Summary

  • ICANN85 recorded 2,195 attendees across in-person and virtual registration, while its 47-person figure was an average physical mid-session room count across a programme of 276 sessions. The figures describe different populations and must not be substituted for one another.
  • Small rooms can produce high-quality technical analysis, expose implementation risks and negotiate workable language. Their legitimacy rests on competence, openness, reasons and review, not on pretending that those present statistically represent affected users.
  • A room-level governance audit should identify unique and repeat entities, institutional roles, remote access, speaking concentration, agenda control, conflicts, decision stages and the path from intervention to outcome without turning participation into demographic voting.
  • ICANN should attach a sample description and authority statement to consequential sessions: who could attend, who was counted, what the gathering could decide, what remained advisory, and how absent interests could later challenge the result.

The number is a room measurement, not a miniature electorate

The most revealing statistic in ICANN's report on its March 2026 Community Forum in Mumbai is not the largest one. The ICANN85 By the Numbers report records 2,195 attendees: 1,517 participating in person and 678 registered to join virtually. It also records 276 sessions, 246 session hours and an average in-person attendance of 47 entities per session. The contrast between thousands registered and dozens in a typical room is not evidence that one figure is true and the other false. It is evidence that they measure different stages of participation.

The 47-person average came from mid-session counts of rooms. The report explains that attendance rankings relied on manual headcounts of people physically present at the moment of the count. It does not say that only 47 people attended ICANN85. It does not identify 47 unique individuals who moved together through every session. It does not include the virtual audience in that average. Nor does it tell us, without further data, whether a particular room contained staff, directors, long-serving community volunteers, first-time visitors, contracted support, speakers or observers in any given proportion.

Those qualifications do not empty the statistic of meaning. They make it usable. Forty-seven is a plausible view of the human scale at which much institutional work occurred: a chair, a panel, a few rows of specialists, observers entering and leaving, remote entities visible through a screen and a microphone queue that only some people use. It reminds us that a global institution often reasons in rooms smaller than a city bus. That fact deserves scrutiny because policies and implementation choices can eventually affect registrants, operators and users who never knew the room existed.

The wrong response is to ridicule the room. Every institution must divide labour. No global meeting can conduct all work in one plenary, and many technical questions are handled better by a focused group than by a stadium. The right response is to match the room's claim to the evidence it can support. A session can identify a defect, compare experience and recommend action. Presence alone cannot transform it into a sovereign chamber.

Four attendance populations are hidden inside two headline figures

The report's attendance page combines several populations that should remain analytically separate. First are registered people, a category created before anyone demonstrates actual engagement. Second are people who physically entered the venue. Third are virtual registrants, some of whom may have joined many sessions and others none. Fourth are session occupants observed in a mid-session snapshot. Each population answers a different operational question.

Registration helps organisers plan, communicate and describe broad reach. Venue attendance indicates travel and physical access. Virtual registration indicates potential remote reach. A room count estimates physical use at one moment. None measures unique active contributors, and none reveals whose evidence changed an outcome. Substituting one for another produces attractive but false stories. A claim that 2,195 people developed a particular position is no more justified by total registration than a claim that only 47 people participated in the entire forum is justified by the average room count.

Repeat participation complicates the picture further. A specialist may attend ten sessions, appearing in ten room snapshots. Another attendee may spend the week in bilateral meetings and enter one formal session. A staff member supporting several groups may be counted repeatedly while having no vote in any of them. A remote entity may contribute decisive evidence without appearing in the physical average. The sum of room counts therefore cannot be treated as a count of unique people unless identifiers are reconciled, and reconciliation itself must respect privacy.

There is also a denominator problem. Forty-seven may be high for a narrow drafting session and low for a cross-community discussion. It may fill a small assigned room or leave a large hall visibly empty. The relevant affected population changes with the issue. A procedural orientation needs no claim to regional representation. A session whose output informs a Board decision with commercial or rights consequences needs a much clearer account of notice, participation and review.

