Your starting point
Prerequisites: Use the ideas and vocabulary from lesson 24; revisit its worked examples if needed.
Definition: An analytical question specifies what is being examined, for whom or what, over which period, and against which comparison. Operationalization translates a broad concept into an observable measurement and a rule for calculating it.
Clarify the decision before the metric
“Improve engagement” is too broad to analyze directly. Ask what action someone might take after seeing the result. A teacher deciding what to reteach needs different evidence from a manager planning server capacity. Define the outcome in terms of the decision, then check whether the available data actually measures it. A convenient field should not silently redefine the goal.
Specify scope and comparison
Name the population, row unit, period, numerator and denominator where relevant. Decide whether the question asks for description, comparison, prediction or causation. “Which channel had more conversions?” differs from “Which channel had a higher conversion rate?” and from “Which channel caused more conversions?” The last question requires stronger design evidence than a table of totals.
Write the analysis plan in plain language
List the needed fields, exclusions, quality checks, calculation and intended display. Make a prediction before running the summary. This helps you notice when a result conflicts with your assumptions instead of rewriting the question after seeing an attractive number. If the necessary evidence is absent, narrow the question or request additional data rather than asking AI to fill the gap with a plausible story.
Two worked examples
A workshop question
Broad request: find out whether the course is successful. A narrower question is: among learners who started week one, what fraction submitted the independent application by the end of week two? This defines a cohort, outcome and deadline. It still measures submission, not necessarily the quality of independent performance.
A campaign comparison
Broad request: identify our best campaign. One measurable question is cost per recorded conversion during the same campaign period, using spend divided by conversions. A campaign with zero conversions has an undefined cost-per-conversion ratio and should be reported separately, not declared free or assigned a fabricated value.
Where this becomes useful
A small organization has many dashboards but little clarity about actions. A one-paragraph analysis question can prevent time spent polishing metrics that do not help anyone decide what to do next.
Make an attempt
Open the complete workbench with Marketing campaigns. Write a question and prediction in the report fields before calculating. Identify the numerator, denominator and scope, then generate one summary that directly answers your stated question.
Misconception check
Claim: The dataset should decide the question automatically.
Inspect the correction
Available fields constrain what can be answered, but the human decision should determine which answer is useful.
Independent application
Rewrite a vague request from your own work into two measurable questions. For each, specify population, time, outcome, comparison, needed fields and a limitation. Choose one and outline the calculation before inspecting any results.
Completion records practice and self-review. It is not certification of statistical competence.
