Summary

  • The next AFRINIC election should begin with a frozen, independently reproducible count of every resource member that satisfies the published eligibility rule. Turnout calculated only from registered or verified voters hides exclusions that occurred before a credential existed.
  • The public reconciliation should follow each eligible organisation through contact, voter designation, documentary validation, identity verification, credential issuance, ballot access, submission and acceptance. Every loss between stages needs a count, a reason and a review route.
  • A ban on formal proxies does not prove that voting power is dispersed. AFRINIC should publish anonymised concentration measures for natural persons representing multiple organisations, common corporate control, shared contact infrastructure and coordinated changes of designation.
  • Rejected applications and ballots require stable reason codes, notice, cure periods and appeal outcomes. Privacy can protect documents and individuals while aggregate counts, timing, reversals and regional effects remain public.
  • The September 2025 statistics of 581 listed voters, 548 completed biometric registrations and 484 votes cast provide a valuable partial reconciliation. They do not alone disclose the original eligible-member universe, rejection categories, attempted but unaccepted ballots or the disposition of every complaint.
  • Independent observation should cover electorate construction and remedies as well as voting software. A cryptographic event log can show that an action occurred without proving that the underlying eligibility decision was lawful, consistently applied or open to effective correction.
  • Recovery should be judged against a precommitted publication schedule and arithmetic that reconciles exactly. AFRINIC should not ask members to infer legitimacy from winner names, a turnout adjective or the absence of a public challenge after deadlines have passed.

The question is not only who won

AFRINIC's September 2025 board election filled eight seats after years in which the organisation had lacked a functioning board. That result mattered. The Number Resource Organization welcomed the election as a step toward restoring governance and acknowledged the staff who had kept services operating. AFRINIC later described the turnout as exceptionally high and set out a programme of institutional repair.

Yet a recovery election carries a heavier burden than an ordinary annual vote. It is asked not only to select directors but also to demonstrate that the institution can define its electorate, apply rules equally, resist improper control, correct error and preserve a record strong enough to survive disagreement. A list of successful candidates answers only the selection question.

The legitimacy question begins one level lower. How many organisations were legally and contractually entitled to vote? How many received notice through a current authoritative contact? How many tried to designate a representative? How many were refused, for what reason, and with what opportunity to cure? Did one human being control credentials for several nominally separate members? Did every submitted ballot enter the count exactly once? Were complaints decided before the result became practically irreversible?

Without those denominators, observers can praise or criticise turnout but cannot test inclusion. A result may be correct. The public record still may be too incomplete to show why it deserves reliance.

June 2025 established why documentation cannot be an afterthought

On 26 June 2025, the Receiver annulled the election held three days earlier, citing feedback and concerns about possible irregularities involving voter documentation and noting that concerns had been reported for investigation. Annulment protected the organisation from treating a disputed count as final, but it did not transform the abandoned vote into a complete diagnosis.

ICANN's subsequent open correspondence about the 2025 election described reports of registration difficulties, questioned an in-person category that could permit one holder of powers of attorney to represent an unlimited number of resource members, and urged a complaint mechanism, routes for correcting registration problems and independent expertise in election design and observation. ICANN carefully distinguished allegations from established findings, but its questions identify the evidential weakness that matters here.

An institution cannot repair an election only by changing the voting channel. It must account for every place authority can be gained, lost or concentrated before the voter reaches the ballot. Documentary authority, member status, contact access, representation, identity checks and complaint handling are all parts of the franchise.

The next vote should not wait for another controversy before assembling that account. Publication fields, custody duties and reconciliation tests must be fixed before members begin to designate voters. Evidence designed after a dispute will always look selective, even when those assembling it act honestly.

September produced a useful but incomplete public funnel

The replacement election made substantial changes. The September 2025 election guidelines required a fresh designation by each eligible member, direct online voting, identity and facial verification, and one vote for each eligible resource member on every open seat. Formal proxy voting and power-of-attorney voting were excluded. A provisional register preceded the final register, and submitted votes were to receive receipts.