A defensible report would present a participation funnel: registered, authenticated on the schedule platform, joined at least one session, attended the relevant session, contributed, received a response and influenced the recorded disposition. The purpose would not be to award votes by activity. It would prevent the widest population at the top of the funnel from being invoked as if it had exercised power at the narrowest point below.

Why a small room can still be institutionally valuable

Small rooms have properties that mass participation cannot easily reproduce. Entities can test assumptions, ask follow-up questions and notice uncertainty in one another's answers. Chairs can identify where disagreement is factual, commercial, legal or terminological. Operators can explain consequences that do not appear in a consultation paper. Lawyers can expose the difference between an intended safeguard and enforceable language. A person responsible for implementation can say which proposed deadline is impossible and why.

Repeated interaction also creates shared technical vocabulary. That can accelerate work and reduce the risk that every intervention begins from first principles. The Internet's coordination institutions depend on such accumulated competence. Treating every experienced entity as suspect merely because they are experienced would waste knowledge and encourage symbolic turnover rather than accountability.

But expertise answers a question about the quality of judgment, not the source of authority. A highly competent group can be unrepresentative. A geographically varied group can misunderstand an operational mechanism. A crowded room can produce weak reasons. A nearly empty room can identify a serious defect. The institution should evaluate each property rather than treating attendance as a universal proxy.

The strongest claim a small expert session can make is often conditional: given the evidence heard, the entities identified these consequences and recommend this course, subject to the wider process, recorded objections and later review. That is not timid language. It accurately locates responsibility. It lets later decision-makers evaluate the result without inventing a popular mandate.

The room becomes dangerous when its convenience is converted into constitutional shorthand. Phrases such as “the community decided” can erase the difference between discussion, consensus within a chartered body, advice, staff implementation and Board action. If 47 people were physically present, the record should not imply that an undefined global community stood behind them. It should show which body acted, under what rule and with what opportunity for challenge.

ICANN85's programme multiplied choice and exclusion at once

The Community Forum structure exists to give Supporting Organisations and Advisory Committees time for internal work, cross-community interaction and sessions of broader interest. A programme of 276 sessions over six days creates extraordinary access to specialised work. It also creates unavoidable competition among rooms. No entity can observe everything. Concurrent sessions force choices, and those choices distribute attention.

Scheduling is therefore part of governance. A high-stakes issue placed against another high-stakes issue will divide relevant entities. A session at the edge of the day may be easier for one time zone and harder for another. A room allocation may signal expected importance before attendance occurs. Changes published late can advantage people already embedded in informal communication networks. The average headcount cannot reveal these effects by itself.

The programme also separates formal and informal power. The published session is visible, recorded and capable of later review. Corridor conversations, meals and private coordination are not. Those interactions can be productive: parties discover concerns, test compromises and avoid wasting plenary time. Yet they can also determine what reaches the microphone as an apparently mature position. A physical attendee has access to this layer in a way that a remote registrant often does not.

An audit should therefore distinguish agenda formation, formal deliberation and disposition. Who proposed the topic? Who chose its title, chair, speakers and duration? Which documents were available in advance? Were competing interpretations visible before the session? What changed in the room? What changed afterward? A headcount sits in the middle of that chain. It cannot explain the beginning or the end.

ICANN's meeting page preserves schedules, briefings and session materials, providing a basis for such reconstruction. The task is not to make every coffee conversation public. It is to ensure that a binding or highly consequential step can be explained through the public record rather than requiring trust in inaccessible social knowledge.

Averages conceal the political difference among rooms

An arithmetic mean flattens a programme. A packed plenary and several sparsely attended technical sessions can produce the same average as uniformly medium-sized rooms. The top-session table in the report partly restores distribution, but the political meaning of attendance still depends on session type.

