Executive Learning · Free learning
Data Analysis
From description to decision, and how to challenge a weak conclusion.
In this module
0 of 6 activities complete
Reading
0% readFour levels of analysis
Descriptive analysis says what happened. Diagnostic analysis says why. Predictive analysis says what is likely next. Prescriptive analysis recommends the action.
Most organisations claim to be predictive while operating at descriptive. Being excellent at descriptive is worth more than being mediocre at predictive.
Correlation, causation and confounding
Two series moving together tells you nothing about direction or cause. Before you act, ask what third factor could drive both, and what experiment would settle it.
Executive questions that expose weak analysis
What is the sample and what does it exclude, what would change your conclusion, what is the confidence interval, and who reviewed this independently.
Real-world case studies
Case study · Marketing
A subscription business stops rewarding customers it was already keeping
Situation
A subscription business attributed most retention to a costly discount campaign because retained customers were far more likely to have received it.
Challenge
The campaign had been targeted at engaged customers, so the analysis was measuring who was selected rather than what the discount achieved.
How data was used
Engagement, tenure, plan type and prior support history were assembled so that comparable customers could be matched, and a genuine holdout group was created.
How AI was applied
An uplift model estimated the incremental effect of the discount per customer rather than the raw retention rate.
Business outcome
Discounting was withdrawn from the segment that would have stayed regardless and concentrated on genuinely at-risk customers, improving margin at equal retention.
Lesson for the learner
Ask what the comparison group is before accepting any performance claim. Without one, you are measuring selection, not effect.
Data Analysis, executive briefing
18:35 · Video briefing
Key takeaways
- Descriptive, diagnostic, predictive and prescriptive answer different questions
- Correlation never establishes causation without design or experiment
- Four standard challenge questions expose most weak analysis in a meeting
Stress-test a recent recommendation
Apply the four challenge questions.
- Take a recent data-backed recommendation you received
- Ask each of the four challenge questions in writing
- Record which one the analysis cannot answer
Module 5 quiz
2 questions · 70% to pass
01Ice cream sales and drowning incidents both rise in summer. This is an example of:
02Which question most quickly exposes a fragile analysis in a board meeting?Applied
Your notes
Lesson discussion
Ask questions, share how you are applying this in your organisation, and learn from other executives.
Sign in to join the discussion. Reading is open to everyone.
Loading discussion…