Skip to content

Experiment-calibrated marketing measurement.

Ongoing doctoral research by Ismail Rhoulam, PhD researcher in Marketing Science at LARMODADD, FSJES Souissi, Mohammed V University in Rabat, since 2026. The questions below are open research questions, not findings, and the doctoral research has no publications yet.

01 Established practice

Credit is not causation

Attribution assigns credit; it does not measure what would have happened without the spend. Controlled experiments, such as holdouts and geo tests, are one of the few ways to observe that difference directly.

02 Ongoing research

Experiment-calibrated measurement

Marketing mix models and everyday reporting are cheap to run but easy to fool. Randomized experiments are reliable but narrow and expensive. This research looks at how to combine the two, so that experiments check and correct everyday measurement wherever a test is feasible.

03 Ongoing research

Transportability

A lift test ran in one market, in one quarter. When does its result hold for the market and period where the budget is actually decided, and how should it be adjusted when it does not? In client work today, every test records its market, period and audience, so its result can be reused with that context in view.

04 Ongoing research

Adaptive experimentation

How can an experiment shift its allocation as data comes in, so it answers the question the decision depends on sooner and at lower cost? In client work today, the decision rule and the test duration are agreed before launch.

05 Ongoing research

Risk-aware budget allocation

Effect estimates come with uncertainty. How should a budget account for the downside instead of optimizing expected return alone? In client work today, recommendations state a range and its downside, not a single return figure.

Questions

The questions, in plain language

  • Did marketing cause the outcome, or merely receive credit for it?
  • Can what we learned from one market or campaign inform decisions elsewhere?
  • How should budget move when performance estimates are uncertain?
  • Can experiments check whether everyday measurement is telling the truth?
  • How can tests adapt as evidence arrives, without wasting spend?

Research and client work stay separate.

Becoming a client never means your data is used for research. Any research use of client data would require separate, explicit consent, anonymization, appropriate contractual terms and ethics approval where required. Participation is never a condition of receiving consulting services.

Research collaborations

Researchers and companies interested in marketing measurement, experimentation or budget allocation under uncertainty are welcome to get in touch.