The published numbers allow three informative calculations. AFRINIC's election statistics show 581 total voters, 548 completed biometric registration, 33 did not complete it and 484 cast votes. Biometric completion was therefore about 94.3 per cent of the listed voter population. Votes cast were about 88.3 per cent of completed biometric registrations and about 83.3 per cent of listed voters.

These are unusually useful stage counts for an institutional election. They show that participation did not simply equal registration and that 64 people who completed biometric registration did not cast a ballot. The final voter register also disclosed organisation, region, designated voter, nominator role and biometric-completion status, subject to stated data-protection limits.

But the first number in this public funnel is 581, not the total number of organisations evaluated under the eligibility rule. The page does not, on its own, reveal how many members existed at the cutoff, how many were found in good standing, how many were contactable, how many designation attempts failed or how many refusals were reversed. The visible funnel begins after several decisive gates.

A turnout percentage is a choice of denominator

Turnout is often presented as if it were an objective property of an election. It is a ratio, and the person selecting its denominator makes a governance choice. Using 484 votes cast over 548 verified voters answers a different question from using 484 over 581 listed voters. Using 484 over every member initially eligible to seek a vote would answer a different and more constitutional question again.

The narrow denominator measures use among those who completed the institution's requirements. It can be valuable for assessing software access or late abstention. It cannot reveal whether those requirements excluded a class of members. A high percentage after a demanding verification stage can coexist with low participation across the full entitled population.

The broad denominator measures the practical availability of the franchise. It includes the member whose notice went to an obsolete contact, the small operator that could not satisfy a document request in time, the organisation incorrectly marked in arrears and the voter whose identity check repeatedly failed. Some nonparticipation will be voluntary. Some will expose institutional friction. The point is to distinguish them rather than blend both into silence.

AFRINIC should publish several turnout ratios, each named by stage. No single percentage should carry the entire claim of legitimacy. The reader should be able to move from the broad constitutional denominator to the narrow accepted-ballot numerator and see exactly where participation changed.

The eligible-member universe must be frozen and reproducible

The first required field is the complete count of organisations evaluated for voting eligibility at a stated time. The election authority should derive it from the member record under the exact rule in force, preserve the version used and have an independent observer reproduce the count from protected evidence. Later changes should create a documented adjustment rather than quietly alter the baseline.

For the September 2025 board vote, the guidelines defined an eligible resource member in good standing by reference to a court order, completion of membership formalities before a specified date, current membership and payment of fees. Each element should have its own public aggregate. How many organisations met the date condition? How many were current? How many were excluded for an unpaid amount? How many payment records were disputed or corrected? How many associate or registered members were outside the voting class?

No private account balance or contract document needs to be published. Aggregate counts can show the rule's effect. Small cells can be combined where disclosure would identify an organisation, while the independent observer verifies the exact protected table.

Reproducibility also requires a membership snapshot identifier, extraction time, governing rule version and adjustment log. If the count cannot be recreated after directors take office, the denominator was only an assertion. A court, member or later committee should not have to reconstruct the electorate from changing live records and personal recollection.

Contactability is a separate institutional duty

Eligibility does not matter if an organisation never receives usable notice. The second layer should count how many eligible members had at least one validated governance contact, how many notices were delivered, how many bounced, how many were acknowledged and how many contact corrections were completed during the designation period.

This is not a demand to treat email opening as consent. It is a demand to measure whether AFRINIC used the published contact points on which its election depended. A message accepted by a mail server may never reach the authorised officer. Conversely, an organisation may receive notice and choose not to participate. Those states need different codes.

Contact quality can have regional and organisational effects. A multinational operator with a dedicated regulatory team is more likely to respond quickly than a small network whose registry contact left years earlier. Public institutions may require correspondence through formal offices. Language and time-zone choices can change response rates. Aggregate reporting by subregion, member scale and contact age can reveal unequal friction without exposing private addresses.

The election authority should also state what it did when notices failed. Did it try a second verified contact, telephone confirmation or a public reminder? Was the cure period the same for all members? How many members recovered access because of that effort? Measuring notice turns outreach from a ceremonial email into an accountable part of the franchise.

