23,624 public agenda records reveal a field that is highly practiced at scheduling attention and far less explicit about what should happen because people gathered.
The direct answer: the average public agenda record scores 21.8/100, and 92.2% show no visible follow-up, tracking, or impact-evidence signal. The largest difference between higher- and lower-scoring records is participant work: people doing something, not only watching something.
Scope: this is an open-web discovery corpus, not a census or probability sample. The 23,624 records map to 14,219 canonical event groups. A missing public signal is not proof that a practice did not occur.
Nearly half of the records show some form of participant work. Far fewer show commitments, follow-through, or impact evidence. Each bar uses the same 23,624-record denominator.
“Any visible follow-through signal” is the complement of records with no visible follow-up, tracking, or impact-evidence signal. The analysis concerns published agendas and related public evidence, not private organizer activity.
Each finding is written to stand on its own. Use “Copy finding” to preserve the number, claim, scope, and source together.
Across 23,624 scored public agenda records, the average Gathering Effectiveness Score is 21.8/100. The rubric rewards visible outcomes, participation, commitments, connection design, learning transfer, and evidence.
Zero of 23,624 records score above 60 under the strict public-evidence rubric. Only 57, or 0.2%, score above 50.
21,781 of 23,624 agenda records show no visible follow-up, tracking, or impact-evidence signal. This measures what the agenda makes visible; it does not prove that organizers did nothing after the event.
Follow-through averages 5.8/100, compared with 30.1/100 for the strongest average pillar. Owners, dates, checkpoints, and evidence plans are rare in published agendas.
100.0% of the top 5% of best-available event-group records include visible participant work, versus 7.2% of the bottom half, a 92.8-point difference.
2,854 agenda records are stage-only: passive formats with no visible participant work or commitment mechanism. A room can deliver information, but its distinct value is what people can do together.
Network design averages 17.5/100. Receptions and breaks create proximity, while matched introductions, peer circles, and structured ask-and-offer exchanges provide visible connection architecture.
Technology / AI / Startup records average 26.6/100 (n=3,229), while Trade Show / Expo records average 16.6/100 (n=1,056). This is descriptive, not a causal ranking of organizers or industries.
The 1,000-record comparison set, selected from higher-quality records after title, source, and duplication filters, averages 37.0/100, 15.2 points above the full corpus. Its follow-through pillar still averages only 7.5/100.
The 2020 United We Transform gathering ran 15 live experiments and recorded 7,545 networking interactions. Results include the misses: only 48% of AI networking recommendations received a positive rating.
The comparison below uses the highest-ranked available public record for each canonical event group: the top 5% (n=711) versus the bottom half (n=7,110). Because these mechanisms contribute to GES, this is a decomposition of the scored groups, not independent evidence that a mechanism caused better outcomes.
Percent of best-available event-group records with each visible mechanism.
Full 23,624-record corpus; every pillar is scored from 0 to 100.
Source: United We Transform public agenda corpus, scoring version ges_agenda_judgment_v2_0. Scope: 14,219 canonical event groups, using the highest-ranked available public record per group.
The research is most useful when it changes a planning decision. These questions convert the largest observed gaps into an agenda review.
Name the decision, behavior, relationship, or artifact the gathering should produce.
Design work, practice, deliberation, building, or exchange instead of continuous broadcast.
Make the owner, action, and date visible before the room closes.
Specify the follow-up moment, checkpoint, and route for keeping momentum alive.
Choose a baseline and a small set of evidence that can show whether anything changed.
External workplace and meeting research helps explain why gathering design matters. These studies are context, not validation: their populations, methods, and outcomes are not directly comparable with this observational audit of public agendas.
Microsoft reports that half of meetings fall in two concentrated windows, while 48% of surveyed employees describe work as chaotic and fragmented. The study combines Microsoft 365 telemetry with a 31-market knowledge-worker survey.
Read the primary reportIn Atlassian’s 5,000-person knowledge-worker survey, 62% said they showed up at most meetings without knowing the goal. The article also recommends documenting decisions, next steps, and owners.
Read the primary reportThe OECD’s cross-country synthesis reports measurable deficits in support, close friendship, and connection. That context strengthens the case for distinguishing unstructured networking time from intentionally designed relationship-building.
Read the primary reportThese sources provide contemporary context only; Microsoft and Atlassian are vendor research, and the OECD study uses different populations and measures. None independently validates United We Transform’s agenda-derived percentages. The share tools on this page apply the NFX sharing framework by reducing friction and making the recipient benefit explicit.
Use the summary as written, inspect the data, build an approved chart, or cite the report. Please preserve the scope: visible public-agenda evidence, not verified attendee outcomes.
United We Transform analyzed 23,624 public agenda records and 316,643 agenda blocks using a published eight-pillar rubric. The average score was 21.8/100; 92.2% showed no visible follow-through signal. Participant work was the largest rubric difference between higher- and lower-scoring records.
United We Transform. (2026). The State of Gatherings 2026: 23,624 public agenda records analyzed. https://unitedwetransform.com/reports/state-of-gatherings-2026/The open-web corpus contains 23,624 public agenda records and 316,643 agenda blocks. Some events have multiple records, such as a main agenda, day schedule, or track page. Comparisons of the top 5% and bottom half use the highest-ranked available public record from each of 14,219 canonical event groups to avoid over-weighting multi-page events.
Each record is scored with the deterministic ges_agenda_judgment_v2_0 rubric across eight pillars. Source-backed evidence is kept separate from inference. The dataset exposes source routes, scores, fields, and caveats under CC BY 4.0.
The statistics packet was generated 2026-07-16 under scoring version ges_agenda_judgment_v2_0. The report was first published 2026-07-02 and substantively updated 2026-07-31. Corrections, versioning, and methodological changes are documented through the trust boundary.
United We Transform began as a live experiment in how a large gathering could create stronger participation, connection, and evidence. Publishing both the wins and failures became the standard for examining the wider field.