Module 2 · 5 minutes

Module 02: Data Analysis Fundamentals

AI Data Analysis Services

READ

Here's a stat that should make you feel better about starting: McKinsey found that most business decisions are still based on gut instinct, not data. Even at companies that have analysts. The bar is lower than you think.

You don't need to understand standard deviations or p-values to be a useful data analyst in 2025. You need to understand five things: what data exists, how to clean it, how to ask good questions, how to spot patterns, and how to explain what you found. AI handles the technical middle. You handle the human bookends.

The Five Types of Business Data That Matter

1. Transaction data — What people buy, when, how much, how often. Every POS system, Shopify store, and Stripe account generates this. It's the most valuable and most underused data for SMBs.

2. Customer data — Who they are, where they're from, how they found the business. CRMs like HubSpot and even basic email lists contain gold.

3. Marketing data — Ad spend, click-through rates, conversion rates, email open rates. Scattered across Google Ads, Meta, Mailchimp, and a dozen other platforms.

4. Operational data — Inventory levels, staffing hours, delivery times, support tickets. Often trapped in spreadsheets or legacy systems.

5. External data — Industry benchmarks, competitor pricing, market trends. Publicly available but rarely collected.

The Data Analysis Process (Simplified)

Forget the textbook seven-step frameworks. Here's what actually happens:

Step 1: Ask "What decision are we trying to make?" Every analysis starts with a business question. "Should we run this promotion again?" beats "Analyse our Q3 sales data" every time. If the client can't tell you what decision the data should inform, help them figure it out.

Step 2: Get the data. Usually this means exporting CSVs, connecting to APIs, or literally asking the client to email you their spreadsheets. It's rarely glamorous.

Step 3: Clean it. This used to be 60-80% of an analyst's time. With AI, it's minutes. Upload a messy spreadsheet to ChatGPT and say "clean this data, identify issues, and standardise formats." Done.

Step 4: Explore and analyse. Ask AI to summarise the data, find patterns, run comparisons. Then use your brain to figure out what matters to the client's actual question.

Step 5: Tell the story. The analysis is worthless if nobody acts on it. Your job is to turn numbers into "here's what to do on Monday morning."

Quick Check

Put the data analysis process steps in the correct order:

The Skill That Actually Matters: Asking Good Questions

Ben Wellington, the data analyst behind "I Quant NY" (featured in the New York Times), built his reputation by asking simple questions of public data. He didn't use fancy algorithms — he asked "where do the most parking tickets get issued?" and found that the city was ticketing legally parked cars, saving New Yorkers millions.

Your superpower isn't statistics. It's curiosity combined with business sense.

When a client gives you a dataset, train yourself to ask:

What surprised me?

What's missing?

If I were the owner, what would I want to know?

What pattern would change a decision?

Common Beginner Mistakes

Mistake 1: Analysing everything. Don't. Find the 2-3 insights that matter and go deep.

Mistake 2: Leading with methodology. Clients don't care how you did it. They care what it means.

Mistake 3: Presenting data without recommendations. "Sales dropped 15% in March" is an observation. "Sales dropped 15% in March because you stopped email campaigns — restart them and expect a recovery within 6 weeks" is analysis.

Quick Check

Complete the key principle of data analysis delivery:

"Sales dropped 15% in March" is an observation. Adding the cause and recommended action makes it .

---

TRY IT

Data Exploration

I have a CSV file with [DESCRIBE YOUR DATA — columns, rows, what it represents]. 

Please:
1. Summarise the dataset (key stats, date range, completeness)
2. Identify data quality issues (missing values, duplicates, outliers)
3. Suggest the top 5 business questions this data could answer
4. Clean the data and flag what you changed

Start with the summary before doing anything else.

Pattern Detection

Analyse this dataset for patterns a business owner should know about:

Focus on:
- Trends over time (growing, declining, seasonal)
- Segments that behave differently
- Anomalies or outliers
- Correlations between variables

For each finding, explain: What's happening, why it might matter, and what action the business should consider.

Executive Summary

Based on this analysis, write an executive summary for a non-technical business owner. 

Rules:
- Maximum 300 words
- Lead with the single most important finding
- Use plain English (no jargon)
- Include 3 specific recommendations with expected impact
- End with "what to do this week"

EXERCISE

Practice Analysis — Real Data:

1. Go to Kaggle.com and download any business dataset (search "retail sales" or "ecommerce" — there are thousands of free ones)
2. Upload it to ChatGPT (Plus required) or paste a sample into Claude
3. Use Prompt 1 to explore it
4. Use Prompt 2 to find patterns
5. Use Prompt 3 to write an executive summary
6. Time yourself — this should take under 45 minutes

This is the exact workflow you'll use with real clients. Practice it until it's muscle memory.

---

KEY TAKEAWAYS

1You need curiosity and business sense more than statistics knowledge
2AI handles data cleaning and pattern detection — you handle the questions and the storytelling
3Every analysis must start with a business decision, not a dataset
4Present recommendations, not just findings — that's what clients pay for
5Practice on free Kaggle datasets to build speed and confidence

Back to course