The method: stages and questions
The detailed form of the five-step method: twelve delivery stages, what each produces and what shows it's done, where each touches the shared library, and the questions asked inside every stage.
1Five steps, twelve stages
The front page describes five steps. The twelve stages below are the same work at delivery grain. Stage 1 covers both of the first two steps, because framing the use case names the owner and their authority.
| Step | Stages |
|---|---|
| 1. Define the outcome | 1. Frame the use case |
| 2. Name the owner | 1. Frame the use case |
| 3. Use the Compass | 2. Identify sources3. Design the analysis4. Prepare and validate data |
| 4. Carry it out | 5. Build a proof of concept6. Test with users and refine7. Document and rehearse operations8. Enable the people9. Release into operation |
| 5. Check and adjust | 10. Observe and improve11. Revalidate periodically12. Communicate results and learning |
2The twelve stages
Each stage names what it produces, what evidence shows it's done, and where it touches the shared library. Library identifiers are shown in mono.
1. Frame the use case
- Produces
- An aim with a range and an owner; the decision it serves. Ask first: what decision needs this data, whose information is involved, and what uses are outside the purpose? Who benefits, who is affected, and who has authority?
- Shown by
- An approved measurement profile
- In the library
Q0, Q1·outcome aim and owner_role·measurement-profile
2. Identify sources
- Produces
- Each requirement marked satisfied, missing or unknown
- Shown by
- A coverage reading, with freshness
- In the library
OP-D01·F20 coverage·F19 evidence trust
3. Design the analysis
- Produces
- A draft detection or analytic, tied to its requirement
- Shown by
- The logic, reviewed against the requirement it serves
- In the library
detection.author·signal patterns·OP-D04
4. Prepare and validate data
- Produces
- Sources parsed, normalized and validated
- Shown by
- Validation results and parse quality
- In the library
data_set.load·telemetry.validate·OP-D02·KPI-EI-17
5. Build a proof of concept
- Produces
- A bounded pilot
- Shown by
- Pilot results against the aim
- In the library
process.pilot
6. Test with users and refine
- Produces
- Accepted, or refined and re-run
- Shown by
- Acceptance against stated criteria, by the user, not the builder
- In the library
deliverable.accept·release.test
7. Document and rehearse operations
- Produces
- A playbook someone else can run
- Shown by
- The playbook, and a drill
- In the library
playbook.draft·OP-D06
8. Enable the people
- Produces
- People who can use it
- Shown by
- The intended users complete a representative task, not just attend training (reuse the acceptance or playbook exercise)
- In the library
person.certify (records only)- Gap
- The library records training; nothing yet delivers it.
9. Release into operation
- Produces
- Live, under change control
- Shown by
- Readiness gate met
- In the library
detection.promote·release.deploy·OP-D09·F23
10. Observe and improve
- Produces
- Tuned, with repairs verified. Feedback from the people affected can change the aim, the measure or the rule, not only tune the system
- Shown by
- Re-measured inside its aim, and held
- In the library
detection.tune·F22 repair verified
11. Revalidate periodically
- Produces
- Still doing what it was built to do
- Shown by
- Dated validation, in production as well as on known samples
- In the library
detection.test (known samples)·KPI-SP-05- Gap
- The library's detection test covers known samples only; production validation needs more.
12. Communicate results and learning
- Produces
- The decision-makers and the people affected know, and the next aim is set
- Shown by
- The decision owner records what continues, what changes, and the next aim and owner (Q8), and says so to the people affected
- In the library
message.send·message.publish·Q8
3The question loop
The same questions come back inside every stage. A question asked is not a stage done: answering "What should we do?" (Q5) is not the same as running a pilot or a user test.
| ID | Question | Stage of the loop | A full answer gives |
|---|---|---|---|
Q0 | What is our aim? | Aim | The range each measure should hold, how long, and who owns it. Without it, Q7 has nothing to measure against. |
Q1 | What exactly are we measuring? | Meaning | The metric, its scope, the services behind it, the data sources, the window and the calculation. |
Q2 | What changed? | Measure | The baseline, the anomaly, the trend, or the deviation from what was expected. |
Q3 | Why is it happening? | Evidence | Evidence correlated across technical and business systems, with confirmed findings kept apart from hypotheses. |
Q4 | Why does it matter? | Consequence | The consequence in operational, financial or experience terms. |
Q5 | What should we do? | Work | A recommendation backed by evidence, with a human approval step. |
Q6 | What happens next? | Trigger and outlook | Patterns and forecasts, where they are justified. |
Q7 | Did it work? | Verify | The outcome measured against the original objective (the Q0 aim). |
Q8 | What did we learn, and what do we change? | Learn | Whether the aim, the definition, the trigger or the owner should change after the incident, and who approved the change. |
Questions that cut across the loop
| ID | Question | A full answer gives |
|---|---|---|
X1 | Who owns it? | The owner of the measure, who decides and who approves. |
X2 | How sure are we, and what is missing? | Trust in the data, confidence in the finding, and the data named as missing. |
X3 | Are we keeping up? | Whether the loop absorbs ambiguity as fast as it arrives (Ambient Breach). |
4Known gaps
- Stage 8, Enable the people: The library records training; nothing yet delivers it.
- Stage 11, Revalidate periodically: The library's detection test covers known samples only; production validation needs more.
5Status and sources
The twelve stages come from ODAM v1's two lifecycle diagrams, in practitioner wording. The question list is the shared library's, at status drafted (October 31, 2026); Q0 ("Question Zero") and Q8 were approved by Chris Perkins on that date.
The method is a guide, not a gate sequence: a team may work several stages at once, and the questions apply throughout. The lineage from v1 is on Reference: lineage.