← Back to work
Experience
pythondjangosqlbashanalyticsteam-buildingautomation

Scaling a Global Analytics Function from One to Fourteen

When I joined Marin Software in 2013, the analytics function was one person: Will Hartley-Booth, a brilliant analyst who knew our bidding algorithms inside out but was moving to professional services. I inherited his responsibilities and an empty toolbox — no platforms, no automation, no internal tooling.

The first problem I spotted: people were spending 8–10 hours producing quarterly business reviews from scratch — screenshots, manual data pulls, a blank page every time. I taught myself bash scripting and wrote automated SQL queries that pulled data into structured Excel templates. It solved the cold-start problem and gave people something to build on rather than starting from nothing.

That initial automation grew into what we called the Analytics App — a Python/Django platform that started as two or three internal tools and expanded rapidly as the wider organisation saw what was possible. We built user journey reports for complex multi-touch attribution paths, keyword recommendation engines, and account auditing tools that provided recommendations on structure, spend efficiency and QA coverage. The 12,000+ reports per year started as on-demand internal pulls and eventually became client-facing quarterly business reviews with polished PDF output.

Over roughly two and a half years I grew the team to fourteen across six offices. Phil in London was my first hire and helped us scale fast. Luke and Simonas joined as developers. Afolabi as an analyst. Mats started in Shanghai — I later supported his relocation to Hamburg. Kate in Austin, Sophia, Michael and Luca in San Francisco, Mihir in Chicago, Wil in Austin, Rob in New York. The automation saved 150+ hours per week and roughly $1.8M annually. The auditing tool delivered an average 26% performance uplift across the customer base.

The hardest part wasn’t technical — it was organisational. We sat between customer success and engineering but belonged to neither. Leadership recognised our impact but never quite resolved where we fit in the reporting structure. That ambiguity taught me more about building influence without positional authority than any leadership course could.