MARKETING · BRAND · CONTENT · PRODUCT MEASUREMENT

Half your growth would have happened anyway. Which half?

Which campaigns, stores, content, and features actually moved revenue — online and in store. Measured with as much rigor as the decision needs, and ending in a next step your leadership can act on.


Same quarter. Same revenue. Two ways of measuring it.

The difference is growth you were going to get anyway.

Illustrative synthetic data. Not a client result.


Why no tool can answer this for you

The method depends on how treatment was actually assigned, what data you can actually get, and what you actually can't do. All three live inside your organization, not in a dataset.

Tools have made the estimation cheap. They can't tell whether your data can answer the question — which is why anything sold as fully automated measurement hands you a confident number instead of a correct one.

Not every decision needs a causal model. Sometimes a descriptive cut is enough; sometimes the honest answer is "not measurable yet." The rigor is sized to the decision, the data, and the pace of the business.


Questions this work answers

  • Did that campaign drive incremental revenue, or buy customers we already had?
  • Our online ads say they drove $X. How much of that showed up in our stores?
  • We spent on TV and podcasts last quarter. Did it move sales, or just awareness?
  • Should we roll this promotion out nationally, or did the pilot region flatter it?
  • Our new stores look great — is that the stores, or the markets we put them in?
  • Is our content program driving sales, or just traffic?
  • We launched the feature to everyone. Did it improve retention, or did our best customers just adopt it first?
  • Should we buy GrowthBook, Statsig, or Eppo — or fix the process first?
  • Can our team run this without outside help in six months?

Capabilities

Geo experiments
GeoLift, Trimmed Match, holdout design, power and geo-count planning
Quasi-experimental estimation
Synthetic control, difference-in-differences, staggered adoption — for promotions, store openings, and launches that weren't randomized
Online experimentation
Experiment design, power and MDE planning, guardrails, SRM and validity diagnostics
Brand and content
Upper-funnel and content impact on online and in-store sales
Product features
Full launches and opt-in features, where early adopters flatter the result
Measurement systems
Metric layer, experiment pipelines, analysis templates, platform selection

Also: uplift targeting, switchback designs, LTV modeling, forecasting.

Engagements start with a short paid diagnostic: what's worth measuring, what's measurable, and what isn't. It ends in a decision memo written for the business leader — the recommendation, how confident it is, what it's worth, and the next three actions. Technical detail goes in the appendix. If the answer is that no engagement is needed, the memo says that.

Delivered by one accountable practitioner, with engineering coordination where a client lacks the resource. Accountability covers the approach, the quality, and the business result.

Selected writing


Find out which half.