Summary

  • APNIC60's official preliminary report recorded 451 in-person attendees, 60 online-only entities, 40 economies and 127 APNIC member organisations represented. Each figure measures a different unit and none alone establishes regional consent.
  • The later figure of around 26,000 members including National Internet Registry members is not an aligned denominator for the 127 represented APNIC member organisations. Direct membership, NIR-served membership, account authority and conference affiliation must first be reconciled.
  • Policy legitimacy should be examined through several lenses: unique organisations, active networks and ASNs, market and dependency exposure, economy and subregion, organisation type, proposal-specific burden, contribution and influence.
  • APNIC can publish a privacy-preserving participation account that connects notice, attendance, speaking, consensus evidence and post-meeting disposition while preserving open participation and refusing to turn network size into voting power.

Four clean numbers, four different stories

The APNIC60 report opens with unusually useful precision. The meeting in Da Nang, Viet Nam, from 4 to 11 September 2025 recorded 451 people attending in person, 60 participating online only, 40 economies represented and 127 APNIC member organisations represented. These figures are preliminary conference statistics, not interchangeable descriptions of a single public.

People are not organisations. Organisations are not economies. Economies are not networks. Registrations are not policy interventions. The four numbers answer operational questions: how many people used each attendance mode, how broad the geographic reach appeared, and how many organisations in a defined institutional relationship had at least one person associated with the event. They do not tell us how many unique people entered the Open Policy Meeting, how many member organisations authorised a policy position, or how much of the region's routed infrastructure was represented.

That distinction matters because conference reporting often moves from reach to legitimacy without an explicit step. A report correctly says that 127 member organisations were represented. A later account may describe “the APNIC community” reaching consensus. Readers can then infer that the represented organisations authorised the result or that the participating room approximated the region. Neither inference follows from the attendance line.

The number 127 still matters. It gives a firmer organisational numerator than a list of countries or a total badge count. It can be compared with future meetings if the definition remains stable. It can show whether outreach changes organisational breadth. It can help identify concentration. The aim is not to discard it but to give it the denominator and process context needed to carry more than promotional weight.

The “rest of the region” in the title is not one silent bloc. It includes direct members, networks served through National Internet Registries, prospective entrants, customers, governments, researchers and users whose exposure differs by proposal. A credible participation account must resist compressing all absence into one imagined interest.

The 26,000 figure is not the missing denominator

APNIC's later summary of its 2026 annual general meeting reported that membership grew 1.6 percent in 2025 and that overall membership, including National Internet Registry members, was around 26,000. It is tempting to divide 127 by 26,000 and announce a participation rate. That calculation would be false precision.

The numerator is “APNIC Member organizations represented” at a conference. The larger figure includes NIR members and may describe a broader membership population with different registry relationships. One organisation may have several accounts or network operations. A person may attend without being the voting or corporate contact. An NIR-served organisation may appear in conference records under an affiliation not coded as a direct APNIC member organisation. The unit and coverage are not aligned.

Before any ratio is published, APNIC would need a crosswalk: direct member organisations eligible for the numerator, NIR-served organisations, duplicate legal entities, related companies and the date at which membership is measured. It would need rules for a entity who lists a parent, subsidiary, university or government department differently from the membership record. It would also need to distinguish mere affiliation from authority to speak for the account holder.

An aligned organisational coverage measure could be useful. The denominator might be active direct member legal entities at the event cut-off date; the numerator, those with at least one verified affiliated entity. A separate NIR measure could use definitions agreed with each NIR. Both should carry missing-data and duplicate-treatment notes.

Even then, coverage would show presence, not consent. A represented organisation may have sent an engineer for training who never entered policy discussion. Several employees may disagree. A entity may speak personally. The denominator solves one measurement problem; it does not manufacture a mandate.

APNIC60 was more than its Open Policy Meeting

APNIC60 ran for eight days and combined workshops, technical sessions, security discussions, Special Interest Groups, networking, elections and the members' meeting. The report identifies the Open Policy Meeting as one part of that programme. Treating all 511 people across in-person and online-only modes as policy entities would erase the reasons they attended.

Training is valuable. Technical exchange improves operations. Social contact can create trust that later supports incident response and policy work. A conference can succeed at these functions even if only a subset participates in rulemaking. The mistake comes when the widest event population is used as the sample for a narrower policy outcome.

