Retail-Forecasting-Talk

Retail Trade Forecasting

View the Live Presentation

AMLC Community Talk — March 2026

Scaling insights with an interpretable, drift-resistant solution of a dynamic system.

This repo contains the slide deck, Jupyter notebooks, and Python source code for a talk on structural econometric forecasting applied to retail and payments data.

Run the Notebooks

Run in Colab

These notebooks are configured to run in Google Colab with zero local setup.

Analysis Module Focus Link
Retail Cyclicality STL decomposition, cross-correlation, and crisis comparison. Cyclicality Analysis
Forecasting Model Structural regression modeling and macroeconomic indexing. Forecasting Model

[!TIP] Using FRED in Colab: When the notebook opens, you will be prompted to enter your FRED API Key to fetch live data. The notebooks include pre-computed outputs if you prefer to just browse.

Run locally

uv run jupyter lab

Open notebooks/Retail_Cyclicality.ipynb or notebooks/Retail_Forecast.ipynb. Both notebooks ship with outputs so you can read them without running, but a FRED API key is required to re-execute cells.

Run Tests

uv run pytest -v                   # All tests (requires FRED_API_KEY in .env)
uv run pytest -v -k "not Live"    # Unit tests only (no network)

💬 Q&A and Discussion

If you’re attending the AMLC talk and have questions about the structural econometric approach or the statsmodels implementation, please open an issue or reach out during the networking session!

Talk Overview

  1. The Problem — sampling bias and simultaneity in operational data
  2. The Structural Solution — decomposing revenue through macro → industry → operational → financial layers
  3. Building It Right — functional form, indexation, and regression rigor
  4. Modeling Dynamics — stationarity, co-integration, and AR models
  5. Using the Forecast — uncertainty, scenarios, and stakeholder communication
  6. Maintaining the Model — tracking drift and maintaining accuracy

What’s Inside

Path Description
decks/build/retail-forecast.html Pre-built slide deck — open in a browser and present
decks/retail-forecast.md Slide deck source (Marp format)
notebooks/Retail_Cyclicality.ipynb Cyclicality analysis: STL decomposition, cross-correlation, crisis comparison
notebooks/Retail_Forecast.ipynb Forecasting model: regression with macroeconomic indices
src/data_loader.py FRED API data loader (FredMacroRetailLoader)
src/regression_dclass/ Versioned regression module (v0 → v1 → v2 progression)
tests/ 26 tests covering the data loader (unit + live FRED API)

Quick Start

Prerequisites

Setup

git clone https://github.com/hb-cam/Retail-Forecasting-Talk.git
cd Retail-Forecasting-Talk

# Install Python dependencies
uv sync

# Configure your FRED API key
cp .env.example .env
# Edit .env and add your key

View the Presentation

The easiest way — just open the pre-built HTML:

open decks/build/retail-forecast.html    # macOS
xdg-open decks/build/retail-forecast.html  # Linux

Press F11 for fullscreen. All diagrams and assets are self-contained.

To run the dev server (requires Node.js):

npm install
npm run deck:serve
# Opens at http://localhost:3000

Tech Stack

🤝 Contributing

We welcome contributions from the community! Whether you are fixing a bug in the regression module or adding a new analysis notebook, please follow these steps to ensure a smooth workflow.

1. Prerequisites

This project uses uv for Python package and project management. If you don’t have it yet, install it via:

curl -LsSf [https://astral.sh/uv/install.sh](https://astral.sh/uv/install.sh) | sh

2. Development Workflow

  1. Fork & Clone: Fork the repository and clone it to your local machine.
  2. Environment Setup: Run uv sync to create a virtual environment and install all dependencies (including dev dependencies like pytest). ```bash uv sync

3. **Create a Branch**:
```bash
git checkout -b feature/your-feature-name
  1. Data Access: Ensure your .env file contains your FRED_API_KEY.
  2. Coding Standards: We use v0 -> v1 -> v2 versioning in the src/regression_dclass to show model evolution. If updating the model, please maintain this progression.

3. Testing

Before submitting a Pull Request, please ensure all tests pass. We use pytest for both unit and live API integration tests.

# Run all tests (requires API key)
uv run pytest

# Run only unit tests (no network required)
uv run pytest -k "not Live"

4. Submitting Changes

License This project is licensed under the MIT License - see the LICENSE file for details.

📬 Contact & Connect

If you’re interested in structural econometrics, property technology, or modern data engineering workflows, I’d love to connect.

[!TIP] Questions about the talk? > Feel free to open an issue or reach out via LinkedIn. I’m especially happy to discuss the v0 → v2 model progression or using uv for reproducible data science.

This site is open source. Improve this page.