The Analysis Workflow from Question to Action
The seven stages of a piece of analysis and why the work moves back and forth between them.
11 min
Activities: 0 of 2
Why a workflow matters
Most weak analysis does not fail because of a wrong formula. It fails because a stage was skipped. Someone opens a file, builds a chart, and only later finds that the data covered the wrong period, or that nobody needed the answer. A clear workflow protects you from this.
Experienced analysts use roughly the same stages, whatever the tool. The names change between organisations, but the logic does not. Each stage has an output that you can check before you move on.
Seven stages of an analysis
- 1
1
Frame
Agree the question, the decision it supports, the person who decides and the deadline. Output: a one-paragraph brief. - 2
2
Acquire
Find the sources, request access and extract the data. Output: raw files or queries, stored unchanged. - 3
3
Prepare
Profile, clean, join and reshape the data. Output: an analysis-ready table and a log of every change. - 4
4
Explore
Plot distributions, look at segments and test first ideas. Output: a short list of patterns worth checking. - 5
5
Analyse
Answer the question with the right comparisons and test other explanations. Output: findings with their size and uncertainty. - 6
6
Communicate
Lead with the answer, show the evidence and recommend an action. Output: a report, briefing or dashboard. - 7
7
Follow up
Check what happened after the decision. Output: a short note on whether the action worked.
The work is not a straight line
The stages are in order, but real work loops back. Exploration often shows that the question was badly framed. Preparation may show that a source cannot answer the question at all. When this happens, go back to the person who asked and agree a new brief. Do not quietly answer a different question and present it as the original one.
Worked example 1. A logistics manager asks why late deliveries rose in the third quarter. During preparation, the analyst finds that the definition of "late" changed in July, from "after the promised day" to "after the promised two-hour window". The rise is mostly a definition change. The analyst returns to the frame stage and agrees with the manager to compare both quarters under the new definition.
Worked example 2. A clinic administrator wants to know which appointment types are most often missed. Exploration shows that one doctor's diary records cancellations as "missed". The analyst fixes the coding in preparation, then reruns the exploration before drawing any conclusion.
Common mistakes and their fixes
During preparation you find that the source cannot answer the question that was agreed. What is the best next step?
Put the stages in order
Drag the stages into the order in which they normally start.
- 1Follow up on the outcome
- 2Explore patterns
- 3Communicate the finding and action
- 4Frame the question and the decision
- 5Acquire the data
- 6Prepare and clean the data
- 7Analyse and test explanations