Study design wizard

The Study Design Wizard is Delphi Studio's canonical 11-step process for preparing a study. It carries one study record from initial setup through protocol review, methodology, content, recruitment, communications, and the final launch gate.

The sequence is deliberate: decisions that affect participants or interpretation are completed before invitations are requested, and workflow-controlled state changes occur on the server rather than through an arbitrary browser edit.

The canonical 11-step approach

The steps are grouped into three working phases:

• Foundation — Steps 1–3 establish the study, its protocol and ethics context, and its methodological approach.
• Study design — Steps 4–8 define the material panelists rate, the analytic rules, and the iterative-round experience.
• Recruitment and launch — Steps 9–11 prepare the panel, participant communications, and final readiness review.

Step reference

#	Step	What you configure	Why it matters
1	Study Setup	Study title, purpose, study lead, institution, and methodological starting point	Creates the study record and establishes its accountable owner and research purpose.
2	Protocol & Ethics	Protocol document, version, registration, ethics status, approval details, and institution	Connects the configured study to the governing protocol and ethics evidence.
3	Method	Delphi type, anonymity model, methodological structure, panel approach, and iteration strategy	Defines how expert judgment will be elicited and protected.
4	Items	Statements, domains, source details, manual entry, and CSV import	Defines the material the panel will evaluate and preserves its structure.
5	Rating	Scale length, labels, dimensions, and response requirements	Defines how panelists express judgment.
6	Consensus	Agreement and disagreement zones, median and dispersion rules, stability, and exclusion thresholds	Establishes the prespecified classification contract.
7	Rounds	Planned and maximum rounds, duration, scheduling, reminders, and stopping behavior	Defines the iterative operating plan.
8	Feedback	Quantitative summaries, controlled comments, prior-rating display, and feedback locking	Makes the between-round intervention explicit and reproducible.
9	Panel	Eligibility criteria, stakeholder roles, recruitment targets, experts, and panel CSV import	Connects panel composition to the research question before contact begins.
10	Communications	Invitation and reminder messages, consent language, re-consent policy, sender details, and quiet hours	Makes participant-facing communication and consent behavior reviewable before delivery.
11	Review & Launch	Readiness checks, acknowledgments, invitation request, and audited launch transitions	Prevents recruitment or Round 1 from starting with unresolved blockers.

The dashboard is outside this sequence. It is the operational workspace researchers return to after setup and during recruitment and rounds.

Study Setup and Protocol & Ethics (Steps 1–2)

Start by naming the study, explaining its purpose, identifying the responsible team and institution, and choosing a methodological starting point. The next step records the protocol version, ethics or IRB status, registration details, and supporting document.

Protocol upload is a real governed operation. Accepted documents are PDF, DOCX, or ODT files. The service validates the filename, declared content type, file signature, and size before storing the document, and records a SHA-256 digest so the exact uploaded version can be identified later. A filename extension alone is not treated as proof of file type.

note

Protocol documents can contain sensitive research information. They are stored through the backend document service, not in browser storage, and should be backed by private object storage, malware scanning, retention policy, and access logging in production deployments.

Method (Step 3)

Method establishes the study's core design without duplicating the detailed settings that follow. Choose the Delphi approach, anonymity model, panel structure, and iterative strategy here; configure the exact scale, thresholds, round limits, and feedback packet in Steps 5–8.

Typical starting approaches include:

Approach	Best used for
Modified e-Delphi	Health services research, competency frameworks, and guideline development
RAND/UCLA Appropriateness	Structured assessment of clinical indications or scenarios
Custom	Studies whose protocol requires explicit configuration of every methodological choice

See Comparing Delphi types before choosing a starting approach.

Items (Step 4)

Items are the statements, competencies, indicators, or scenarios the panel rates. Each active item should use clear wording and belong to a meaningful domain.

You can enter items individually or import a validated CSV. The minimum import columns are:

Column	Example
statement	The service should document consent before each remote encounter.
domain	Clinical practice

The import service checks required columns, row values, duplicate statements, file size, and row count before committing the result. Import failures return row-level feedback rather than silently accepting a partial or malformed list.

Item wording and lineage remain important after setup. A material revision between rounds should create a new version or an audited amendment rather than overwrite the wording panelists previously rated.

Rating and Consensus (Steps 5–6)

These are separate steps because they answer different questions:

• Rating defines how participants respond: scale length, labels, dimensions, and required or optional responses.
• Consensus defines how those responses are interpreted: agreement zones, disagreement rules, minimum median, maximum IQR, stability, bimodality, and any cross-stakeholder condition.

Consensus settings form the study's statistical contract. The rules in effect for a round are frozen with that round's analytic snapshot. Later permitted changes require an audited protocol amendment; they do not rewrite historical classifications.

See Consensus analytics for the classification model and Interpreting results for derived re-analysis.

