Framing Business Questions, Data and Metrics · Lesson 1

From a Business Need to an Answerable Analytical Question

4 min read

Lesson objectives

Translate a broad business need into an analytical question defining the decision, measure, period and segments.
Distinguish an answerable descriptive question from a causal claim not supported by available data.
Record a measurement gap instead of inventing a variable or unsupported conclusion.

Start with the decision, not the tool

Business data analysis does not begin with “Which chart should we use?” or “Which software should we open?” It begins with a business need and a possible decision. A manager may say, “Performance is declining.” That is a broad observation, not yet an analytical question. We need to know which decision could change, which measure, period and groups will be compared, and which data actually exist. This translation prevents collecting large amounts of irrelevant data and makes the scope testable.

Separate the symptom, the decision and the analytical question

Element	Weak wording	Better wording
Symptom	Sales are bad	Completed-order value declined compared with the previous period.
Decision	We need to do something	We want to determine whether a region or customer segment should be prioritized for intervention.
Analytical question	Why is performance worse?	How did average completed-order value and completion rate change from Q1 to Q2 by region and customer type?

The better question does not promise a cause we have not measured. “Why” may require causal variables that are absent; the comparison in the table can be answered from the current order log. If a difference appears, it identifies an area worth investigating, not automatic proof of the cause.

Exercise

Exercise 1 — about ten minutes: Convert “customers are unhappy” into an answerable analytical question. Use a fictional dataset containing a satisfaction rating from 1 to 5, date, region and customer type. State the decision, period, segments and measure. Do not invent variables that were not provided.

Check your answer

Guided solution: “To decide which region should be prioritized for service improvement, how did average satisfaction and the share of ratings 1–2 change over the last three months versus the previous three months, by region and customer type?” The decision is explicit, both measures are computable from the stated data, and periods and segments are defined. It does not claim to explain the cause of dissatisfaction because no cause variable is available.

Before moving to data, write the question as if another analyst will receive it. They should understand what to calculate, for which period, on what unit, and which decision the result will support. If words such as “better,” “success,” or “impact” remain undefined, stop and operationalize them before building a table or chart.

Lesson summary

A good analytical question connects a business need to a decision, measure, period, segments and explicit data limits. It describes what can be measured, does not jump from association to causation, and records a measurement gap when data prevent a stronger answer.