Some sessions are informational. Some are internal working meetings. Some solicit public questions. Some prepare advice. Some test consensus. Some present decisions already made elsewhere. Counting each room as one observation may be appropriate for venue planning, but it is not appropriate for evaluating representative reach. A public plenary and a sign-up room do not carry the same institutional role merely because both occupy time and space.

The report itself distinguishes open sessions, closed sessions and sign-up rooms. That distinction should become part of governance analysis. Closed sessions may be legitimate where confidential personnel, legal or security matters require protection, yet their output should be traceable to an authorised body and followed by an adequate public explanation. Open sessions offer observer access, but openness does not establish that affected constituencies attended. Sign-up rooms enable initiative, but they may have no formal decision status at all.

A session audit should weight no one person's opinion by room type. Instead, it should classify the authority of the output. Was the room authorised to decide, to recommend, to gather information or merely to exchange views? Was any poll binding? Did a chair declare consensus, and under which rules? Was the outcome forwarded to another body? These questions prevent a large audience from giving false authority to a presentation and a small audience from obscuring a formally delegated decision.

Publication should avoid a single “average participation” claim when consequential sessions differ radically. A distribution by authority class, with medians and ranges, would tell readers much more. Privacy can be protected because the unit is the session, not an identified attendee. The result would remain descriptive, but at least it would describe the institution people are trying to assess.

The virtual entities missing from forty-seven

ICANN85 registered 678 people for virtual participation. The 47-person average explicitly concerns in-person attendance. Any reading that folds virtual entities into the average or treats their absence from the room as absence from the meeting is wrong. Yet simply adding virtual logins to physical counts would create another misleading total.

Remote participation has several levels. A registered person may not connect. A connection may remain open while its user is elsewhere. A entity may listen closely without speaking. Someone may use a shared room or reconnect through multiple devices. A contributor may enter only to make one carefully prepared intervention. Connection data must therefore be labelled as connections, unique authenticated accounts or active contributions rather than casually translated into people.

Influence differs too. Remote entities can hear proceedings, use chat and request the microphone, but they cannot reproduce every informal encounter. Their queue may be visible differently to the chair. Latency, audio failure or interpretation delay can make intervention harder. A physical room can react through body language that a remote speaker cannot see. When discussion runs late, people in distant time zones bear a cost invisible in the venue.

These are design questions, not arguments for abandoning hybrid meetings. Robust virtual access expands participation when travel, visas, disability, care obligations, cost or disruption prevents physical presence. The official retrospective specifically noted travel disruption and the importance of remote participation. The institutional question is whether the remote channel allows timely influence, not whether a video feed exists.

Useful measures include the proportion of remote questions acknowledged, response time by channel, speaking order, interventions carried into summaries, technical failures during decisive moments and whether remote contributors returned for later stages. Aggregate reporting can expose structural disadvantage without profiling individuals. The 47-person room should be analysed as one component of a hybrid deliberative space, not its entirety.

The chair controls more than the microphone

In a small room, chairing decisions are unusually visible and consequential. The chair recognises speakers, manages time, frames disagreements, decides when an answer is sufficient and summarises what the room appears to believe. Even when the chair has no substantive vote, these acts shape the record from which later authority may be drawn.

A raw count treats every occupant as equally present. In practice, one panellist may speak for twenty minutes, several established entities may exchange rapid interventions, and most people may remain silent. Silence can mean agreement, uncertainty, deference, fatigue, language difficulty or simple observation. It cannot safely be counted as consent.

The microphone queue is a better but incomplete measure. Who entered it, who withdrew, who was recognised and who received a substantive response? Did the chair alternate channels and perspectives? Were newcomers helped to frame a question, or did jargon consume their time? Did a late intervention reopen an issue or get deferred beyond the point of influence? Recordings and transcripts make much of this observable.

Chair summaries need special treatment. A sentence such as “there was broad support” should identify the evidence: a poll, interventions, a chartered consensus method or the chair's assessment. The record should preserve material objections and explain their disposition. A summary that turns a complex room into unanimous community voice is more constitutionally consequential than the headcount itself.

