Research Design

Research Design

Use When

• Use when a research project needs formal design, research questions, method selection, case logic, trend analysis, MROC design, or report assembly.
• Use when the project should draw from design-research methods for complex problems, innovation, service design, product discovery, usability, journey mapping, or evidence-backed recommendation design.

Do Not Use When

• Do not use for a quick source lookup that does not need formal research design.

Initial Design Inputs

• Research topic, intended output, timeframe, audience, evidence types, and constraints.

Design Entry Sequence

• Select the relevant design layer below and pair with research-orchestration for long-running projects.

Design Principle

• Research questions, methods, evidence, and contribution must be explicitly connected.

Early Design Warnings

• Do not skip the design document on multi-week projects or overclaim from weak data.

Initial Design Deliverables

• Research design document, method plan, case-selection logic, trend-analysis design, or report-builder route.

Book-derived additions

Pair with dissertation-writing-process when a formal research design must be carried through literature, methods, findings, discussion, and submission gates.

References

• Use the routing table below to load only the relevant design reference.

Formal research-design layer. The orchestrator (research-orchestration) decides what kind of research is being run; this skill governs how that research is designed from question to artifact, when the project demands a formal design rather than an opportunistic sweep.

Research design starts with reasoning discipline. Use critical-reasoning-and-argument when framing empirical puzzles, theoretical puzzles, research questions, hypotheses, case-selection logic, method justification, and expected contribution.

When to load

The canonical design document (always for projects ≥1 week)

For any project longer than a one-day sweep, load references/research-design-document.md and produce the 3–4 page document with:

• Empirical puzzle — what is the engine observing that is not yet explained?
• Theoretical puzzle — which existing frame breaks down here?
• Research questions — concrete, answerable, derived from the puzzles.
• Action list — sub-agent waves, methods, evidence targets.
• Argument logic — the puzzle-question-method-contribution chain tested through critical-reasoning-and-argument.

The document is the engine's north-star artefact. Every sub-agent reads it and may update it.

Six design layers (load when relevant)

1. Historical methods

For historical / archival research: primary-source priority, source appraisal (Burke pentad — load source-evaluation), chronology discipline, historiographical-debate framing.

2. Trend analysis

For forecasting / signal extraction: durable trend vs fad, driver vs symptom, leading vs coincident indicators, signal-to-noise filters.

3. MROC (Market Research Online Community)

For longitudinal qualitative research with a recruited panel: recruitment screen, moderator playbook, prompts cadence, qualitative coding pipeline, quantitative scaffolding (polls, ratings), synthesis cadence.

4. Knowledge lifecycle (Bergeron)

8-stage state machine: discovery → capture → curation → storage → retrieval → application → re-use → retirement. Each stage has its own issues, support mechanisms, and value contribution.

5. Research design document (Abbott)

The 3–4 page artefact that names the empirical puzzle, theoretical puzzle, research questions, and action list. Maintained for the life of the project.

6. Research report builder

The pipeline that pulls from research/, analysis/, and opportunities/ files, applies a topic-appropriate report schema, and emits a designed .docx via the engine's professional-word-output layer.

7. Systems and process description

For research that studies how a system, institution, market process, workflow, data architecture, or operating model works, load systems-process-requirements. It supplies boundary discipline, process maps, state/event tables, requirements catalogues, data dictionaries, and traceability.

8. Universal design methods

For complex problems, innovation, service design, product discovery, usability, journey mapping, and recommendation design, load references/universal-methods-of-design-router.md. It routes among Hanington and Martin's 100 methods and forces a method plan that names the question answered, required participants or artifacts, output artifact, triangulation partner, and limitation.

Orchestration

new long-running research project
├─ load research-orchestration (type / discipline / reading mode / wave plan)
├─ load research-design
│   └─ produce references/research-design-document.md artefact
├─ during waves: load matching design layer
│   ├─ historical → references/historical-research-methods.md
│   ├─ forecasting → references/trend-analysis.md
│   ├─ design-method selection → references/universal-methods-of-design-router.md
│   ├─ panel-based → references/mroc-design-and-management.md
│   └─ artefact-flow modelling → references/knowledge-lifecycle-pipeline.md
└─ at finish
    └─ load references/research-report-builder.md to assemble the deliverable

Universal anti-patterns

• Skipping the design document; running a multi-week project as if it were a single sweep.
• Treating a trend analysis as "list of things happening recently" without leading-vs-coincident discipline.
• Running an MROC as a focus group; it is longitudinal and qualitative + quantitative.
• Selecting design methods because they sound impressive rather than because they answer a specific research question.
• Mixing knowledge-lifecycle stages in the same artifact (capture vs curation vs application have different rules).
• Building the final Word document by hand-stitching files; use the report-builder pipeline.
• Letting the design document drift; it must be updated as the project evolves, not snapshot at kickoff.
• Research question, method, or case selection that is plausible but not argued.

Companion skills

• research-orchestration — the dispatcher; this skill is its formal-design layer.
• critical-reasoning-and-argument — mandatory for puzzle framing, research questions, hypotheses, method justification, case selection, and contribution logic.
• research-techniques — specific named techniques applied inside design phases.
• systems-process-requirements — systems, processes, workflows, scope boundaries, requirements, interfaces, state models, data architecture, and design-system descriptions.
• source-evaluation — every source consulted.
• academic-writing, report-and-proposal-craft — output containers.
• uganda-academic-research, kenya-academic-research — East African institutional research-handbook requirements.
• data-quality-pipeline, dataset-discovery-and-analysis — for empirical projects.

Inputs

Outputs

Evidence Produced

Capability Contract

Design and critique default to read-only. Recruitment, participant contact, data collection, ethics submission, file mutation, or publication requires explicit authority and applicable approval.

Degraded Mode

When evidence access, participants, tools, or institutional rules are unavailable, return a qualified design with infeasible methods removed and gaps marked. Do not treat an untested method or missing ethics check as approved.

Decision Rules

Worked Example

For a historical case study, map the question to archival and secondary sources, select the historical-method and case-selection references, state missing-record limitations, and preserve the chain from evidence to contribution.

Anti-Patterns

• Designing without an answerable question. Fix: resolve the puzzle and question.
• Selecting a method by fashion. Fix: map method to evidence need.
• Ignoring infeasible access. Fix: revise the design.
• Claiming causality from weak design. Fix: narrow the claim.
• Treating missing ethics review as approval. Fix: mark it unresolved.

Workflow

1. Define the question, audience, timeframe, and contribution; stop if the question is not answerable.
2. Select methods and cases that can produce the required evidence.
3. Map each question to evidence, analysis, limitation, and output.
4. Test feasibility, ethics, and contradictory evidence.
5. Recover from missing access by narrowing the question, changing method, and recording the gap.

Quality Standards

Every question maps observably to a feasible method, evidence source, analysis step, limitation, and intended contribution; unresolved ethics or access checks block approval.