A participation account should follow people through stages without publishing personal trails. How many unique accounts accessed policy materials? How many attended the relevant session? How many subscribed to or posted on the Policy SIG list during the proposal period? How many interventions introduced evidence? How many distinct organisations were associated with those stages? Aggregate funnels can answer these questions while keeping ordinary attendance private.

Programme design affects the funnel. A policy session scheduled against technical content may lose operators most affected by a proposal. A workshop entity may not know that policy is open to them. Online-only attendees may join a technical stream but miss a consensus call. The conference total hides these choices.

APNIC should therefore celebrate conference reach in its own terms and report policy reach separately. Separating them strengthens both claims. The conference can be broad without pretending every attendee governed, and the policy process can be evaluated on its actual notice, evidence and consensus rather than borrowing the workshop audience.

Six proposals reveal why policy-specific denominators matter

The report says six proposals were discussed at the Open Policy Meeting. Three reached consensus, while three returned to the mailing list for further discussion. This is more informative than a single claim that the meeting made policy. It shows discrimination: the process did not convert every proposal into an outcome.

Yet each proposal activates a different affected population. A proposal concerning directory privacy touches registered contacts, abuse reporters, operators, researchers and people exposed by publication. A proposal involving RPKI service continuity affects certificate holders and routing security dependencies. Directory usage statistics implicate service operation and transparency. IPv6 allocation boundaries concern applicants and routing practice. Transition resources for IPv6-only networks engage a different deployment population.

One conference-wide denominator cannot describe these constituencies. The relevant question is not merely how many organisations attended APNIC60, but whether evidence from the organisations and people bearing each proposal's principal costs and risks entered before consensus. A small specialised sample may be appropriate if expertise is rare, provided missing interests are identified and later review exists.

Proposal records should carry an affected-population map prepared early enough to shape outreach. It would identify current resource holders, likely applicants, implementing networks, downstream dependants, public-interest effects and counterpart regions where applicable. It would not assign votes. It would show chairs which silences require active testing.

When three proposals returned to the list, the record should show what evidence was missing or contested and whom further discussion sought to reach. A return is not failure. It is evidence that the process preserved uncertainty rather than laundering a room reaction into finality. That restraint deserves as much attention as the three consensuses.

Organisations are a better unit than badges—and still incomplete

Counting unique organisations corrects one distortion in conference totals. A large operator may send a team, while a small network sends one person. A badge count can make the larger organisation appear as several independent interests. Organisational deduplication reveals concentration.

But legal entities are not automatically independent interests. Subsidiaries may share control. Associations may aggregate many members. Consultants may advise several clients. Universities and government departments may have decentralised operations. An employee can speak personally rather than under instruction. The participation record needs layers rather than one deduplication rule.

A useful hierarchy would show unique people, declared affiliations, ultimate organisational groups where publicly verifiable, and claimed mandates. “Claimed” matters: an affiliation answers who employs or hosts a person, not whom the person represents in that intervention. A entity should be able to mark a statement as personal, organisational, client-authorised or made in a formal community role.

Privacy and safety require restraint. APNIC should not infer hidden ownership, employment or opinion. Public aggregates can suppress small cells. Sensitive client relationships may be disclosed confidentially to a conflict function without public naming. The aim is to prevent ten badges from becoming ten constituencies, not to construct dossiers.

Organisation counts also risk privileging corporate form. Individual experts and civil society contributors can provide decisive evidence without a member account. Open policy deliberately permits participation beyond membership. Reports should present organisational coverage as one view, then separately describe non-member contribution rather than treating it as noise.

Counting ASNs brings operations closer, but not all the way

Autonomous System Numbers offer an appealing operational denominator. They identify networks participating in interdomain routing and can help estimate whether policy discussions include the operators who implement number-resource decisions. Compared with conference passes, an ASN count is closer to the infrastructure at stake.

It is not a constituency roll. One organisation may operate many ASNs for history, geography or network design. Another may operate critical infrastructure under one. Some resource holders do not originate an ASN. Not every assigned ASN is visible or active. A network's customer dependencies can be much larger than its ASN count suggests. Weighting views by ASNs would reward administrative topology rather than public mandate.

ASN linkage is best used diagnostically. APNIC could report, at aggregate level, how many active routed networks were associated with participating organisations, the distribution across subregions and organisation types, and concentration among the largest organisational groups. It could identify proposal-specific gaps: for a routing-security issue, were small edge networks and large transit providers both heard?