Designation is where organisational authority becomes human control

A corporate member cannot physically vote. A natural person exercises the vote on its behalf. The designation stage therefore converts an organisational entitlement into human control, and that conversion deserves more scrutiny than a name on a register.

The public account should show designation attempts, complete applications, incomplete applications, requests for additional evidence, acceptances, refusals, withdrawals, replacements and late confirmations. Each accepted designation should connect to one eligible member and one current authorising officer. Each rejected attempt should carry a stable reason code and a notice time.

The September final-register notice said four designated voters were added after the provisional publication because late confirmations had been received from nominating executives, while stating that no new applications had been accepted after the deadline. That may be entirely consistent with a cure for timely applications. A versioned adjustment table would make the distinction visible: application received before cutoff, confirmation received later, rule permitting cure, decision time and effect on the count.

The authority also needs to distinguish replacement from duplication. If an organisation changes its representative, the former authority must end before the new credential becomes active. The public aggregate should report replacement events, attempted concurrent designations and any votes prevented because authority was uncertain. The protected record should preserve who authorised each change and which version was effective at every moment.

Verification attrition needs reasons, not one residual number

Biometric completion provides one measure of readiness, but the gap between designation and verification can contain several different experiences. A voter may never start, abandon the attempt, fail document quality checks, encounter device incompatibility, face a mismatch, seek human review or complete the check after several attempts. Combining all of them as incomplete hides whether the barrier was choice, technology or decision.

The next election should publish counts for invitation delivered, account activated, identity review started, automated check completed, manual review requested, verified, refused, timed out and voluntarily withdrawn. Technical failures should be separated from identity decisions. Repeated attempts should be counted both as voter cases and as events so that one struggling user does not distort the number of affected organisations.

Verification accuracy has two sides. A false acceptance can give an unauthorised person a ballot. A false rejection can disenfranchise a legitimate member. Security language often focuses only on the first. An accountable election reports both suspected impersonation cases and legitimate voters restored after review.

The service provider should disclose accessibility options, supported devices, human-review availability and median and longest resolution times. AFRINIC remains responsible for the franchise even when a contractor operates the identity service. Outsourcing a check cannot outsource the duty to explain its aggregate effect.

Credential issuance must reconcile with verified voters

After verification comes authority to enter the voting environment. The election record should state how many credentials were created, delivered, activated, reissued, revoked and unused. The totals must reconcile with the verified-voter count, allowing only documented timing differences.

Credential events can reveal risks not visible in a ballot tally. Multiple reissues to one voter may indicate delivery trouble or attempted takeover. Several credentials routed through one contact point may reveal common administration. A credential created after the voting period opened may be legitimate following a successful review, but its timing should be visible. A revoked credential should be linked to a reason and a determination of whether any earlier ballot remained valid.

Receipts also need careful interpretation. A receipt can confirm that the voting service recorded a submission without exposing the vote. It should include a verification method that lets the voter confirm inclusion in the accepted set while preventing outsiders from proving selections for coercion or sale. The public report should count issued submission receipts, successful inclusion checks, failed checks and corrections.

The essential arithmetic is simple: every verified voter either receives one usable credential, has an explained pending case or has a reasoned refusal. Every credential either remains unused, produces one accepted ballot or is revoked with its ballot status resolved. Unexplained residuals should prevent certification until reconciled.

No formal proxies does not mean no concentration

The September rules removed proxies and powers of attorney, a sensible response to the concerns surrounding the June vote. But legal labels do not exhaust the concentration question. One natural person may be duly designated by several organisations. Several member companies may share beneficial control. A service company may administer contacts for unrelated networks. A coordinated group may replace designated voters at the same time.

None of these facts automatically proves wrongdoing. Corporate groups legitimately centralise governance functions, and small operators may rely on shared expertise. The governance concern is undisclosed voting concentration, not the existence of cooperation. Members need to know how much electoral power was controlled through common people and control groups before judging whether one-member-one-vote produced dispersed choice in practice.

