Too many growth decisions rest on numbers that cannot tell cause from coincidence. I help teams find out what really drives their results and put their budget behind it.
Attribution often credits ads for conversions that would have happened without them. I built a system that separated the two across $3.5 billion a year in ad spend.
A flawed test can make a model look better than it is. Before one rollout, I found target leakage and a misapplied variance reduction method that would have overstated the lift.
When return on ad spend fell for several large advertisers, the targeting model took the blame. The real cause was budgets scaling into lower-intent audiences. Finding it saved the company a costly model rebuild.
Every test should inform a specific decision. If the result would not change anything, the test is not worth running.
I compare every result to what would have happened without the change, using holdouts, matched groups, or causal models.
That includes flat and unexpected results. A clear no is as useful as a yes.
I build testing and measurement your team can keep using after the engagement ends.
I design A/B tests with adequate sample sizes and automate setup and reporting, so teams can test more often.
When a fixed-horizon test is too costly, I use multi-armed bandits to shift traffic toward the better-performing option during the test.
I measure the lift a campaign actually caused. One self-serve lift testing program I built covered more than 4,000 advertisers.
I build attribution and media mix models that show which channels drive sales and where to reallocate budget.
I analyze where new users drop off, which cohorts retain, and what customers are worth over their lifetime.
I also build the data infrastructure behind this work, including ad log pipelines, audience data, and clean room reporting.
Tell me about your goals, and we can find a time to talk.