Training and common reporting standards can improve this without destroying discretion. Chairs need freedom to manage real conversations. They also need to show how that freedom was used when their assessment enters a decision chain. An annotated outcome note—questions considered, views expressed, unresolved evidence and next authority—would do more for legitimacy than a larger attendance number.

Speaking concentration is measurable without policing opinion

The institution can examine room concentration without assigning political value to identities. Transcripts can be used to calculate speaking time, number of interventions, reply networks and the share of discussion occupied by panellists, staff, chairs and floor entities. The aim is not to declare frequent speakers illegitimate. Some speak often because they know the issue and answer questions. The aim is to understand whether an apparently open discussion actually relied on a narrow conversational core.

Organisation-level context matters where it is voluntarily or officially disclosed. Ten speakers may come from ten independent organisations, from three related interests or from no organisation at all. A consultant may speak personally in one intervention and for a client in another. Staff may explain implementation without advocating an outcome. Any analysis must preserve these distinctions rather than guessing allegiance from an email domain or job title.

Concentration should trigger inquiry, not automatic discounting. If the same few entities dominate because other people lack information, the remedy may be earlier materials and onboarding. If language access is the barrier, interpretation and translated summaries matter. If the subject truly requires rare expertise, the institution should seek independent review rather than pretend the expertise is common. If time allocation favours panellists, the format can change.

The most useful indicator is not a diversity score attached to a decision. It is a map of where reasons came from and how they were tested. A claim raised by one person can be decisive if the evidence is strong. A position repeated by many people does not become technically correct through volume. Measurement should reveal the conditions of deliberation while leaving the substance open to reasoned assessment.

This is why meeting analytics should be paired with issue coding. Which interventions introduced new evidence, challenged an assumption, requested clarification or proposed language? Which received answers? Which appeared in the next document? The trace makes influence visible without turning a policy forum into a plebiscite.

Staff, Board and community presence need separate interpretation

The ICANN85 report notes that attendee data includes organisation staff, support staff, Board members and community members. This is sensible for operational reporting: everyone uses venue capacity and meeting services. It is limited public evidence for a claim about community participation because the roles have different institutional meanings.

Staff may present analysis, support a session, implement policy or answer operational questions. Board members may listen, deliberate or explain decisions. Community volunteers may hold formal offices, represent recognised structures or participate personally. Contractors may provide technical, language or logistical support. Observers may be learning. A room containing all these groups can be healthy, but its headcount should not be described as 47 independent community voices.

Role separation also protects staff. Employees should be able to contribute expertise without their attendance being interpreted as support for a policy position. Board members should be able to hear discussion without implying predetermination. Community members should know when an answer represents current institutional practice and when it is one entity's view.

Aggregate role categories can be reported for consequential sessions. The categories should allow multiple roles and changes over time. They should be based on self-declaration or official office, not inferred personal data. Small cells should be suppressed to prevent identification. The purpose is a sample description, not a roster of suspected interests.

This distinction helps a later reader reconstruct authority. If staff presented a proposal and most floor interventions questioned it, a bland attendance total conceals the event. If a community working group presented a recommendation after months of public work, the session may be reporting rather than originating consensus. Governance depends on these verbs: propose, explain, advise, decide and implement. Counting bodies without roles erases them.

Geography shows reach, not delegation

ICANN's retrospective celebrated participation from many countries and territories and all regions. Geographic reach is valuable. A meeting located in Mumbai can improve access for people in the surrounding region and expose the institution to experience often filtered through distant venues. The report's regional distribution helps readers see host effects and persistent imbalance.

But a country label does not create a delegation. An attendee from a country may represent a company, a government, a civil society organisation, a technical community or only themselves. They may work on networks spanning several jurisdictions. Another country may have one attendee whose perspective cannot encompass its operators, registrants, users and public authorities. Counting countries is therefore a breadth measure, not proof that national populations consented.

