Open to Roles Consulting via Zenith Data
Roles
I'm currently open to senior and staff IC roles, and head-of-data roles, at companies where measurement is load-bearing — where the hard part isn't shipping the feature but knowing whether it worked, for whom, and what to try next.
That usually means experimentation platforms, causal inference, and the infrastructure between behavior and decisions. The best fit is a team that wants to own that stack rather than lease a vendor's black box.
Consulting
Through Zenith Data I take on e-commerce measurement work for outdoor and other high-consideration brands — the categories where purchases are deliberate, infrequent, and genuinely hard to attribute.
Engagements are audit-first: a data infrastructure audit, then a roadmap, then ongoing fractional work if it makes sense to keep going. I'd rather show you what's broken and what it's worth fixing than sell you a retainer.
The Shape of the Work
Article was spending millions a year on advertising with no way to see past the platforms' own attribution. I designed and ran a geo-experiment program using regional holdouts, backed by Bayesian structural time series and synthetic-control models, to measure what each channel actually moved. Platform-reported numbers turned out to be significantly over-attributed, and the program redirected millions in annual ad spend toward channels that held up under causal measurement.
A major publisher had millions of daily active users and a single proof-of-concept A/B test. I architected their experimentation platform — real-time assignment under 200ms, a Bayesian analysis pipeline, causal trees for heterogeneous treatment effects — and helped build the testing culture around it. Two years later: roughly a hundred experiments across seven studios, with double-digit revenue-per-user gains from targeted interventions.
For a brand whose customers deliberate for 30–90 days before buying, last-click attribution was the only measurement in place. I built the stack — server-side first-party tracking, geo-holdout experiments, cross-channel attribution — and taught the team to run it. Testing identified 15–25% of ad spend as non-incremental, and within three months the team was designing and reading its own experiments without me.
For a longer look at how I approach this work, the writing is the best place to start.
Contact
Email is best — tim@tradcliffe.com. Tell me the shape of the problem and I'll tell you whether I'm the right person for it.