Methods must be time-bound. Routing tables change. ASN-to-organisation mapping is uncertain and can conflict with membership records. NIR arrangements complicate attribution. A published data dictionary should explain active-status windows, related-entity treatment and missing links.

The result would not say that a policy was legitimate because a percentage of ASNs appeared. It would say whether the deliberation drew evidence from a broad or narrow part of the operating surface. Chairs could use that information to adjust confidence, extend discussion or require post-implementation review.

Customers and dependencies resist simple counting

Network governance affects people through dependency chains. A small wholesale provider may support many local networks. A large consumer operator may connect millions of users. A university network supports research and public services beyond subscriber counts. Cloud, content, exchange and transit relationships spread consequences across borders.

Customer weighting therefore looks attractive but quickly becomes speculative. Subscriber figures use different definitions and reporting dates. Wholesale and retail customers cannot be added. Market share can hide resilience importance. A policy position does not become correct because its proponent serves more users.

Dependency evidence should inform impact, not allocate authority. For each proposal, entities can identify which services, routes, registration functions or customer groups rely on the affected mechanism. Staff analysis can present ranges and uncertainty. Independent evidence can test claims made by large operators.

This approach prevents two errors. It avoids treating every member organisation as equally exposed when burdens differ sharply. It also avoids converting commercial scale into votes. A small network can reveal a design flaw that harms an entire class of operators; a large network can provide implementation data without receiving constitutional priority.

The public record should distinguish affected volume from representational mandate. “This organisation supplies evidence drawn from a large deployment” is legitimate. “This organisation speaks for every user behind that deployment” requires authorisation that customer relationships do not provide.

Forty economies measure geography, not national voice

APNIC60's 40 represented economies demonstrate geographic reach across a region of exceptional scale and diversity. The host location in Viet Nam likely altered travel accessibility and regional composition. Economy coverage can reveal repeated absence and guide venue, fellowship, language and outreach choices.

An economy label does not mean the attendee represents its government, operators or population. One person may be affiliated with a multinational. Several organisations from one economy may hold opposed views. Networks often operate across borders. Territories and economies vary dramatically in population, market structure and registry arrangements.

Economy counts also conceal within-country concentration. Attendance from a capital-based incumbent, ministry and research network differs from one corporate delegation, even if both cases add one to the headline. A defensible report would show organisation types and subregions at safe aggregation, plus the host-economy effect.

Longitudinal data can ask whether geographic entry persists. Did organisations from newly represented economies participate on lists after the conference? Did operational examples enter policy documents? Were remote sessions scheduled accessibly once the meeting moved? Geographic inclusion is stronger when it changes evidence, not merely the map.

The institution should avoid using country diversity as a proxy for consent. It is evidence that access reached places. Authority still depends on the role of entities, the process used and the reasons supporting the outcome.

National Internet Registries require their own participation account

NIRs are central to understanding the gap between direct APNIC membership and the wider membership figure. They create additional relationships among APNIC, national registry arrangements and organisations receiving resources or services. Collapsing these relationships into one denominator obscures where participation rights, fees, contacts and policy communication sit.

An NIR-served organisation may receive notice through a national channel, through APNIC or both. Language and local operational support can improve access. At the same time, an organisation may be counted differently in regional and national records. The path by which its views reach a regional policy discussion may be less visible than a direct member's.

APNIC and NIRs could publish aligned aggregate participation tables without exposing account details. For each arrangement: eligible or active organisations under a stable definition, organisations reached by notice, conference affiliations, policy contributors and known missing data. Methods should be co-designed rather than imposed, because records and legal contexts differ.

The report should also clarify mandate. An NIR representative may explain institutional practice without speaking for every served organisation. An NIR-served member attending directly may speak personally or for its network. Neither channel should erase the other.

This separate account would make the “around 26,000” figure analytically useful while preventing a false 127-over-26,000 ratio. It would also expose whether regional consultations systematically hear direct members more clearly than NIR-served networks.

Elections at the same conference use another denominator

APNIC60 hosted an NRO Number Council election and several SIG leadership selections. Election participation has a formal electorate or selection mechanism different from an open policy room. Conference presence may permit voting or candidate contact, but it should not define electoral legitimacy by itself.

