Free lesson of "Data Analysis for Managers" by Tomasz Zielinski, 8 min
Managers rarely say "I need a cohort analysis". They say "sales feel slow" or "customers seem unhappy". Your first job is to translate a feeling into a question with four parts: a metric, a period, a comparison and a decision it could change. "Sales feel slow" becomes "Did monthly revenue from new customers fall in the last quarter compared with the same quarter last year, and should we move budget from events to search ads?"
Write the question before you open any file. Then list what data you would need to answer it and where it lives: orders, customers, ad spend. Often you discover that a column is missing, for example the source of each customer, and you save hours that you would have spent staring at the wrong table.
State your hypothesis and what would change your mind. If you believe the drop comes from fewer orders, not smaller orders, say so, and say what number would prove it wrong. Revenue equals orders times average order value, so you can split a change into those two parts and see which moved. For example, orders down 12 percent and average value up 4 percent gives a net drop of about 8.5 percent.
Beware questions that cannot be answered with the data you have. "Why do customers leave?" cannot come from an orders table. "Which customers did not order again within 90 days?" can.
Task: take one worry from your own work and rewrite it as a four-part question. List the two or three columns you need and where you would find them.
This was the first of 12 lessons. The full course is €199, with a certificate at the end.
Sales open soon