Regional statistics also require care because meeting location changes who can travel. A high host-region share may demonstrate reduced distance rather than durable representation. That is still a success if access was the goal. It becomes misleading only when presented as evidence that the region collectively authorised outcomes.

Longitudinal analysis can separate host effect from sustained engagement. Did first-time entities join later work? Did local operational examples enter documents? Did people contribute remotely after the meeting moved elsewhere? Did leadership pathways broaden? These questions test whether a meeting created capacity rather than a one-week photograph.

The relevant geography may also be issue-specific. A policy affecting internationalised domain names has language and script dimensions that global regional categories flatten. A registry-contract issue may concentrate operational exposure differently. A session report should identify the populations the issue actually touches instead of relying on five-region coverage as a universal answer.

First-time attendance is an opening, not an outcome

ICANN reported hundreds of first-time in-person attendees at ICANN85. New entry counters the tendency of specialist communities to become closed through habit rather than rule. It can bring local knowledge, expose unexplained assumptions and create a future volunteer base. Yet first-time presence is the start of an institutional relationship, not evidence of influence.

Newcomers face a vocabulary, a map of bodies, years of prior documents and dense personal networks. They may spend their first meeting learning where decisions occur. That is not failure. Expecting immediate authorship would be unrealistic. But an organisation claiming inclusive participation should measure whether entry becomes a usable path.

The path can be observed through aggregate retention: subsequent remote participation, mailing-list contribution, working-group involvement, public-comment submission, leadership applications and continued engagement with no formal role. The institution should also ask why people leave. Cost, time, hostility, employer withdrawal or repeated non-response can remove entities without appearing in attendance totals.

Onboarding should explain authority as well as acronyms. A newcomer needs to know whether a session is informational, where operative text lives, when comments matter and who must respond. Otherwise openness becomes a tour of visible meetings while consequential work remains socially legible only to regulars.

No one owes the institution permanent volunteer labour. Retention is not a demand that every newcomer become a professional entity. It is a diagnostic: among people who wanted to continue, could they find a route, and did their evidence receive fair treatment? That question respects agency while testing the institution.

The room's authority changes across the decision chain

One of the easiest mistakes in meeting coverage is to locate a decision in the most visible room rather than in the body authorised to make it. A session may discuss an issue whose formal decision lies with a Supporting Organisation council, Advisory Committee, Board, member vote or staff implementation function. The room may narrow options without formally selecting one. It may ratify work completed elsewhere. It may merely inform.

Every consequential session should therefore carry an authority label in its public materials: purpose, chartered body, decision status, next step and review route. If the session can assess consensus, the applicable method should be linked. If it cannot decide, the record should avoid verbs implying that it did.

Decision-chain mapping also reveals handoffs. A community recommendation can change during legal review, Board consideration or implementation. Those changes may be justified, but they should not be retroactively attributed to the meeting. Conversely, a room's intervention may be decisive even if the final act occurs months later. Traceability gives credit and responsibility accurately.

The authority statement should include who may challenge the next step. An absent affected party may learn about the issue only after the meeting. If the room's output enters a later public-comment period, that is an opportunity for correction. If no later route exists, the burden on notice and participation at the session is higher.

This approach permits small-room work without constitutional fiction. Forty-seven people can do important work when everyone understands what the work is, how it travels and where it can be contested. Ambiguity, not smallness alone, creates the legitimacy risk.

Polls do not turn observers into a sampled public

Meeting polls are useful for sensing a room, choosing agenda time or discovering whether language is understood. They are weak evidence of wider support unless the eligible population and sampling method are defined. A poll among self-selected session occupants does not become representative because a percentage is displayed.

The report should preserve the exact question, response options, number responding, channel and time. It should note whether staff, directors and panellists could vote, whether remote entities were included and whether multiple responses were controlled. These details do not make the poll statistically representative. They make it honestly interpretable.