For each election, the institution should report eligible voters, ballots or consensus entities, turnout, invalid ballots, candidate access and relevant organisational concentration. Those figures must not be merged with 451 in-person attendees or 127 represented member organisations unless the rules make them the electorate.

Leadership selection also changes later influence. Chairs organise agendas, facilitate discussion and assess consensus. Their mandate should be traceable to the selection process, term and removal rules. A person acclaimed in a room may be entirely legitimate under published rules, but the record should not imply a region-wide popular election.

The coexistence of elections, policy and training at one event makes precise verbs essential. Attendees learned, discussed, voted, advised and observed in different combinations. “APNIC60 decided” is too broad unless the specific body and mechanism are named.

Membership accountability improves when these denominators remain separate. Members can evaluate electoral participation, the wider community can evaluate open policy access, and everyone can see where authority transfers from one process to another.

Consensus needs a sample description, not a vote count

APNIC's Policy Development Process and SIG Guidelines place open discussion and consensus at the centre of policy work. Consensus is not simple majority voting. It requires chairs to assess whether objections have been addressed and whether a proposal is acceptable to entities.

That method can value reasons over numbers. One operational objection may reveal a serious defect even if many people initially support a proposal. Returning three APNIC60 proposals to the list illustrates that a meeting reaction need not close discussion.

Yet rough or community consensus still needs a sample description. How many unique people and organisations contributed? Did support come from distinct operational contexts? Which affected groups were absent? Were remote objections heard? What material concerns remained? A chair can answer these questions without converting consensus into arithmetic.

The declaration should distinguish silence from support. People may observe, lack sufficient evidence or avoid speaking against an employer or established group. A room poll can inform a chair but should be published with its exact question, respondents and channel. Mailing-list evidence should be integrated, not treated as secondary to visible reactions.

When participation is narrow, the chair may still find consensus among entities. The outcome note should limit the claim and attach safeguards: further notice, implementation monitoring, review or a return point if missing evidence emerges. This is calibrated authority, not paralysis.

Notice is the first denominator

No participation measure begins at the registration desk. The first population is people and organisations given a realistic chance to know that a proposal matters. Notice must arrive through channels they use, in time to understand the mechanism and consult colleagues.

Mailing-list publication is essential but may reach a professional core more reliably than small operators. Conference promotion may attract technical attendees without explaining which policies affect them. NIR, network operator group and industry-association channels can broaden reach, provided their role is transparent.

For each proposal, APNIC could publish a notice ledger: dates, languages, channels, audience size where responsibly measurable, material versions and response windows. The ledger should not claim that delivery equals reading. It would show what the institution did to reach the relevant denominator.

Notice quality matters as much as volume. A title may obscure operational consequences. A plain-language mechanism summary and affected-population note can help people decide whether to engage. Exact proposal text and staff analysis must remain available for scrutiny.

If evidence from a materially affected group is absent, targeted outreach should ask for operational cases, not endorsements. The objective is to improve the record rather than recruit supportive voices. Publication of the outreach and responses protects against selective consultation.

Attendance, contribution and influence are three separate ledgers

An organisation can be present without contributing, contribute without attending in person and influence an outcome through one concise piece of evidence. Reporting should therefore preserve three ledgers.

The attendance ledger uses privacy-preserving event and session aggregates. The contribution ledger records public messages, microphone interventions, submitted evidence and proposal authorship, with declared role context where relevant. The influence ledger links issues to revisions, chair responses, implementation analysis and later review.

The ledgers should not become rankings of volunteers. Frequent contribution may reflect service, not domination. Quiet entities may learn or support implementation. Influence coding requires human review because not every accepted edit is consequential and not every rejected comment was ignored.

Together, however, the ledgers expose institutional patterns. One organisation may send many people but rarely shape policy. A small group may write most proposals and hold leadership roles. Fellows may attend broadly but lack pathways into drafting. NIR-served networks may appear geographically but not in issue resolution.

Annual reporting can aggregate these patterns and connect them to action. If first-time organisations do not return, improve onboarding. If remote contributions receive fewer responses, change chair practice. If proposal authorship is concentrated, build capacity and solicit problems from missing operator types.

Measuring organisation type without freezing identity

The Asia Pacific Internet ecosystem includes access providers, mobile operators, hosting and cloud services, exchanges, universities, research networks, governments, registries, vendors, civil society and individual experts. Organisation type helps show whether a proposal heard the operational contexts it affects.

