Lesson 1 of 4
The analyst's job: from business question to decision
2 min read
A data analyst turns data into decisions. Tools change; this loop doesn't:
- Ask: what decision will this analysis inform? "Why did applications drop 20% last month?" beats "analyse the applications data".
- Get: find the data (SQL, exports, APIs, spreadsheets) and understand each column.
- Clean: fix types, missing values, duplicates, inconsistent labels.
- Explore: summarise, slice, chart, look for patterns and outliers.
- Analyse: compare segments, test hypotheses, quantify impact.
- Communicate: a clear answer, the evidence, and a recommendation.
Turning vague asks into questions
| Stakeholder says | You clarify |
|---|---|
| "How are sales doing?" | Compared to what: last month, target, same month last year? Which regions or products matter? |
| "Users are leaving." | How do we define churn? Over what period? Which users: new, paying, everyone? |
| "Make a dashboard." | Who looks at it, how often, and what will they do differently after seeing it? |
Metrics that matter
- Counts and sums: orders, revenue, applications.
- Rates: conversion rate (applied ÷ viewed), churn rate, click-through rate.
- Averages vs medians: salaries and order values are skewed; the median is often more honest.
- Growth: month-over-month %, year-over-year % (removes seasonality).
- Funnels: viewed → applied → shortlisted → hired, with the conversion at each step.
- Cohorts: group users by when they joined and track them over time; reveals whether the product is actually getting better.
The toolkit
- Excel / Google Sheets: quick analysis, pivots, small datasets, sharing with business teams.
- SQL: pulling and aggregating data from databases (the most important analyst skill).
- Python (pandas) or R: cleaning and analysing larger or messier data, automation, statistics.
- Power BI / Tableau / Looker Studio: dashboards and self-serve reporting.
You don't need all of them on day one. SQL + Excel + one BI tool gets most freshers hired; Python makes you stand out.
Check yourself
- Rewrite "look into our website traffic" as a specific question tied to a decision.
- Why might median order value be more useful than the mean?
- What is a cohort analysis and what does it reveal?
Sign up free to save your progress.