Chairs should avoid treating abstention as agreement. People may decline because the question is unclear, they lack authority to state an organisational view or they have just entered the room. A binary question can conceal conditional positions. Public display can create conformity pressure even when voting is technically anonymous.

For formal consensus processes, the poll's role must be stated. It may be one signal alongside reasoned objections, mailing-list discussion and charter criteria. If a single substantiated objection can expose a defect, majority volume should not erase it. If a body uses a formal vote, eligibility and mandate should be documented separately.

A session count and a poll result are therefore two samples layered together: who happened to be in the room, and who among them responded. Publishing 80 percent without both denominators manufactures certainty. The remedy is not to ban polls. It is to label them as instruments of facilitation rather than miniature referendums unless a genuine electorate exists.

Transcripts are evidence, but not the whole event

ICANN's schedule infrastructure links recordings and transcripts after sessions. This creates a valuable public memory. A person unable to attend can inspect reasons, identify speakers and compare a summary with what was said. Researchers can test claims about concentration and issue treatment. Entities can correct the record.

Transcripts also have limits. Automated or live text can mishear names and technical terms. Side conversations and visual reactions disappear. Slides may carry qualifications not spoken aloud. A recording starts and stops at formal boundaries, missing preparation and aftermath. The most important document may be revised later without a clear link to the intervention that caused the change.

The answer is a layered record. Preserve transcript, recording, presentation, attendance methodology, chair summary, action items and later disposition. Link versions of operative documents. Allow corrections without silently replacing the original. Identify when a session contains no decision rather than leaving readers to infer one.

This record should remain usable. Hundreds of hours of video are transparency in a formal sense but impose high search costs. Time-stamped issue indexes and concise outcome notes can reduce that burden. Accessibility requires captions, stable links and formats that do not depend on a proprietary interface.

Reason-giving is the bridge between the room and the absent public. People who were not present cannot recreate every social cue, but they can evaluate an argument if evidence and objections are preserved. The stronger the public record, the less legitimacy depends on accepting the room's self-description.

A minimum room-level participation record

ICANN could publish a compact participation card for sessions that feed policy, advice or governance decisions. The card would not identify ordinary attendees. It would state session type, authority, count method, room capacity, physical mid-session count, unique remote accounts present during a defined interval, number of speakers by channel, declared institutional roles at an aggregate level, polls used and material access incidents.

A second section would describe process: publication date, supporting materials, schedule changes, interpretation and captioning, chair, speakers, conflicts disclosed, exact output and next decision point. A third would list substantive effects: questions answered, evidence requested, objections preserved, text changes proposed and matters deferred.

Privacy rules are essential. Aggregate affiliation should not expose a lone entity from a small country or sensitive employer. Anonymous and pseudonymous contributions may be necessary for safety. Raw platform logs should have limited retention and access. Public analysis should suppress small groups and avoid joining records across meetings merely to profile individuals.

The card should include uncertainty. A manual snapshot misses movement. Remote-account counts may include idle connections. Affiliation information may be incomplete. Speaking time does not equal influence. Declaring these limits increases trust because readers know which inferences remain unsafe.

Most importantly, the card would state what the session cannot claim. It might say that attendance was self-selected, the session was advisory and no statistical inference to users or countries is intended. Such language does not weaken the outcome. It prevents communications material from expanding the mandate after the fact.

An influence trace should follow the meeting

Participation reporting usually ends when the venue closes. Governance analysis should continue until the room's output reaches its institutional destination. If a concern led to new text, link the version. If staff rejected a suggestion, publish the reason. If a council or Board chose another course, identify the decision and response.

An influence trace can classify outcomes without pretending causation is simple: adopted, partly adopted, prompted further study, addressed independently, deferred, rejected with reasons or unanswered. Entities should be able to challenge misclassification. The classification concerns public interventions, not private attribution.

