Skip to content
Was It Causal?

Decisioning · Decide

What should happen after the click?

Who should receive which action, if any? The optimization signal ladder, lifecycle experimentation, uplift modeling, and the systems that turn estimates into governed decisions. Where causal ML stops being an estimate and becomes a policy.

The decision this discipline makes

Two ways to spend one contact budget on one audience. Ranking by who is likely to convert against ranking by who is likely to be moved. Same people, same cost, different result.

0 500 1,000 1,500 2,000 2,500 0% 25% 50% 75% 100% Share of the audience contacted Incremental conversions Contact budget · 40% 2.4× at the same cost Uplift 2,346 at budget Propensity 996 at budget
Both policies contact the same number of people and spend the same money. Ranking by who is likely to convert mostly selects people who were going to convert anyway. Ranking by who is movable finds 2.4 times the incremental conversions. The curves meet on the right because contacting everyone is the same decision either way. Simulated data · scripts/lifecycle_uplift.py

Worked examples

Full analyses in this discipline, carried from the business decision through to a recommendation and an explicit account of what the result does not establish.

Applied Worked example

Broad Nurture, Propensity, or Uplift?

One trial-to-paid nurture sequence, four targeting policies, and a randomized holdout. Compared on incremental conversions rather than model accuracy, which is the only comparison that changes the answer.

Simulated data Read  →

Lessons

Applied

Choosing the Right Campaign Optimization Event

The deepest event that stays reliable, timely, and frequent enough to train on. Alignment sets a ceiling you cannot buy your way past, precision decides how close you get, and only one of them improves with volume.

Read  →
Foundation

The Case for Doing Nothing

Contact fatigue, capacity, and cost mean no-action is a real treatment arm. Why it belongs in every decisioning system by default, and why organizations reliably leave it out.

Read  →
Technical

Uplift Modeling in Practice

Where to cut the ranking, how much to withhold, and why the offline number is larger than the money. Estimating heterogeneous treatment effects when you have a randomized holdout, and validating the policy prospectively.

Read  →
Applied

From Segment to Audience to Treatment Policy

Segments, audiences, scores, assignments, and policies are five different objects. A result is only comparable to a previous one if you can say which of the five changed.

Read  →
Applied

Why Agentic Marketing Still Needs Holdouts

Two agents, the same budget and the same population, differing only in what they are scored on. One improves its dashboard by 65% while destroying value, and the one doing the right thing looks worse.

Read  →

The machinery behind this

Lessons say how to answer the question once. These are the standing systems that answer it on a schedule, each built around the unit it is allowed to withhold.