1. What data-driven really means
Data-driven decision·Building decisions on a verifiable evidence chain: hypothesis first, then data, then decision and review. Data calibrates experience; it does not replace judgment.
Experience proposes hypotheses and explains anomalies; data falsifies and corrects. Experience alone calcifies, data alone fragments. The best operators make the two challenge each other.
2. Four rungs of analytics
| Rung | Question answered | Typical tools |
|---|---|---|
| Descriptive | What happened | Reports, dashboards |
| Diagnostic | Why it happened | Drill-down, funnels |
| Predictive | What will happen | Regression, ML models |
| Prescriptive | What to do | Simulation, optimization |
Most firms stall between rung one and two. Diagnostic skill is the watershed: decomposing a sales drop into price, channel, product and competition hypotheses is where data literacy begins.
3. An operating checklist
Frame a falsifiable hypothesis
Slow landing page causes low conversion is a hypothesis. Conversion is low is just a complaint.
Find minimal sufficient evidence
Use existing data first; do not build a warehouse for a small decision.
Run comparisons
A/B tests are the gold standard; otherwise approximate causality with control groups.
Audit the data
Check definitions, outliers, survivorship bias. When numbers conflict, align definitions first.
4. AI cuts both ways
Language models democratize data access: natural-language queries, automatic reports, anomaly alerts. The cost of consuming data collapsed; the cost of governing it rose. AI hallucinates numbers and mixes definitions, so key conclusions must be checked against source data.