// Case study / Experimentation culture
Introducing A/B testing as a leadership practice.
Five years building marketing analytics from scratch at a 280-location franchise, including the warehouse foundation, identity model, and experimentation culture that came with it.
What happened
A marketing analytics function and testing culture were built from the ground up.
At Stanley Steemer, a 280-franchise-location home services brand, I helped build the marketing analytics function from scratch on an Azure SQL Server foundation, including the single-customer / householding data model that gave the business its first reliable identity resolution across the multi-location estate.
I introduced A/B testing as a leadership decision-making practice, not a marketing tactic. By year two, leadership was actively requesting evidence before decisions; the culture had shifted from intuition-first toward test-first. The original A/B framework was still in use when I left.
Context
The business had no shared customer definition.
Stanley Steemer had no experimentation culture, no marketing analytics function, and a legacy BI stack that needed replacement. Across 280 franchise locations, there was no reliable identity resolution across the estate, which meant no strong foundation for marketing automation, CRM strategy, or consistent customer-level reporting.
Task
Build the architecture and the decision practice together.
The work was not a tool migration. It was a cultural and architectural rebuild: put analytical data in the warehouse, define customer identity, make reporting usable, and introduce experimentation as decision support.
FoundationAzure SQL
Universal Analytics API ingestion in Python, then GA4 beta export through BigQuery back into Azure.
IdentityCustomer model
Single-customer concept and householding model created reliable identity resolution across locations.
PracticeA/B testing
Experimentation framed as decision support for leadership, not isolated marketing optimisation.
Built on Azure SQL Server with SSMS for development and SSRS for reporting.
Wrote the Python API calls that pulled Universal Analytics data into the Azure warehouse.
When GA4 beta became available, moved the company onto GA4 export through BigQuery and connected that stream back into Azure.
Introduced A/B testing as leadership decision support and stewarded the framework through year-two adoption.
Action
The work linked data architecture to executive decision-making.
I designed the single-customer concept and householding model, wrote the Python API ingestion that brought Universal Analytics data into Azure, later moved the company onto GA4 beta with BigQuery export back into the warehouse, and introduced A/B testing as a way for leadership to reduce decision risk. The point was not only to run tests; it was to make evidence a normal part of deciding.
Outcome
The framework outlived the build phase.
- Marketing analytics function built from scratch over five years.
- 280 franchise locations supported by the framework.
- Single-customer / householding model enabled reliable identity resolution across the estate.
- By year two, leadership was actively requesting evidence before decisions.
- Original A/B framework was still in use when I left.
Design lesson
Experimentation is cultural infrastructure.
A/B testing is often described as optimisation. In this case, it functioned as governance: a repeatable way for leadership to ask better questions, define evidence before the decision, and avoid treating intuition as the only available input.