AFRINIC should publish anonymised concentration bands: number of natural persons representing one, two, three, four, five and more organisations; number of member votes grouped by declared common control; and the largest, second-largest and aggregate top-ten shares. It should also report shared email domains, phone numbers or authorising officers as risk indicators after removing false matches and protecting identities.

The independent observer should inspect the named cases and state whether each concentration was disclosed, verified and consistent with the rules. Public data can remain aggregated. What cannot remain unknown is whether 581 listed organisations corresponded to 581 independent human voting relationships or a materially smaller set of controllers.

Proxy concentration and representative concentration must be reported together

Formal proxy limits are easy to count. Representative concentration is harder because authority may arise through employment, company office, delegated administration or common ownership rather than a document titled proxy. Reporting only the formal category creates an incentive to move influence into another legal form.

The data contract should therefore use a functional definition. A person controls a ballot when that person can access the credential, make the selection or direct the individual who does. An organisation belongs to a common-control group when the same beneficial owner or controlling body can determine its governance decisions. These tests should be narrow enough to avoid speculation and strong enough to reveal obvious aggregation.

Three counts are needed. The legal count states accepted and refused formal proxies under the applicable rule. The human-control count states how many ballots each designated person could cast. The organisational-control count aggregates member votes under verified common control. Differences between the counts are informative rather than contradictory.

Publication should not accuse a group merely because it votes coherently. Secret ballots prevent direct attribution, and common positions may arise independently. Concentration data measures capacity, not proof of coordinated selections. It allows members to assess capture risk while preserving the right of affiliated organisations to exercise valid votes where the governing rules permit them.

Rejection codes should identify the gate that closed

A refusal described only as ineligible is not reviewable. The next election should use stable, mutually exclusive primary reason codes supplemented by secondary codes where necessary. Categories should include membership class, cutoff date, inactive membership, unpaid amount, disputed payment, missing executive authority, late application, incomplete evidence, identity mismatch, duplicate designation, unresolved common-control conflict, security hold and voluntary withdrawal.

Each code should have a published definition, evidence standard, decision owner, cure route and appeal deadline. The authority should not add a new category during the election without publishing the change, identifying affected cases and allowing review. Free-form explanations may remain protected because they can contain personal or commercial information, but the coded outcome must be public in aggregate.

Reason codes reveal institutional choices. If most refusals concern stale contacts, the remedy is better member-record maintenance. If payment disputes dominate, the organisation needs an election-specific reconciliation window. If identity checks cause regional imbalance, the provider and review arrangements require attention. A single rejection total cannot guide any of those corrections.

The report should show decisions reversed during cure and appeal. A high reversal rate may indicate that the initial gate was too brittle. A zero reversal rate is not automatically reassuring; it may mean decisions were perfect, reviews were inaccessible or members gave up. Timing and participation in review complete the picture.

Invalid ballots require a precise vocabulary

Online elections are sometimes described as having no invalid ballots because the interface prevents malformed selections. That statement confuses ballot formatting with every reason a submission might not enter the final tally. A voter can lose connectivity, submit after closure, use a revoked credential, encounter a duplicate event, abandon a partially completed ballot or receive a receipt for an event later quarantined during verification.

The next election should distinguish attempted sessions, opened ballots, submitted ballots, accepted ballots, duplicate submissions blocked, late submissions refused, technically incomplete submissions, security-quarantined submissions and ballots excluded after review. If the system truly makes some categories impossible, the report should state a zero and explain the control that prevented them.

Blank votes, abstentions and undervotes also need explicit treatment. Eight separate board seats can produce different participation counts even when 484 people access the election. A voter may select a candidate for one seat and deliberately leave another unselected. Candidate totals should reconcile to accepted ballots, blank selections and any permitted none-of-the-above choice for each contest.

The 2025 guidelines said the result announcement would include total votes for each candidate. The durable published winner announcement names the elected directors but does not display those totals. Future certification should place contest-level arithmetic in the same permanent record as the winner names.

Complaints are part of the denominator

A member denied a vote is not restored by a complaint channel that decides the case after the result is irreversible. Complaint data must therefore be integrated with the electoral timetable. The public report should count complaints by stage, issue, time received, interim action, disposition, reversal and resolution time.