Timing matters. A technically sound answer delivered after implementation may not repair lost influence. The trace should record response and disposition dates. It should also show whether an issue returns at a later meeting, preventing repeated discussion from being mistaken for resolution.

This is where small-room legitimacy can become strong. A session with modest attendance that identifies a material problem, receives a reasoned institutional response and changes an outcome has demonstrated value. A large session that produces applause but no trace may have less governance significance. Influence evidence is more informative than spectacle.

The trace also disciplines claims of consultation. An institution should not cite a session as community engagement if it cannot show what the engagement did. Consultation may legitimately confirm an existing view, but the record should explain how conflicting evidence was considered. The right to speak is incomplete when the recipient has no duty to account for listening.

Meeting analytics should be independently reproducible

ICANN's data-report page explains that meeting data supports improvement and that collection has become more standardised. It also warns that historical manual headcounts can be affected by collection and entry practices. This candour should be extended into reproducible methods for current reports.

For each metric, publish the unit, collection moment, inclusion rule, duplicate treatment, missing-data rule and known change from previous meetings. “Attendee,” “entity,” “connection” and “session attendance” should not shift meaning between charts. If a platform changes, the series should disclose the break.

The session average should show its denominator. Were all 276 sessions counted, or only rooms with available snapshots? How were closed and sign-up rooms treated? Was one snapshot used per session? The public report gives important clues, but a concise methods appendix and machine-readable aggregate table would allow independent checks.

Reproducibility does not require releasing personal attendance logs. Session-level aggregates, schedules and methods are enough for many questions. Access to sensitive raw material can remain tightly controlled, with an independent auditor checking privacy and accuracy.

An advisory group including meeting operations, privacy expertise, community entities and accessibility specialists could review the method periodically. Its job would be measurement integrity, not judging political outcomes. The result would let ICANN improve meetings while preventing operational statistics from becoming unexamined legitimacy claims.

Comparisons across meetings need stable definitions

ICANN85 took place in a particular city, season and global travel environment. Its host region, programme and disruptions shaped attendance. Comparing its 47-person average with another meeting can be informative only if collection methods, session types and programme density are comparable.

A meeting with fewer parallel sessions may have higher average room attendance without broader participation. A Policy Forum and a Community Forum have different structures. Virtual-only meetings produce no physical room average. A host-region surge can alter total registration while the core of repeat entities remains stable. Simple rankings reward conditions rather than inclusion.

Longitudinal reporting should separate total unique attendance, session concentration, newcomer entry, remote engagement, role mix and issue-specific influence. Stable definitions should survive platform changes. When they cannot, old and new series should not be spliced without warning.

Comparisons should also resist a false race toward larger rooms. Some work benefits from small groups. The objective is not to maximise every headcount but to fit participation design to authority and affected interests. A confidential safety discussion may appropriately be small; a region-wide policy consultation may require wider evidence even if its final drafting team is small.

The best benchmark is whether participation limits were recognised and mitigated. Did an under-attended consequential session trigger additional notice? Did a remote failure lead to extension? Did a host effect produce sustained contribution? These are governance outcomes, not event-marketing totals.

Absent interests require procedural representation

No meeting can contain everyone affected. Some users will not know ICANN exists. Small businesses may lack staff time. People facing language, disability, visa or connectivity barriers may be unable to attend. Future registrants cannot participate yet. Their absence does not mean their interests are unknowable or that present entities may freely claim them.

Institutions can represent absent interests procedurally. Impact assessments can identify affected groups. Chairs can ask proponents to address predictable burdens. Independent reviewers can test consumer, competition, accessibility and security effects. Public-comment periods can remain open long enough for intermediary organisations to consult. Post-implementation review can detect consequences the room missed.

These safeguards are not substitutes for direct participation where it is possible. They are protections against equating silence with acceptance. A session summary should identify materially affected groups not heard and explain how their interests will be examined before final action.

