Reference
Templates and checklists
Part of the Primary Research Synthesis skill · loaded on demand from SKILL.md
Copy-ready structures for every deliverable this skill produces. Placeholders in angle brackets are permitted here and nowhere else. Read this file when you are producing an output and want the exact structure; the reasoning behind each is in SKILL.md and references/frameworks.md.
Contents
- Research design sheet
- Interview log
- Coding table for an excerpt
- Codebook entry
- Analytic memo
- Saturation log
- Axial paradigm table
- Theme hierarchy with evidence counts
- Insight statement
- Espoused / in-use gap table
- Quote bank entry
- Synthesis pack structure
- Pre-fieldwork checklist
- Pre-delivery checklist
1. Research design sheet
Engagement: <client, workstream>
Question the research must answer:
<one sentence, from problem-framing>
Hypotheses to test: <from hypothesis-stress-testing, or "none: exploratory">
Segments and counts: <segment>: <n> | <segment>: <n> | ... Total: <n>
Site or unit spread: <how sites were chosen; name the deviant cases>
Observation plan: <n> sessions of <type>, <duration>, at <where>
System / document data for triangulation:
<source>: <what it will corroborate>
Format: <duration>, <structure>, <recording and consent>, <transcription SLA>
Sequencing: <n> fixed up front; <n> held back for theoretical sampling
Stop rule (set now): <per-segment criterion>; minimum <n> per segment before saturation claimed
Anonymisation rule: <attribution level, e.g. role + region; minimum group size for attribution>
Team: <interviewer>, <note-taker>, <coder(s)>; second coder on <n> percent
2. Interview log
| Int | Date | Segment | Role | Site / region | Duration | Transcript ref | Coded on | Coder | Notes |
|---|---|---|---|---|---|---|---|---|---|
| 01 |
3. Coding table for an excerpt
Excerpt <n>. <Role>, <region or site>, interview <nn>.
"<verbatim text, fillers removed, no other edits>"
| Segment of text | Open code | Tag (E / U) |
|---|---|---|
| "<fragment>" | <code> |
4. Codebook entry
| Code | Definition | Include | Exclude | Example (source) | Tag | Count |
|---|---|---|---|---|---|---|
| <kebab-name> | <one sentence: what an instance is> | <what counts> | <what does not; name the code it goes to instead> | "<short verbatim>" (<int>) | E / U | <n interviews> |
Rule: a second coder must be able to apply the entry without asking you. If the Exclude column is empty, the code is probably too broad.
5. Analytic memo
Memo <n> Date: <date> After interview: <nn>
Trigger: <what in the data prompted this memo: a conflict, a repetition, a surprise>
Observation: <what you noticed, with interview numbers>
Provisional claim: <the relationship you think is there>
What it would take to disconfirm:
<the data that would show this is wrong>
Next sampling move: <who to interview or observe next, and why>
6. Saturation log
| Int | Segment | New codes | Cumulative codes | New properties on top themes? (which) |
|---|---|---|---|---|
| 01 | n/a |
Below the table, state: code saturation reached at interview <n>; meaning saturation on top themes reached at interview <n>; segments meeting the stop rule: <list>; segments below minimum size: <list, reported as "consistent across n" rather than "saturated">.
7. Axial paradigm table
| Paradigm element | Content | Evidence (interviews, observations, data) |
|---|---|---|
| Phenomenon | ||
| Causal conditions | ||
| Context | ||
| Intervening conditions | ||
| Action / interaction strategies | ||
| Consequences |
Leave a row blank rather than invent content. A blank row is a finding about what the data does not show.
8. Theme hierarchy with evidence counts
| Theme | Interviews (n / N) | <seg 1> | <seg 2> | <seg 3> | Intensity | Triangulation |
|---|---|---|---|---|---|---|
| A. <theme stated as a claim, not a topic> | High / Medium / Low | <non-interview source> | ||||
| Counter: <the strongest opposing view> |
Intensity definitions: High = raised unprompted and with emphasis by most who mentioned it; Medium = raised when probed, or unprompted by a minority; Low = acknowledged when put to them.