There should be at least three routes. A rapid service route handles access and technical problems while voting remains open. An eligibility review route examines designation and member-status decisions before credential cutoff. A certification challenge route addresses tally, rule or integrity concerns after closure but before directors are finally confirmed. Each route needs a decision-maker independent from the original decision where practical.

The system should issue a case receipt and state whether the complaint pauses any act. Most individual disputes should not freeze the whole election. A complaint affecting one organisation can preserve that ballot opportunity; a system-wide defect may justify a broader pause. The decision and scope should be reasoned.

Aggregate publication must include unresolved cases at each milestone. Certification should state how many complaints remained open, why none could change the outcome or what contingent remedy applies. Silence cannot be classified as satisfaction. A member who lacked access, did not understand the route or expected no timely remedy may never file. Surveys and outreach to nonparticipants can identify that hidden denominator.

Appeals need outcomes, not merely availability

An appeal clause is weak evidence unless members can see whether it worked. AFRINIC should publish the number of appeals filed, accepted for review, dismissed for lateness, decided on the merits, upheld, partly upheld, rejected, withdrawn and unresolved. It should state the median and maximum decision time and the number of cases that changed voter access or ballot status.

The appellate body should receive the original reason, evidence considered, member response and complete timing record. It should not receive the member's likely political preference. Recusal rules must cover relationships with candidates, member groups and election contractors. A protected register can hold the details while the public report supplies aggregate assurance.

Effective review also requires a remedy. Before voting, the remedy may be corrected registration, extra time equivalent to the time lost or a replacement credential. During voting, it may be restored access and preservation of the full voting window. After closure, it may be inclusion of a demonstrably timely ballot, a rerun of an affected contest or a finding that the error could not alter the result but still requires correction and compensation.

Every appeal should end with a durable disposition. The election archive should not preserve only successful complaints or cases mentioned publicly. Completeness is what allows later reviewers to tell whether similar members received similar treatment.

Privacy should shape publication, not erase accountability

Election records contain identification documents, signatures, contact details, biometric results and commercial information. Publishing all of them would create serious security and privacy harm. That does not justify withholding counts, definitions, timings and aggregate effects.

The data contract should define three views. The public view contains rules, stage totals, reason-code aggregates, concentration measures, regional comparisons, complaint outcomes, observer findings and cryptographic commitments to the underlying records. The member view lets each organisation inspect its own status, events and decisions. The protected audit view gives authorised independent reviewers access to named evidence under confidentiality and retention controls.

Small-cell protection can prevent indirect identification. Where only one organisation in a subregion received a rare rejection code, the public table can combine categories or delay a detailed breakdown. The exact case remains available to the observer and the affected member. Suppression rules should be stated in advance so they cannot be used selectively.

Biometric material requires particularly narrow handling. AFRINIC should disclose what was collected, which party controlled it, where it was processed, how long it will be retained, how deletion is verified and how a voter can challenge an identity result. Public accountability concerns the operation and impact of the check, not the publication of faces or documents.

Independent observation begins before the ballot opens

An observer who watches the final tally cannot assess whether the electorate was lawfully constructed. The mandate should begin before the eligibility snapshot and continue through designation, verification, credential issuance, voting, counting, complaints, certification and record retention.

The observer should be selected through disclosed criteria, publish conflicts and contractual limits, and retain the ability to report disagreement. Independence is weakened if the election authority can edit findings, terminate access after an adverse observation or prevent publication beyond a bland assurance statement. Funding should be disclosed and should not depend on a favourable conclusion.

The report should distinguish direct examination from reliance. The observer may inspect a sample of member files, reproduce all aggregate counts, witness key ceremonies, test the voting service, review complaint files and verify the final reconciliation. For areas outside the mandate, the report should say so. A signed opinion that names its evidence is more valuable than a universal claim that the election was free and fair.

ICANN's 2025 correspondence recommended independent election expertise and an observation role over the electronic system. AFRINIC should extend that logic to the whole electorate funnel. Software integrity is one part of the opinion; equal access and reasoned exclusion are equally important parts.

