SAP + Databricks Possibility: Machine Learning for Store Optimization | by THE BRICK LEARNING | Mar, 2025

In the first two parts of this series, we:

• Defined the store optimization use case
• Integrated and prepared SAP retail data in Databricks

Now it’s time to unlock predictive power with machine learning (ML).

In this article, you’ll learn how to:

1. Cluster stores based on inventory behavior
2. Predict slow-moving SKUs
3. Optimize markdown timing
   All backed by Databricks code snippets, real-world logic, and business outcomes.

Use Case 1: Store Clustering Based on Inventory Turnover

Segment stores based on how they manage inventory — to apply differentiated business strategies (e.g., reorder fast sellers, redistribute slow movers, or automate markdowns).

Features to Engineer

Dataset Structure

Columns: WERKS (Store), turnover_ratio, aged_stock_pct, sales_velocity

PySpark ML Code (KMeans)

from pyspark.ml.feature import VectorAssembler
from pyspark.ml.clustering import KMeans
# Assemble features
vec_assembler = VectorAssembler(
inputCols=["turnover_ratio", "aged_stock_pct", "sales_velocity"],
outputCol="features"
)
vectorized_df = vec_assembler.transform(store_metrics_df)
# Train KMeans
kmeans = KMeans(k=3, seed=42)
model = kmeans.fit(vectorized_df)
# Predict store clusters
result = model.transform(vectorized_df)
result.select("WERKS", "prediction").show()

Use this insight to drive merchandising, ops, and pricing actions.

Use Case 2: Predicting Slow-Moving Inventory