Talk to an Expert
Back to Cases
FMCG

Tea Price Forecasting and Procurement Insights

Ensemble machine learning–based tea price forecasting with external weather signals to improve procurement timing and pricing strategy.

≥90% price forecast accuracy

Improvement

The Problem

Volatile tea prices driven by seasonality, weather, and demand–supply shifts made it hard to forecast accurately, affecting procurement timing and pricing strategy.

Key Challenges

  • Price Volatility due to climatic conditions and supply–demand changes.
  • Procurement Timing: identifying optimal bulk-buy windows.
  • Pricing Strategy: aligning with market conditions to protect competitiveness and profitability.
  • Demand Impact: understanding brand-level demand shifts from tea price changes.

Our Solution

Built an ensemble ML pipeline using 5 years of daily tea data plus AccuWeather signals to forecast prices 12 months ahead and derive demand-uplift rules.

  • Ensemble ML
  • Temporal Feature Engineering
  • AccuWeather Data

Implementation Approach

  1. Data Prep: 5 years of historical daily tea; integrated AccuWeather data; engineered temporal features.
  2. Forecasting: Ensemble ML model to predict tea prices 12 months ahead with quarterly refresh; validated via cross-validation and blackout periods, achieving ≥90% accuracy one quarter out.
  3. Demand Uplift Patterns: Uplift/downlift rule sets layered on top of demand forecasts.

Results & Impact

  • High Accuracy: identified optimal bulk purchase periods to reduce procurement costs.
  • Enhanced Pricing Strategy: proactive price adjustments to maintain competitiveness and margins.
  • Demand Forecast: better projections by incorporating commodity-price-driven uplift.