Cryptographic records cannot decide whether a rule was fair

The 2025 election FAQ said the online system logged actions cryptographically and that opening and closing would be livestreamed. Such controls can strengthen event integrity. They can show that a credential was issued at a time, that a ballot commitment existed before closure or that an archive was not later altered.

They cannot determine whether the organisation receiving the credential was eligible, whether a payment dispute was resolved consistently, whether one person controlled several members or whether a rejected identity match deserved reversal. Cryptography protects the statement placed into the record; it does not supply the institutional judgment that made the statement legitimate.

The public report should therefore connect technical proof to governance proof. Hash commitments can bind the voter-register versions, reason-code table, accepted-ballot set and complaint ledger. The observer then examines the underlying protected records and attests that the committed totals follow the published rules. Members can verify that later publications correspond to the same preserved evidence.

This combined approach avoids two errors. One is trusting administrative assurances without tamper evidence. The other is treating a secure voting service as a substitute for accountable electorate construction. Recovery requires both reliable events and reviewable reasons.

The election service provider needs its own accountability record

A contracted voting or identity company controls important parts of the member experience. The provider should publish, through AFRINIC, the service scope, relevant assurance reports, incident definition, availability, support coverage, accessibility, data locations, subcontractors, retention schedule and independent test results. Commercial confidentiality can protect exploitable detail without hiding performance.

The post-election report should count service incidents, affected voters, duration, retries, manual interventions, late resolutions and any event that changed the voting window. It should distinguish failures in AFRINIC member data from failures in the provider's identity or ballot service. Members need one integrated account even when responsibility is divided contractually.

Trustees should not rely only on a provider-generated dashboard. Independent evidence should include time sources, access logs, credential totals, encrypted ballot commitments, opening and closing records, key-custody events and reconciliation against the member register. No trustee should be able to alter a ballot alone or silently extend the election.

Provider replacement and data return should also be tested. Election evidence must remain available if the vendor ceases trading, disputes payment or ends the contract. A recovery election whose audit trail depends indefinitely on one foreign commercial service has exchanged one concentration risk for another.

Regional analysis can expose unequal friction without regionalising the vote

AFRINIC's membership spans different languages, legal systems, connectivity conditions and organisational capacities. Reporting by subregion can reveal whether one verification method or notice practice affected members unevenly. It does not imply that votes should have unequal weight.

For each stage, the public table should show eligible organisations, successful contacts, designations, verifications, credentials, accepted ballots, refusals, complaints and reversals by subregion. Percentages should always carry both numerator and denominator. Where counts are small, categories can be combined under the preannounced privacy rule.

Other useful cohorts include member tenure, organisation scale, public-sector status and age of the registered contact. The purpose is diagnostic. If newer members are excluded by a court-defined cutoff, that should be visible as a rule effect. If older contacts drive nonparticipation, member maintenance needs repair. If one region has repeated biometric failures, accessibility and support should be investigated.

Analysis must not become profiling. The report should not infer political preference from region, company type or resource holdings. It should not publish the size of address portfolios as a proxy for importance. Cohorts reveal access conditions, not the content of the secret ballot.

Changes to rules and registers need a public version history

Election rules often change for legitimate reasons: a court order, discovered ambiguity, service failure or correction to a date. The legitimacy problem arises when members cannot determine which rule governed an event. Every published document should carry a version, effective time, authority, change summary and link to the previous version.

The voter register needs the same discipline. The provisional count, corrections, late confirmations, removals and final freeze should reconcile. After the freeze, mandatory changes should appear in an exception ledger stating the rule, decision time and effect. The original version must remain available.

The September guidelines themselves noted that they replaced earlier July guidelines and later corrected a constitutional reference. Recording that correction is better than silently editing the text. Future elections should make the practice systematic across guidance, forms, eligibility lists, candidate materials, result tables and complaint rules.

Version discipline protects both members and officials. A member can show which instruction it followed. An election officer can demonstrate that similarly timed cases used the same rule. An observer can reproduce stage counts without guessing which live page existed on a past date. A court can assess the actual decision environment rather than a later cleaned-up archive.

