data across market segments had various sources and column structures. A standardized data model was essential for consistent product placement decisions.
With limited technical expertise, the client required a non-code solution for easy maintenance and rerunning. The solution needed a user interface for data cleaning, running the machine learning model, and visualizing results.
The Compass Analytics team leveraged Alteryx and Tableau. Alteryx was used to build workflows that were turned into Analytical Apps to create a point-and-click interface for data standardization and to run the clustering model using scikit-learn. Tableau visualized the results and key metrics for intuitive interpretation.
For data cleaning, we built an Alteryx workflow to normalize all data inputs into our defined data model and an Analytical App, a user interface allowing business users to select data sources and common columns, ensuring a standardized format for clustering. Validation steps confirmed the correct columns were included in the final dataset. The workflow allowed a non-technical user to run the data standardization process.
For clustering, the user interface enabled business users to select parameters such as store scope and number of clusters. The Python Tool in Alteryx implemented the clustering model which uses scikit-learn. Key metrics were converted into a diagnostic report, providing insights into cluster quality and product characteristics. Instructions were provided to business users on how to interpret the results.
The model's output was exported to Tableau, highlighting key metrics and visualizing the geographic distribution of store clusters.
The solution developed by Compass Analytics has enabled the beverage company to implement a robust product segmentation model, facilitating data-driven decisions and improving product placement across stores.
Built primarily in Alteryx, the solution offers a user interface for data standardization and clustering model parameter setting. This ensures the company can create business value from product assortment and allocation without needing technical coding expertise.
The multinational beverage company now has a standardized process for clustering and optimal product allocation, enabling Category Managers to continue setting optimal product distributions across stores, securing market share and sustained revenue.
Our standardized data cleaning process ensured consistent data ingestion across various market segments, facilitating uniform clustering and insightful results across regions.
The solution groups stores with similar product distribution and provides easy-to-understand visuals, helping the company quickly comprehend their product segments. This data-driven approach allows Category Managers to optimize product assortment, reduce lost institutional knowledge, and maintains market share even if key team members depart.
John Molson Supply Chain and
Business Technology Association
.png)