Skip to main content

The Delphi lifecycle

Every Delphi study in Delphi Studio moves through five distinct stages: Design, Run, Analyze, Audit, and Report. Rather than treating these as informal notes or external spreadsheets, Delphi Studio treats each stage as a first-class, guided workflow. This structure helps researchers maintain methodological rigor while reducing administrative burden.

The lifecycle is inherently iterative. Most studies move through multiple cycles of running rounds and analyzing results before reaching a point where all items are classified and the study can conclude.

1. Design

The Design stage is where the foundation of the study is established. Delphi Studio provides an 11-step wizard that guides researchers through defining the study's identity, objectives, methodology, content, and governance before any panelist is invited.

AreaKey decisions and examples
IdentityPrincipal investigator, institution, IRB approval, protocol version
ObjectiveResearch aim, consensus question, target outputs (e.g., competency framework, quality indicators)
MethodDelphi type (e.g., Modified e-Delphi), level of anonymity, planned number of rounds, stopping rules
ContentItem bank creation or import, taxonomy tagging, rating scales, consensus thresholds and zones
GovernanceConsent language, email templates, quiet hours, comment approval mode, export and data retention settings

Checkpoint: Once the researcher completes the Review & launch step, the protocol is marked as complete. Recruitment can then begin, and upon activation the full methodology is frozen into a launch snapshot. This snapshot becomes the immutable reference for all future analysis and reporting.

2. Run

The Run stage covers all activities related to engaging the expert panel and collecting data across rounds.

TaskLocation in Delphi StudioNotes
Invite and manage panelistsPanelSupports stratified recruitment and expertise tagging
Collect informed consentPanelist magic linkConsent is recorded before any ratings are collected
Open and close roundsRoundsCan be manual or scheduled
Send remindersAutomated via cadence engineConfigurable timing and quiet hours
Lock feedback (Round 2+)Feedback packetsEnsures every panelist receives identical, approved information

Throughout the Run stage, response anonymity is structurally protected: panelists never see the identities or individual responses of peers. Researchers ordinarily work with stable pseudonyms (for example, Panelist 1), while any authorized identity reveal is reasoned and audit logged. This preserves confidentiality while allowing longitudinal response tracking and necessary panel administration.

3. Analyze

When a round closes, the Analyze stage begins. Delphi Studio's analytics engine automatically computes a comprehensive set of results for each item:

  • Median, interquartile range (IQR), and zone percentages
  • Classification into Agree, Disagree, or Indeterminate zones
  • Additional metrics such as Kendall's W, stability between rounds, and bimodality flags
  • Subgroup breakdowns by expertise domain or demographic characteristics
  • A stopping recommendation based on the proportion of items classified and observed stability

Researchers can review these results through visual dashboards (including heatmaps) and may override dispositions when justified, with the rationale recorded in the audit log. For studies that collect open-text comments, researchers can optionally run a governed, AI-assisted qualitative analysis pipeline to identify themes before preparing feedback for the next round.

4. Audit

The Audit stage ensures that every decision and event that influenced the final results is documented in a transparent and tamper-evident way.

Delphi Studio automatically records the following types of events:

Event categoryExamples recorded
Design changesUpdates to consensus rules (with before/after values)
Panel eventsInvitations, consent, withdrawals, unavailable marks, stakeholder role updates
Round eventsRound opened, closed, deadline extended, feedback locked
Ratings and commentsAll submissions (stored pseudonymized), including unable-to-rate
AI-assisted analysisModel used, prompt version, researcher adjudication
Study eventsResearcher-recorded external evidence, outages, deviations (separate from the hash chain)
ExportsTrace bundles, CREDES reports, REFI-QDA files, re-analysis versions

The audit log is hash-chained, meaning each entry includes a cryptographic hash of the previous entry. This creates a tamper-evident record suitable for IRB closeout, journal review, or any situation where the integrity of the research process must be demonstrated. See Audit trail.

5. Report

In the Report stage, the completed study is transformed into publishable and reusable outputs.

Delphi Studio supports the generation of multiple artifacts:

OutputFormatPurpose
Study reportIn-app view + DOCXComprehensive internal or funder report
Manuscript sectionsAI-assisted drafts (human reviewed)Methods, results, and discussion sections
CREDES-aligned exportMarkdownMethods section structured around the CREDES domains
Consensus tablesCSV / embedded in reportsReady for publication or supplementary materials
Qualitative dataREFI-QDA (XML)Import into NVivo, MAXQDA, or ATLAS.ti
Trace bundleChecksummed archive (per-file SHA-256 manifest)Complete reproducibility package

These outputs are designed to reduce the manual effort required to prepare findings for publication while maintaining full traceability back to the original data and decisions. See Reporting & exports.

Iteration loop

Delphi is fundamentally an iterative process. After the Analyze stage, items that remain indeterminate flow back into the Run stage for another round of rating and feedback.

Researchers can follow the platform's stopping recommendation or make their own determination based on the pattern of remaining indeterminate items and overall stability.

Study status vs. lifecycle stage

Delphi Studio uses study status labels that map to the broader lifecycle stages:

Study statusTypical lifecycle stageDescription
draftDesignStudy is being configured in the wizard
recruitingDesign complete → Run (recruitment)Panelists are being invited and consented
activeRun + Analyze (iterating)Rounds are being conducted and analyzed
completedReport (+ Audit archive)All rounds finished and final outputs generated
terminatedEnded earlyStudy was stopped before completing all planned rounds; no further data collection

In addition, any study can be archived. Archived is an overlay on the current status rather than a status of its own: it locks all fields against editing while preserving the full record.

Next steps