Seams Boutique

Lifestyle Retailer

Seasonality was difficult to manage for this small retailer
Tools Used: Apache Airflow, AWS, Tableau

The Problem

Seams Boutique, a local California retail store with three locations, specializing in outdoor apparel and gear. Facing seasonal demand swings—especially around summer and holiday tourism—the company struggled with inventory shortages during peak months and overstock in off-seasons. Leadership sought a data-driven solution to forecast next month’s sales more accurately, reduce stockouts, and optimize purchasing.

CLIENT

Seams Boutique

YEAR

2022

TECH STACK

Python, AWS, Airflow, Tableau

The Solution

Our team worked with key stakeholders to:

Data Ingestion and enrichment

The team worked to ingest POS transactions nightly into an S3 bucket. We also loaded weather data from the NOAA API and Events data from a company spreadsheet

Feature Engineering

Using various python scripts, we were able to generate monthly data aggregates and develop features such as rolling statistics, lagged sales, and weather indicators for machine learning

Modeling

We chose to fit a Random Forest Regressor model due to its performance with non-linear patterns. Using ML best practices, we found a forecast that produced 15% improvements over the manual forecasting done by the team.

Deployment and Automation

Once the model was fit, we loaded the model into S3 with joblib and created an API service in AWS. CI/CD was handled on a monthly trigger, and forecasts are shown to the purchasing team as a Tableau Dashboard