Rounds and Feedback (Steps 7–8)

Rounds defines the cadence and stopping envelope. Feedback defines what panelists see between rounds. Keeping these concerns separate makes it possible to report both the operating schedule and the intervention that may influence subsequent ratings.

Step	Representative settings
Rounds	Planned rounds, maximum rounds, duration, open and close behavior, grace window, reminder cadence, stopping rule
Feedback	Median or distribution display, own prior rating, approved comments, redaction and approval policy, packet locking

Feedback is controlled study material. Lock the approved packet before release so every eligible panelist receives the intended version and the study retains evidence of what was shown.

Panel (Step 9)

Define eligibility, exclusions, expertise, stakeholder roles, recruitment targets, and any conflict-of-interest requirement before requesting invitations. Experts can be entered individually or imported from CSV.

The minimum panel import columns are:

Column	Example
email	expert@example.org
first_name	Alex
last_name	Rivera
specialty	Clinical informatics
institution	Example University

Import validates addresses, required values, duplicate rows, and study-level duplicates. Adding an expert creates a study relationship; it does not itself send an email. Selection, invitation, acceptance, consent, COI disclosure, and enrollment remain separate events.

See Panel management and Expert directory and recruitment.

Communications (Step 10)

Review invitation, invitation-reminder, round, re-consent, withdrawal, and completion messages before recruitment. Configure sender identity, quiet hours, access-link lifetimes, consent language, and the default behavior after a material protocol amendment.

Invitation delivery uses an outbox rather than sending mail inside the study update. The email worker claims a queued message, receives the one-time link only for delivery, and reports a delivered or failed result through an authenticated callback. Delivery callbacks are idempotent, so a provider retry does not double-advance the study or duplicate an audit event.

Panelist onboarding is gated:

1. The invitation must be delivered and accepted.
2. Consent records the protocol version, attestation, signature name, actor, and time.
3. A required COI declaration must be completed.
4. Enrollment occurs only after those prerequisites are satisfied.

Review & Launch (Step 11)

The final step aggregates readiness across all previous steps. A blocker must be resolved; a warning may require explicit acknowledgment and rationale.

Common blockers include:

• no active items or unassigned required domains;
• missing protocol or ethics details required by the selected status;
• unset rating or consensus rules;
• incomplete consent or participant communication content;
• no eligible panelists or unmet recruitment thresholds; and
• required COI or preflight acknowledgments that are incomplete.

Launch is not a generic status edit. The server exposes explicit, permission-checked transitions:

1. Request invitations validates preflight, creates hashed one-time invitation credentials, and queues messages.
2. Recruiting begins from confirmed delivery events, not from a browser-only flag.
3. Open Round 1 is allowed only when the study is recruiting and the enrolled, consented, and required-COI counts meet the configured gate.

Each transition records the actor, role, action, affected entity, prior and resulting revision, timestamp, and bounded transition metadata.

Changes after recruitment

After recruitment begins, fields use one of three policies:

Policy	Meaning
Editable	The value can still be changed through the normal study API.
Audited	The change is allowed but records before-and-after values and may require a rationale.
Locked	The value cannot be changed directly because participants or results already depend on it.

A material locked change uses the amendment workflow:

1. Submit an allowlisted patch with a substantive reason.
2. Store it as pending without changing the active protocol.
3. Require an authorized owner or reviewer to approve it, or record rejection with its reason.
4. Apply the approved change as a new audited revision and trigger the configured re-consent behavior when applicable.

Historical round, protocol, consent, and feedback snapshots remain unchanged.

Templates

Delphi Studio supports two forms of reuse:

Type	Where it enters the workflow	Purpose
Method preset	Study Setup (Step 1)	Seeds a methodological starting point such as Modified e-Delphi or RAND/UCLA Appropriateness.
Study template	Study Setup and Panel	Applies organization-scoped taxonomy, panel, rating, consensus, round, feedback, consent, and communication defaults.

Templates standardize structure and policy; they do not silently send invitations or bypass Review & Launch. To reuse item wording from an earlier study, use an explicit duplication or item-import workflow so lineage and validation remain visible.

Authoritative study record

When connected, the backend study record is authoritative. Every mutation requires an authenticated identity, study membership, permission for the requested action, and the expected study revision. A stale update is rejected as a conflict and the client refreshes from the server.

Browser storage may retain a last-known cache for resilience, but it is not the production source of truth and cannot directly advance recruitment, launch, consent, COI, or amendment state.

See Governed study record for the full authority, access, validation, and audit model.

Next steps

• Governed study record — backend authority, access control, and workflow integrity
• Panel management — recruit and manage experts
• Running rounds — open, close, and iterate
• Audit trail — inspect lifecycle and amendment evidence
• Collaboration & roles — understand study permissions