Not dashboards for their own sake. Each is a case where the analysis produced a structural outcome the organisation committed to.
The hardest part is rarely the model. It is turning a vague, strategic question, the kind where the decision is expensive to get wrong, into one that can actually be answered. I take the messy version (“where should we be placing our bets?”) and return the specific evidence the decision turns on.
Forecasting models, propensity scoring, A/B testing: chosen to answer the question, not to show their working. Each lands on a clear recommendation with its uncertainty stated, so you know not just what to do but how confident to be.
Analysis only matters if it survives me leaving. I build the pipeline, document the methodology, and translate the work into a repeatable operating model the internal team runs independently.
On decision quality and the gap between a correct analysis and a good decision. Published on Substack.
What B2B decisions need beyond the p-value. Moves from statistical significance to what actually decides: expected value, downside risk, and reversibility.
Extends the framework to system-level uncertainty using Monte Carlo, with worked examples on uncertainty propagation and input sensitivity.
A Bayesian comparison of two conversion rates, the right way to ask “which campaign won?” when a higher rate might just be luck. Computed exactly, in the browser. A demonstration of method you can run on your own numbers.