How to Run and Report a Live Delphi Voting Session
A meeting can move quickly from a first rating to discussion and a second vote. Keep track of who responded, which wording they assessed, and what changed so the final result remains interpretable.
Specify what the session is allowed to decide
Before the meeting, identify the items for review, the people eligible to vote, the decision rule, and the time available for reflection. Agree what happens to items that remain unresolved.
A live poll is a collection method. A Delphi voting session also needs a defined relationship between independent responses, controlled feedback, discussion, and reconsideration. Continuous asynchronous Real-Time Delphi gives participants a longer active window; a facilitated live session puts those steps inside a shared meeting schedule.
State whether the meeting makes a final study decision, recommends revisions to the wider panel, or prepares questions for another online round. A small meeting group should not silently become the full panel.
What a hybrid polling study contributes
A 2026 methods paper describes four modified Delphi rounds combining online surveys with anonymous polling during two hybrid meetings. The researchers checked vote counts against attendance and revised items before repeat votes. They also acknowledged changing participation, possible social influence, and no formal evaluation of polling's added value. Read the methods paper.
The practical contribution is a documented workflow. It provides a useful example to adapt and evaluate; it does not establish that faster meetings necessarily produce more valid consensus.
Reconcile voters and item denominators
Record who is eligible for the session and how eligibility is checked. For each item, distinguish the people eligible to rate it, substantive ratings, explicit non-ratings, and missing responses. Use the denominator specified in the protocol.
Original example: more support or a different group?
An item receives 12 supporting votes from 16 valid ratings: 75.0%. Later, the same 12 supporting votes among 14 valid ratings produce 85.7%. Under an 80% rule the label changes, although the supporting count has not increased.
Check whether people revised their ratings, left the session, abstained, or became ineligible after an item changed. Where permitted, use stable analysis identifiers to distinguish individual changes from a different voting group. Preserve confidentiality in published outputs.
Keep each wording version and vote distinct
Give every item and vote a stable identifier. Record whether a second vote evaluates the original wording, a clarification, or a substantive change. Save the displayed wording and decision rule before collecting responses.
| Record | Purpose |
|---|---|
| Item ID, wording version, vote number | Identify exactly what was evaluated. |
| Eligible group and response counts | Explain the denominator and missing input. |
| Feedback and discussion summary | Document what informed reconsideration. |
| Rule, classification, and next action | Separate the vote outcome from later team decisions. |
When wording changes, report the next vote as a judgment on that version. An increase in support cannot be attributed solely to discussion if the statement or group also changed. Preserve important reasons for dissent.
Protect reflection time and independent judgment
Collect an independent initial response before discussion when the design calls for one. Present the agreed aggregate feedback, invite balanced discussion, and collect reconsidered responses privately. Avoid visible pressure to join the apparent majority.
Give remote participants equivalent access to the wording, evidence, discussion, and voting window. Explain what happens if a connection fails or a person misses part of an item. Attendance alone does not demonstrate that everyone had a fair opportunity to respond.
Set a reasonable limit on repeat votes. If an item needs substantial evidence review or redrafting, carry it into an asynchronous follow-up under the protocol. Persistent disagreement can be an informative result.
Report the session as part of the study
Report attendance and stakeholder composition, the number and timing of votes, the feedback shown, revisions, item-level denominators, classifications, and unresolved questions. Distinguish matched respondent changes from comparisons based on everyone responding at each time point.
For a Surveylet Live Consensus Session, plan the questions, feedback settings, participant access, and export requirements during preparation. Rehearse the sequence with the facilitator. Agree how the study team will retain the session log and reconcile saved results before producing the report.
The Live Consensus service can support preparation and delivery. A detailed session report or additional analysis can be scoped through Delphi Analysis & Final Report.
Source and further reading
O’Hagan E, Sheahen B, Cho K, et al. Integrating real-time polling into modified Delphi methodology: lessons from a consensus study on cardiac implantable electronic device remote monitoring. BMC Medical Research Methodology. 2026;26:73. First published February 25, 2026.
The denominator example and session-record table are original planning examples. For choosing the overall sequence of online rounds, dynamic feedback, and meetings, see our hybrid Delphi guide.
Prepare the meeting and the reporting together
Tell us which decisions the group needs to make, who will take part, and what the final report must show. We can help plan and configure the session.