Ebrahim AlhamedFrameworks Library

m.05 · II · Inference & Prediction · Synthesis & Communication

From Data to Decision

The last mile: turning confidence intervals into a board recommendation.

The core idea

An analytics project is only as good as the decision it enables. The final skill is translating a model into a recommendation a non-technical stakeholder can act on — emphasising the decision, the uncertainty, and the cost of being wrong. A model that is statistically significant but makes a £2 difference per transaction is a fascinating curiosity, not a £10M rollout. — after decades of dashboards nobody used

The hero diagram

The analytics maturity ladder.

Each rung requires the one below it.

Linear flow Left-to-right sequential flow: Raw → Reporting → Statistical → Predictive → Prescriptive. Raw → Reporting → Statistical → Predictive → Prescriptive
Raw
↓
Reporting
↓
Statistical
↓
Predictive
↓
Prescriptive

The tools on the bench

Ideas that pay rent.

Analytics Workflow · Christodoulou
Problem → Data → Analysis → Action
Without the fourth step, it is homework.
Known-Unknowns · Rumsfeld · 2002
known-knowns · known-unknowns · unknown-knowns · unknown-unknowns
Use data to move unknowns into knowns. Guard against unknown-knowns (gut feels).
Data Maturity · Analytics leadership
raw → cleaned → statistical → predictive → prescriptive
Climb one rung at a time. Level 5 requires all four below.
Statistical vs Practical · Executive decision-making
significant: real, not noise · practical: worth the rollout
Both must be true before you act.

How to apply

Writing the one-page executive memo.

  1. Lead with the decision. Not the method. The decision.
  2. State the uncertainty in business units. "We are 95% confident profit is between £1.2M and £1.8M" — not t-statistics.
  3. Name the cost of being wrong. If the CI straddles breakeven, say so. Recommend a pilot, not a rollout.
  4. Leave equations in the appendix. The CFO is not impressed by your R².

Key reading · Session 5 · Christodoulou

From data to decisions.

Most analytics projects die in the hand-off. The best regression is worthless if it confuses the decision-maker. Simplicity, business context and honest communication of uncertainty are not optional — they are the whole deliverable.

What should we do? Always answer the question being asked.

← m.04 Dummies, Interactions & Validation ··· back to DA