The burden rises with irreversibility. A reversible pilot with monitoring can proceed on narrower evidence than a permanent change imposing high switching costs. Emergency action may require speed, but it should carry sunset and review. Small-room judgment becomes more legitimate when its uncertainty changes the design of the decision.

This approach respects both expertise and the absent public. It does not demand that every room mirror humanity. It demands that the institution know the limits of its sample and build remedies into the authority exercised.

Communications language should stop laundering scale

Public institutions naturally celebrate reach. Numbers of attendees, countries, sessions and hours demonstrate operational achievement and help justify investment. Problems arise when those figures migrate into claims that “the global community” produced or endorsed a substantive outcome.

Communications should use precise nouns. ICANN85 registered 2,195 attendees. A particular chartered body reached a decision. A session heard specified interventions. The Board adopted a resolution. Staff implemented it. Precision makes institutional complexity legible and allocates accountability.

The phrase “multistakeholder community” can describe an institutional ecosystem without implying a census. When used as an actor—“the community decided”—it should identify the mechanism. Was there a vote, rough consensus, advice from several bodies or a Board assessment after consultation? Readers should not have to infer authority from branding.

The 47-person figure is a useful test. If a sentence sounds uncomfortable when rewritten as “the average physical session room of 47 decided,” that discomfort may reveal an overbroad claim. The remedy is not always to deny the decision. It is to name the authorised body and its reasons.

Honest communications can still celebrate expertise, first-time access, regional reach and robust hybrid services. Those achievements matter. They become more credible when they are not asked to perform the unrelated work of proving popular sovereignty.

A practical audit of ICANN85

A focused audit could begin with the 276-session programme and classify each session by purpose and authority. It would attach available physical snapshots, remote participation aggregates, room capacity, schedule conflicts and material access services. It would then sample consequential sessions for transcript coding and decision tracing.

The audit would not seek one legitimacy score. It would identify patterns: sessions where speaker concentration was high; outputs whose authority was unclear; remote interventions that received delayed responses; issue areas with strong traceability; and sessions where absent affected interests prompted safeguards. Qualitative review would prevent numerical patterns from being misread.

An independent reviewer should test the published methods and invite corrections. Community bodies should be able to explain context. Privacy officers should prevent re-identification. Findings should connect to specific actions—schedule design, chair guidance, remote queue handling, records, notice or later review—rather than becoming another report with no owner.

The audit should also record success. A small session may demonstrate excellent preparation, balanced facilitation, clear authority and visible influence. Publishing strong examples gives other chairs a practical standard and avoids treating measurement as accusation.

ICANN85 is especially useful because the official report already supplies the essential factual boundary: 47 was an average in-person session count, while the meeting involved thousands across two participation modes. The audit's task is to preserve that boundary through the entire decision narrative.

Conclusion: forty-seven can deliberate, but cannot become everyone

The image of 47 people in a room is neither a scandal nor a constitution. It is the ordinary human scale of specialised governance. People gather, compare experience, argue over language and sometimes improve decisions that matter far beyond the venue. Their work deserves respect when it is competent, open to challenge and tied to an authorised process.

The same image warns against rhetorical inflation. A manual mid-session count does not reveal unique participation, representation, mandate or influence. It excludes virtual entities from the physical average, combines institutional roles and sits inside a programme whose parallel rooms distribute attention. Total meeting registration cannot be poured into each session to enlarge its authority, and a session average cannot be used to shrink the meeting to 47 people.

Legitimacy comes from matching claim to mechanism. Publish who could participate, what was counted, what the room could do, how speech and reasons were handled, which interests were absent, where the output travelled and how it could be challenged. Protect privacy while making methods reproducible. Measure influence after the meeting rather than declaring success at the door.

Under that standard, a room of 47 may produce excellent advice or a defensible decision within a delegated body. It simply does not become the Internet public. ICANN's strongest response to the gap is not to hide the room or romanticise it, but to show, step by step, why its work deserved to travel further than its walls.

Sources