Publication must occur before, during and after voting

A complete report released months later is useful for history but weak for remedy. The data contract should have a schedule. Before designation opens, AFRINIC publishes the eligibility rule, initial denominator, reason codes, privacy thresholds, observer mandate and complaint routes. At provisional registration, it publishes stage counts and concentration indicators. At the final freeze, it reconciles corrections and unresolved cases.

During voting, public updates should be limited enough to avoid influencing participation or revealing choices. They can report system availability, credential delivery, aggregate ballots received and incidents without candidate totals. Any extension, pause or restart requires a reason, authority and equivalent opportunity for affected voters.

At closure, the trustees publish accepted-ballot totals, rejected-event categories and a commitment to the sealed result before revealing candidate figures. Certification then adds per-contest totals, blank selections, complaint status, observer findings and exact arithmetic. A later final report closes appeals, deletion duties and remediation.

This staged approach makes transparency useful when members can still act. It also limits retrospective discretion. Officials know in advance which numbers must appear and cannot omit an inconvenient category merely because it became controversial.

The minimum public reconciliation should be exact

The final report should include a table in which every row has a defined relationship to the next. At minimum, it should disclose:

Stage Required public evidence
Member universe Total organisations evaluated, governing cutoff, membership classes and snapshot time
Eligibility Eligible, ineligible and pending counts with primary reason codes and adjustments
Contact Notices delivered, failed, corrected and acknowledged, with outreach methods
Designation Attempts, complete files, cures, accepted representatives, replacements, refusals and withdrawals
Concentration Representation per natural person, verified common-control bands and largest aggregate shares
Verification Started, completed, manually reviewed, refused, abandoned and technically failed cases
Credentials Created, delivered, activated, reissued, revoked and unused credentials
Voting Sessions opened, submissions, accepted ballots, blocked duplicates, late or quarantined events
Contests Accepted ballots, candidate votes, blank selections and any permitted abstention for every seat
Remedies Complaints, appeals, reversals, unresolved cases, decision times and practical remedies
Assurance Observer scope, exceptions, service incidents, preserved commitments and final reconciliation opinion

The arithmetic should include formulas, not just prose. Eligible equals accepted eligibility plus unresolved eligibility at the relevant cutoff. Accepted designations plus withdrawals and refusals should reconcile to complete designation cases. Accepted ballots plus unused valid credentials and resolved revoked credentials should reconcile to credentialed voters. Every contest should reconcile to its accepted ballot population.

Where categories overlap, the table must say so and provide a non-overlapping primary classification for totals. An unexplained difference of one matters because the institution is asserting one vote per member. Exactness is not pedantry; it is the substance of equal treatment.

A worked example shows what the contract changes

Imagine that the initial snapshot contains 760 organisations evaluated under the rule. Six hundred and fifty satisfy all eligibility conditions, 80 fall outside the permitted membership class or date, 20 have payment questions and 10 require membership-status review. Notice reaches 630 eligible organisations, while 20 contacts fail and enter assisted correction.

Five hundred and ninety members submit designation files. Twenty are incomplete; 14 cure the defect, four withdraw and two are refused after review. The final register contains 584 organisations. Concentration analysis finds that 560 representatives each serve one organisation, eight people serve two organisations and two people serve four organisations within disclosed corporate groups. The public sees the distribution, not private identity papers.

Five hundred and sixty representatives complete identity verification. Twelve never start, seven abandon after technical difficulty and five are refused, of whom two win appeal and receive equivalent time. Five hundred and sixty-two credentials become valid. Five hundred and twenty ballots are accepted, 40 credentials remain unused and two are revoked before submission. One duplicate event is blocked and no accepted ballot is removed.

For each seat, 520 accepted ballots reconcile to candidate selections and blank choices. Nine complaints are resolved before certification, one remains open but cannot alter the margin, and the observer explains why. Readers can now identify where participation fell and whether remedy worked. The winner is no longer detached from the electorate that produced the result.

