Running rounds

Rounds are the operational core of any Delphi study. They represent the structured, iterative cycles through which expert judgment is collected, refined, and synthesized toward consensus. Delphi Studio provides clear workflows for creating, scheduling, opening, closing, and iterating rounds while maintaining methodological rigor and full auditability.

Round lifecycle

Every round progresses through five defined states:

State	Description	Researcher actions
draft	Round is being prepared	Define timing, confirm items, set windows
scheduled	Round is configured and will open automatically at the scheduled time	Monitor countdown; can still edit before opening
open	Panelists can submit ratings and comments	Monitor response rates in real time
closed	No new ratings accepted	Trigger analysis or review results
analyzed	Statistics, classifications, and dispositions have been computed	Review results, prepare feedback, plan next round

This lifecycle ensures that every round follows a consistent, documented process from preparation through analysis.

The Rounds view shows each round's state, timing window, response rate, and available actions.

Creating Round 1

After panel recruitment and consent are complete, you can create the first round:

1. Go to Rounds → Create round (or use the round automatically created from your wizard defaults).
2. Confirm that the correct active items from the item bank are included.
3. Set the opensAt and closesAt dates to define the rating window (for example, a 10-day period).
4. Open the round manually, or let the scheduler open it at opensAt. When the scheduler opens a round, it sends round_open emails to eligible panelists, respecting configured quiet hours; opening a round manually does not send emails. On both paths, panelists with a pending re-consent requirement are held back until they re-acknowledge.

Once opened, panelists can begin rating immediately.

Planning subsequent rounds (Round 2+)

Before opening Round 2 or later, use the Plan next round feature. It becomes available only after every disposition from the previous round has been confirmed. The tool automatically selects items based on their disposition from the previous round:

Disposition	Default inclusion in next round	Notes
re_rate	Included	Items that need another round of rating
revise	Included (with revised wording)	Researcher has updated the item
Indeterminate	Included	Items that did not yet reach consensus
retain	Excluded	Consensus achieved
drop	Excluded	Item removed from the study

You can manually override the default inclusion for any item. Excluding an item that would otherwise carry forward requires a rationale, and the complete plan is recorded in the audit log before the next round's items are created.

Panelist-proposed statements

When the method allows new item suggestions, panelists may propose statements during a round. Proposals are de-identified and appear in Plan next round for researcher adjudication:

Outcome	Effect
Approved	Creates a rateable item with panel provenance retained
Rejected	Kept as a rejected suggestion for audit; not added to the item bank

Approved proposals can be scheduled into the next round like any other active item.

Feedback packet lock (Round 2+)

For studies using controlled feedback, you can require that feedback packets be locked before opening Round 2 or later. Pre-open locking is opt-in, controlled by the feedback lock required setting (feedbackLockRequired).

Step	Purpose
Preview	Review medians, histograms, zone distributions, and approved de-identified comments
Lock (optional)	Create an immutable snapshot of the feedback for each panelist before the round opens
Open round	Panelists see feedback immediately upon entering the round

How freezing works: When the pre-open lock is required, the system blocks round opening until packets are locked, so every panelist receives exactly the same information from the start. When it is not required, each panelist's feedback packet is frozen the first time they view it — it cannot change for them afterward — but panelists who first view at different times may see feedback generated at different moments. Requiring the pre-open lock removes that possibility.

Here is the typical sequence:

Deadlines, grace windows, and extensions

Round timing is configured in the wizard and refined when you create or open each round.

Capability	What it does	Why it matters
Timezone preview	Before you commit a deadline, preview what that instant means in every panelist's local timezone	Fairness for international panels
Grace window	Optional prespecified minutes after closesAt (up to 24 hours) during which late submissions are still accepted and labeled as in-grace; the close sweep waits out the window	Avoids “cron lottery” acceptance of late ratings
Deadline extension	Extend an open round’s closesAt with an audited change and fair notice to the panel	Clinical schedules slip; extensions keep the process defensible

Grace minutes apply only when accept-late / grace is enabled for the study, and are capped at 24 hours — for anything longer, use an audited deadline extension instead. Extensions never rewrite ratings already submitted; they only move the close boundary forward.

Automated operations

Delphi Studio's background worker automates several round-related tasks on a schedule:

Job	What happens	Researcher control
Round open	Moves scheduled round to open at opensAt	Can also open manually
Round close	Moves open round to closed after closesAt (plus any grace window)	Can also close manually
Analytics	Runs consensus engine after round closes	Automatic, but results are reviewed by researcher
Reminders	Sends pre-close nudges to non-responders	Configurable cadence and quiet hours

Researchers retain the ability to manually open or close rounds when needed.

Response-rate floors

On the Analysis & reporting screen, you can set a minimum response rate threshold for each round. If a round reaches its close time below this floor:

• The round is held open — closing is blocked — until you acknowledge the low response with a documented rationale.
• An attrition caveat is added to reports and the stopping recommendation.

Example:

Round	Responses	Floor	Action
1	12 / 15 (80%)	70%	Close normally
2	8 / 15 (53%)	70%	Round held open until low response is acknowledged

This feature helps maintain transparency about participation levels.

Stopping recommendations

After each round is analyzed, Delphi Studio provides a stopping recommendation based on the current state of the study:

Recommendation	Typical recommended action
consensus_reached	Move to final reporting and mark the study complete
tentative_consensus_low_response	Treat consensus as provisional; document response-rate caveats
stable_contested	Review remaining indeterminate items; consider accepting limitations or adjudicating
still_moving	Plan another round of rating
max_rounds_reached	Conclude the study and report with caveats on unresolved items
insufficient_history	Not enough round history for a stability-based stop

These recommendations are advisory. Researchers can make their own determination while documenting their reasoning.

Item revision between rounds

When items remain indeterminate or require improvement, you can revise them between rounds:

1. Review indeterminate items on the Results dashboard.
2. Edit the item wording in the item bank (this creates a new version linked to the original).
3. Optionally run qualitative analysis on panelist comments to inform the revision.
4. Use Plan next round to include the revised items in the upcoming round.

All item revisions are tracked with full version history for transparency and auditability.

Round monitoring

While a round is open, you can monitor progress through several views:

View	Information provided
Round detail	Completion progress, number of comments received
Dashboard	Overall study-level participation metrics
Tasks	Pending actions such as reminders or blockers

These views help you stay informed and take timely action (such as sending reminders) without disrupting the panelists' experience.

Next steps

• Consensus analytics — how statistics are computed
• Interpreting results — heatmaps, overrides, re-analysis, stopping
• Panelist experience — feedback and proposals from the panelist side
• Multi-round worked example — end-to-end scenario