Categories can mislead. A company may operate networks and sell software. A university may serve as an ISP. A government body may run infrastructure. Entities should self-select multiple roles, and reports should publish the classification method and allow correction.

Type is not a proxy for policy position. Two small ISPs can disagree. Civil society does not speak for all users. Government participation does not represent every citizen. The measure reveals the evidence base, not a voting bloc.

Small categories need suppression to protect identities. Historical comparisons should note taxonomy changes. APNIC should not derive sensitive categories from names or public profiles. Voluntary disclosure and broad operational groupings are sufficient.

The useful question is proposal-specific: did the discussion include the types of organisation expected to implement, pay for, rely on or be constrained by the change? If not, what substitute evidence and review were used? That inquiry connects diversity to mechanism rather than display.

Market concentration must not become vote weighting

Participation by large networks can dominate because they have policy staff, travel budgets and extensive deployment evidence. Their knowledge is important. Their commercial scale does not entitle them to define public consensus.

APNIC can report organisational concentration using bands rather than public rankings: share of contributors associated with the largest groups, repeat leadership roles and proposal authorship. It can compare those patterns with the broader membership distribution. The result should prompt outreach and conflict management, not discount an argument solely by source.

Small networks need protection from participation cost. Meeting time, English drafting, list culture and technical analysis all impose burdens. Remote access and fellowships help, but proposal mentoring, impact templates and staff-supported evidence calls may matter more.

No single weighting solves the legitimacy problem. One organisation, one vote ignores operational diversity and legal structures. One ASN, one vote rewards topology. Customer weighting rewards incumbency. Headcount rewards employers able to send teams. Consensus should remain reason-sensitive while the institution publishes concentration and safeguards missing interests.

Transparency is the constraint. A chair can say that support was voiced by many people associated with fewer organisational groups and still judge the evidence on its merits. Readers can then distinguish breadth from repetition.

Future entrants are absent by definition

Number-resource policy affects organisations that do not yet hold resources or membership. Start-ups, community networks and future operators cannot appear in a current-member denominator. Scarcity and transfer rules may determine whether they ever enter.

An institution accountable only to current holders risks treating incumbency as constituency. Open participation partly addresses this, but future entrants are difficult to identify and may lack expertise to follow policy before they need it.

Impact analysis should therefore include an entrant test: application burden, cost, availability, documentation, transition and appeal. Business formation and network-development evidence can supply ranges without pretending to count an unknown population. Small-operator associations and technical training programmes can contribute cases.

Policies with uncertain entrant effects should include review after sufficient applications occur. Public metrics can compare processing, rejection and burden by organisation size without exposing applicants. Sunset or revision triggers preserve adaptability.

The absence of future entrants is not a reason to halt policy. It is a reason not to call current conference attendance the complete regional mandate. Procedural protection carries interests that no badge list can record.

Language and time determine who can use openness

Asia Pacific's linguistic and time-zone diversity makes formal openness an incomplete measure. English materials and rapid discussion favour entities who can process specialised terms publicly. Online attendance removes travel cost but may place decisive sessions at difficult hours.

Translation of announcements improves discovery; translation of mechanism summaries and material revisions improves deliberation. The timing of each edition matters. A summary published after discussion cannot restore agenda-setting power.

Participation reporting should identify source language, translated materials, publication times, interpretation use and contributions received through supported languages. It should not assume that a person from an economy prefers a particular language.

Chairs need procedures for translated intervention: confirmation of technical meaning, equivalent response and extensions when institutional delay leaves too little time. Original and translated versions should remain linked with correction history.

Language evidence belongs beside the 40-economy figure. A map may show geographic breadth while the discussion remains linguistically narrow. Reporting both conditions makes investment choices visible and gives members a basis for accountability.

A regional participation matrix

APNIC could replace the hunt for one definitive percentage with a matrix. Rows would represent policy proposals or institutional decisions. Columns would cover notice, direct and NIR-served organisational reach, active-network evidence, economy and subregion, organisation type, participation mode, first-time contribution, speaking concentration, unresolved absence and post-decision review.

Each cell would contain a bounded measure or a candid “not known,” plus method notes. No row would collapse into a legitimacy score. Chairs and members could see where confidence is strong and where safeguards are needed.

The matrix should distinguish counts from rates and rates from weights. It should mark misaligned denominators rather than calculating them. It should identify whether one organisation is associated with several entities and whether participation came through formal mandate or personal contribution, where declared.

