Geospatial & Customer Data Enrichment via Public Statistical Sources

Stage 1. Hypothesis & Tiering Framework Design

Initiated the project with a core hypothesis: not all geographic areas generate equal value, and performance heterogeneity can be systematically identified and leveraged.

Developed a structured tiering framework to segment regions based on underlying economic and performance characteristics. Drafted key contributing components, including earnings proxies, historical performance data, revenue contribution, and volume metrics.

To ensure comparability across regions of varying scale, normalised variables against revenue contribution and transaction volume to prevent large geographies from distorting the analysis.

Initially implemented K-means clustering to identify natural groupings, but after evaluating interpretability and commercial applicability, transitioned to a structured deterministic tiering approach better aligned with decision-making.

Stage 2. External Enrichment & Probabilistic Optimisation

Enhanced the internally aggregated county-level model through structured enrichment using publicly available Office for National Statistics (ONS) data and anonymised customer-level aggregates.

Integrated socio-economic and demographic variables to refine regional performance expectations and reduce allocation bias.

Applied Bayesian statistical modelling to better estimate regional performance uncertainty and inform budget allocation under probabilistic constraints, enabling a more robust and forward-looking geo strategy.