← All case studies
Data Public-Demo 2024

AutoML Coffee Price Forecasting

A decision-support data product for coffee price forecasting using Python, Streamlit, and PostgreSQL with a management-facing dashboard.

RoleProduct Direction · Developer
ClientCoffee trading & processing
StackPython · Streamlit · PostgreSQL · AutoML

01 Problem

Business Context

A coffee trading and processing environment needing decision support from historical commodity data and market-related indicators.

Why it mattered

Management needed a more structured way to analyze historical coffee price data and generate forecast direction for short-term decisions.

Problem Statement

Price analysis was manual and ad-hoc, with no repeatable way to turn historical data into forward-looking direction.

02 Solution

An upload-to-forecast pipeline: tabular ingestion → PostgreSQL → forecasting model → Streamlit dashboard for decision support.

Objectives

  • Provide an upload flow for CSV/Excel historical data
  • Store data in a database-backed model
  • Define a forecasting output direction
  • Create a simple management-facing interface

Architecture / Topology

CSV/Excel Upload PostgreSQL Forecast Model Streamlit UI
CSV/Excel Upload ingest PostgreSQL
PostgreSQL train Forecast Model
Forecast Model forecast Streamlit UI
AutoML Coffee Price Forecasting architecture diagram

Implementation Agile / Sprint

  1. Sprint 1Done1 week

    Product Direction

    Frame the decision the forecast must support.

    • Product brief
    • Data flow sketch
  2. Sprint 2Done2 weeks

    Data & DB Design

    Define ingestion and storage.

    • Schema
    • Upload flow
  3. Sprint 3Done2 weeks

    Forecast Prototype

    Produce a working forecast direction.

    • Prototype
    • Streamlit UI

Selected Visuals

03 Role

Product Direction · Developer

Stakeholders

ManagementTrading deskData / IT

04 Tech / Tools

PythonStreamlitPostgreSQLAutoML

Key Components

  • CSV/Excel upload
  • PostgreSQL store
  • AutoML / forecasting
  • Streamlit dashboard

05 Business Impact

Turned scattered spreadsheets into a repeatable forecast workflow management could open and read.

Reduced manual reporting dependency for price analysis.

Improved decision-support visibility for management.

Created a repeatable forecasting workflow.

06 Evidence Policy

Public-Demo

Public demo only with dummy data. Internal commodity datasets are not published.

No credentials, server IPs, internal production URLs, financial figures, or confidential documents are shown. Detailed artifacts are available upon request under NDA.