← All case studies ◆ Commercial analytics · Statistical modelling

Hierarchical sales analysis across key markets

Umbizo built hierarchical models to capture how multiple variables interact to affect product sales across markets, quantifying both direct and indirect effects.

Illustration of overlapping distribution curves, one per market, with one market highlighted

The challenge

The engagement involved a complex analysis of how multiple variables interact to affect product sales across markets. The goal was to build models that capture both direct and indirect effects through hierarchical analysis, turning complex relationships into usable insight.

What we set out to do

  • Develop hierarchical models for sales analysis
  • Identify the variables that contribute most
  • Optimise model fit and performance
  • Quantify interactions between variables
  • Turn complex relationships into actionable insights

Our approach

Variable analysis

Carried out principal component analysis (PCA), correlation matrix evaluation, selection of optimal variables, and identification of key interaction effects.

Model development

Built hierarchical structural models, integrated seasonal components, incorporated regional variations, and validated model assumptions.

Performance analysis

Benchmarked variable contributions, assessed latent variable impacts, measured model fit metrics, and evaluated prediction accuracy.

Results and impact

The hierarchical models achieved a strong correlation of 0.89 in sales prediction, identified significant seasonal effects, mapped regional variation patterns, and quantified key variable interactions. This gave the client a better understanding of market dynamics, improved prediction capabilities, and actionable market insights.

“The hierarchical modelling approach revealed crucial insights about our market dynamics.”

Olivia Lau, Lead Market Analyst

Timeline

  • Initial analysis: 2 weeks
  • Model development: 1 month
  • Validation: 2 weeks
  • Final implementation: 1 week

Looking forward

Expanding the model to more market segments, developing more sophisticated interaction analyses for emerging markets, and continuing to improve prediction accuracy through machine learning integration.