Skip to main content

Amethyst Technologies

Led development of AI-powered forecasting pipeline that improved prediction accuracy by ~40%

Product ManagementAnalytics

Role

Product Manager

Timeframe

6 months

Team Size

8 engineers, 2 designers

Tools & Technologies

PythonTensorFlowTableauJira

Key Impact

Improved forecasting accuracy by ~40%

Context

Amethyst Technologies needed to improve their demand forecasting capabilities to reduce inventory costs and improve customer satisfaction. The existing manual forecasting process was time-consuming and prone to human error.

Problem

The company was experiencing significant inventory management issues due to inaccurate demand predictions, leading to both stockouts and overstock situations that impacted profitability.

Approach

  • Conducted stakeholder interviews to understand current forecasting pain points

  • Analyzed historical sales data to identify patterns and seasonality trends

  • Collaborated with data science team to design machine learning pipeline

  • Implemented A/B testing framework to validate model performance

  • Created dashboard for real-time forecast monitoring and adjustments

Outcome

  • Achieved ~40% improvement in forecasting accuracy across all product categories

  • Reduced inventory holding costs by 25% through better demand prediction

  • Decreased stockout incidents by 60% leading to improved customer satisfaction

  • Established automated forecasting pipeline reducing manual effort by 80%

Next Steps

  • Expand forecasting model to include external market factors

  • Implement real-time inventory optimization recommendations

  • Develop mobile dashboard for field sales team access

Project Gallery