Strong objections support a narrower and better design

One objection is cost. Detailed reporting, observation and protected review require money. But AFRINIC already bears the cost of election technology, legal advice, committees and institutional uncertainty. Reconstructing incomplete evidence after a disputed vote is more expensive than designing consistent records once. The contract should reuse ordinary member-account events and generate aggregate reports without exposing documents.

Another objection is that publication will fuel litigation. Ambiguity fuels litigation more effectively. Exact definitions and preserved evidence narrow factual disputes. Members may still disagree about law or policy, but they will not need to guess how many cases existed or which rule applied.

A third objection is privacy. That concern is valid and argues for layered access, small-cell protection and independent review, not for hiding the denominator. Aggregate rejection counts and concentration bands do not require publishing passports, balances or ballot choices.

A final objection is that no election can satisfy every critic. True. The objective is not universal approval. It is a record that lets a reasonable outsider test inclusion, concentration, count and remedy. Legitimacy is strengthened when disagreement can focus on disclosed facts rather than competing anecdotes.

The 2026 rules make denominator quality a continuing issue

AFRINIC's 2026 election guidelines carried forward the September 2025 board-election register as a baseline for certain elections and member resolutions, invited members who had not participated in the August 2025 registration exercise to register, allowed confirmation and correction, and required provisional and final registers. They also clarified how designated representatives differ from formal proxies.

That continuity makes the 2025 denominator more than a historical concern. A baseline can preserve verified relationships and reduce repeated burden. It can also preserve old exclusions, stale designations or undisclosed concentration if inherited without a complete reconciliation. Every reuse should state additions, removals, confirmations, deemed confirmations, replacements and mandatory exclusions.

Different elections may have different electorates. Resource-member voting for a governance committee is not identical to a community register for an NRO Number Council seat. Each contest needs its own denominator and transition table. Combining participation figures across unlike electorates would mislead.

By the next board rotation, AFRINIC should be able to show a continuous history from the recovery election through subsequent corrections. A durable franchise is maintained, not recreated through hurried contact each time a seat opens.

What independent observers should certify

The observer's final opinion should answer a finite set of questions. Was the eligibility rule fixed and consistently applied? Did the independently reproduced eligible count match the public denominator? Did every accepted designation have valid organisational authority? Were concentration cases identified and handled under disclosed rules? Did verification and credential totals reconcile?

It should then address the ballot. Did every valid credential produce at most one accepted ballot? Were all exclusions and quarantines resolved? Did each contest total reconcile? Were opening, closing and tally events jointly controlled? Could any identified incident have changed a result?

Finally, it should address remedy and preservation. Did affected members receive notice, cure and timely review? Were unresolved cases disclosed before certification? Are the underlying records preserved under a retention schedule, and can the public commitments be verified later?

The opinion should list exceptions and material weaknesses. A qualified opinion is more credible than an absolute blessing that conceals limits. If the observer could not inspect beneficial-control evidence or contractor logs, members should know. Certification is useful because it compresses extensive examination into a reasoned conclusion, not because it replaces the evidence tables.

Recovery requires a denominator that can withstand an unfriendly reading

Institutions often publish election information for supportive readers: dates, candidates, turnout, winners and photographs. A recovery election must also satisfy the sceptical member, losing candidate, court, network operator and future board. Each should be able to test the record without accepting the election authority's reputation as proof.

That is why the denominator matters. It exposes exclusion before voting, concentration behind nominally separate members, attrition inside verification, uncounted events after submission and ineffective remedies after complaint. It also protects the institution from exaggerated allegations by showing the full scale and disposition of problems.

AFRINIC does not need to disclose secret votes or personal evidence to meet this standard. It needs a frozen electorate, named stages, exact arithmetic, coded decisions, independent access and timely publication. Those are ordinary controls adapted to an organisation whose governance decisions affect network operators across a continent.

The next AFRINIC election should not be described as successful merely because it finishes. Success should mean that every entitled organisation can find itself in the account, every exclusion has a reviewable reason, every accepted ballot reconciles and every concentration risk is visible at the right level. A board elected under that standard will inherit more than seats. It will inherit evidence of mandate.

Sources