Public data would be aggregate. Independent reviewers could inspect protected linkage methods. Entities should be able to correct affiliation without changing the historical substance of comments. Retention limits would prevent conference analytics from becoming permanent personal tracking.

Over time, the matrix would show whether interventions work. Does rotating location change policy authorship? Does translation reduce delay? Do fellows return as contributors? Do NIR channels produce issue-specific evidence? The purpose is institutional learning constrained by accountability.

What the APNIC60 report can legitimately support

The preliminary report supports several strong claims. APNIC60 brought hundreds of people together in person and online. It reached entities associated with 40 economies and 127 APNIC member organisations. It hosted open technical and policy work, elections and a members' meeting. The Open Policy Meeting distinguished among six proposals rather than approving all of them.

It cannot, from those statistics alone, support a claim that 127 organisations authorised each policy result, that 40 economies consented, that the attendees represented the region's users, or that conference attendance covered a known share of all APNIC and NIR-served membership. Those require aligned records and evidence of role and influence.

This is not a criticism of the report for failing to be something it was not. Event reports serve operational and community communication. The governance task is to stop later rhetoric from expanding their numbers beyond their definitions.

APNIC's 2025 Annual Report provides broader institutional context. Membership, services and participation should be analysed across those records with careful units. The public should be able to reproduce any claimed ratio.

The best institutional sentence may be modest: the meeting demonstrated substantial but self-selected reach, and proposal legitimacy rests on the open process, reasons, recorded consensus assessment and review—not on conference statistics as a regional plebiscite.

The numerator can survive stricter interpretation

The 127 organisations should remain in APNIC's public record. It is more informative than a photograph and more disciplined than an undefined claim of community. It can become a valuable baseline if its definition remains stable and its limits remain visible.

The next step is not to find one enormous denominator and declare coverage. It is to align several denominators with the questions they answer. Direct organisations show institutional reach. NIR-served membership requires a separate account. ASNs illuminate operating networks but not votes. Economies show geography but not national mandate. Customer and dependency evidence informs impact without allocating authority. Proposal-specific populations reveal whose experience was needed.

Attendance then needs connection to contribution and influence. Notice should be recorded before the meeting; chair reasoning and missing interests during it; revisions, implementation and review afterward. Elections should use their formal electorates. Open policy should preserve reasoned consensus and describe its sample. Privacy should constrain every linkage.

APNIC60's figures show a real regional gathering doing valuable work. They also reveal the scale of what remains unmeasured. By publishing a multi-denominator participation account, APNIC would not weaken its community. It would let 127 organisations mean exactly what the evidence supports—and make the interests beyond the room visible without pretending they are one voice.

Conclusion: make denominator disputes resolvable

A participation account will earn trust only if an outside reader can reproduce its categories and challenge mistakes. APNIC should publish a compact methods note beside each annual series: the date of the membership snapshot, the meaning of “organisation represented,” the treatment of online-only attendance, whether related legal entities are combined, how NIR-served organisations are handled and which records could not be matched. When a definition changes, both the old and new series should remain visible rather than being silently joined.

Correction rights matter because affiliation records are imperfect. A entity may register under a trading name, parent, subsidiary or former employer. An organisation may be acquired between the meeting and publication. The public aggregate can be corrected without rewriting what a person said. Material corrections should carry dates and a short explanation so that apparent changes in coverage are not mistaken for changes in participation.

Independent access need not mean public access to personal records. A reviewer can receive protected, purpose-limited data under confidentiality, test deduplication and publish findings about accuracy and bias. Raw attendance and affiliation data should be retained only as long as the stated governance purpose requires. The reviewer should also test whether small-category suppression actually prevents re-identification across tables.

Denominator disputes should have an owner. Members and community entities need a documented route to question a published rate, ask for its calculation and receive an answer. If APNIC cannot align a numerator and denominator, the answer should be “not calculable from current records,” followed by a plan to improve the records—not a ratio assembled from convenient but incompatible totals.

This discipline would improve more than meeting analysis. It would clarify election turnout, consultation reach, service adoption and outreach outcomes. A stable measurement constitution—definitions, privacy limits, correction, independent review and version history—would stop each report from inventing its own public. The result is not statistical perfection. It is a shared basis on which the Secretariat, members, NIR communities and independent operators can disagree productively about what participation evidence means.

Sources