Thinking Like an Analyst

The Mindset That Turns Data Into Insight

Tools and techniques matter, but the analytical mindset matters more. This chapter develops the thinking habits that make analysis valuable.

Start With Questions

Question-First Thinking

Don't start with data. Start with questions.

What do you want to know? What decision will this inform? What would change your mind?

Good Questions

Specific: "What is the average order value for repeat customers vs. first-time customers?"

Answerable: Can your data actually answer this?

Actionable: If you find the answer, what would you do differently?

Question Hierarchy

Big questions break into smaller questions:

Big: "How can we increase revenue?" Smaller: "Where is revenue coming from?" Smaller: "Which products have highest margins?" Smaller: "What drives customers to buy high-margin products?"

Start big, then decompose.

Be Skeptical

Of Your Data

  • Is this data complete?
  • How was it collected?
  • What biases might exist?
  • Are there errors or anomalies?

Of Your Analysis

  • Am I measuring what I think I'm measuring?
  • Could there be another explanation?
  • Am I confirming what I expected?
  • What would change my conclusion?

Of Your Conclusions

  • How confident should I be?
  • What are the limitations?
  • What don't I know?

Understand Causation

Correlation ≠ Causation

Ice cream sales and drowning deaths both increase in summer. Ice cream doesn't cause drowning. Summer causes both.

Finding Causation

Causation requires:

  1. Correlation (they move together)
  2. Time order (cause precedes effect)
  3. No confounding variables (nothing else explains it)
  4. Mechanism (plausible reason why)

When to Be Careful

Marketing attribution: Did the ad cause the sale, or would they have bought anyway?

A/B testing: The gold standard for causation. If possible, run experiments.

Observational data: Most business data is observational. Be humble about causation claims.

Think in Distributions

Not Just Averages

The average can mislead. Two groups with the same average can look completely different.

Distribution Questions

  • What's the spread?
  • Are there outliers?
  • Is it symmetric or skewed?
  • What's typical vs. extreme?

Example

Average customer lifetime value: $500

But: 50% of customers are worth $100, and a few whales are worth $10,000+. The "average customer" doesn't exist.

Consider Base Rates

What's Normal?

Before getting excited about a finding, ask: Is this unusual?

Example

"Our conversion rate is 3%!" Is that good? What's typical for your industry? What was it last month?

Comparison Is Essential

Numbers alone mean little. Compare to:

  • Previous periods
  • Other segments
  • Industry benchmarks
  • What you expected

Control for Confounds

What Else Could Explain This?

If sales increased after a marketing campaign, was it the campaign? Or seasonality? Or a competitor's mistake? Or a price change?

Isolate Variables

Try to compare like with like. Control for factors that might muddy the picture.

Ask "What Else?"

Always ask what alternative explanations exist.

Accept Uncertainty

You Rarely Have Certainty

Most analysis produces probabilistic conclusions, not absolute truth.

Communicate Uncertainty

"We're confident that..." vs. "It appears that..." vs. "The data suggests..."

Make Decisions Anyway

Uncertainty doesn't mean paralysis. Make the best decision with available information, then update as you learn more.

Common Thinking Traps

Confirmation Bias

Seeing what you expect to see. Look for disconfirming evidence.

Survivorship Bias

Looking only at successes. What about failures you don't see in the data?

Anchoring

Being overly influenced by the first number you see.

Cherry-Picking

Selecting data that supports your conclusion.

Over-Precision

False precision suggests false confidence. "Sales increased 17.34%" sounds more precise than knowledge allows.

AI Prompt: Thinking Check

Help me think through my analysis critically.

My finding: [What you concluded]
My data: [What you analyzed]
My method: [How you analyzed it]

Please challenge my thinking:
1. What alternative explanations exist?
2. What assumptions am I making?
3. What could be wrong with my data?
4. How confident should I actually be?
5. What would disconfirm this conclusion?

What's Next

Good thinking requires good data. Let's talk about getting it ready.

Next chapter: Getting and cleaning data.

Data Analysis FundamentalsGetting and Cleaning Data