9. Insight statement
Insight <n>. <One sentence. A claim about mechanism or behaviour, not a topic.>
Evidence: <n> of <N> interviews (<segment split>); <observations>; <system data>.
Intensity: <High / Medium / Low>, and among whom.
Counter-evidence: <who disagreed and why>, or "none found".
Strength: <High / Medium / Low> on <existence / magnitude / prevalence>, separately if they differ.
Would be overturned by:
<the specific evidence that would refute it>.
Implication: <one sentence: what it means for the decision>.
10. Espoused / in-use gap table
| Topic | Espoused theory (policy, leadership statements) | Theory-in-use (incidents, observation, records) | Evidence | Gap type | Governing variable |
|---|---|---|---|---|---|
| <topic> | <one sentence in the organisation's own words if possible> | <one sentence describing observed behaviour, with a number where one exists> | <int, obs, data refs> | Unaware / Known and tolerated / Known and concealed / Policy is wrong or ambiguous | <Model I value, or other> |
Gap types and the response each implies:
| Gap type | Closes with |
|---|---|
| Unaware | Information: show the incident data to the people who hold the espoused theory |
| Known and tolerated | A change to the incentive or the governing value; information alone will not move it |
| Known and concealed | Removing the reason to conceal; workshops will not surface it |
| Policy is wrong or ambiguous | Changing the policy to match the better practice, or resolving the ambiguity |
11. Quote bank entry
| Quote ID | Verbatim text | Attribution (role, region) | Interview / line | Theme(s) | Cleared for client? | Cleared for external use? |
|---|---|---|---|---|---|---|
| Q<nn> | "<exact words; fillers removed only>" | <role>, <region> | <int nn>, lines <a-b> | <A, C> | yes / no | yes / no |
Rules: never merge two speakers into one quote; never alter wording beyond removing fillers and marking omissions with an ellipsis; never attribute at a level that identifies an individual (if a segment has fewer than three people, attribute one level up); keep the transcript pointer so any quote can be verified on request.
12. Synthesis pack structure
1. Governing insight (one page)
- The core category as a headline sentence
- Three to five supporting insights, each with its count
- What this means for the decision at hand
2. Insight statements (one per page, template 9)
3. Theme hierarchy with evidence counts (template 8)
4. Espoused / in-use gap table (template 10)
5. Negative cases and counter-evidence (what did not fit, and what it changed)
6. Method appendix
- Research design sheet
- Interview log (anonymised)
- Saturation log
- Codebook (consolidated)
7. Quote bank (cleared quotes only)
Hand the pack to executive-communication for the storyline; the pack is the evidence base, not the presentation.
13. Pre-fieldwork checklist
- The question the research must answer is written in one sentence and agreed with the engagement lead.
- Every hypothesis the research is expected to test is listed, with what would count as disconfirming evidence.
- Segments include every party in the mechanism under study, including at least one external or counterparty view where feasible.
- At least one deviant case (site, team or person where the outcome differs) is in the sample on purpose.
- At least one non-interview source (observation, log, document sample) is planned for each major expected theme.
- The interview guide pairs every policy question with an incident question on the same topic.
- No question in the guide can be answered yes or no without a follow-up.
- The stop rule is written down, with a minimum per segment.
- Consent, recording and anonymisation rules are agreed with the client and stated to each interviewee.
- Coding begins after the second interview, not after the last.
14. Pre-delivery checklist
- Every theme is stated as a claim, not a topic heading from the guide.
- Every insight carries a count (n of N), a segment split, an intensity rating and a counter-evidence line.
- Every count can be reproduced from the codebook and interview log.
- The core category explains the negative cases, or the write-up says why it does not.
- The espoused/in-use table has at least one non-interview source on the in-use side for each row where one exists.
- Every quote is verbatim, attributed by role and region only, and traceable to a transcript line.
- No quote has been chosen because it is vivid rather than representative; where a vivid minority quote is used, the text says it is a minority view.
- The saturation log is included, and any segment below the minimum is labelled "consistent across n", not "saturated".
- The lead hypothesis from upstream is explicitly reported as supported, not supported, or not testable with this data.
- Hand-offs to
value-driver-modeling,strategic-choiceandexecutive-communicationare listed with what is passed.