Summary

  • AFRINIC's 18 August Members and Stakeholders Survey announcement says more responses will make the results more representative and useful in guiding future actions and priorities.
  • The checked launch pages describe a public, forwardable survey whose answers are anonymous unless a respondent identifies voluntarily. They do not publish the denominator, unit of analysis, duplicate rules, member/stakeholder split, weighting, coding or result-release method.
  • That omission does not prove manipulation. The survey was six days old at the evidence cutoff, hidden technical controls may exist, and an open anonymous channel can produce candid service intelligence.
  • Before the Receiver, the current group described by NRS as the purported Board, management or their supporters invoke the results as legitimacy, AFRINIC should publish the method and state that the exercise is advisory rather than a member vote.

AFRINIC has put the conclusion before the method.

The regional registry's new Members and Stakeholders Survey tells readers that every response will help shape its direction. Its 18 August announcement goes further: the more responses AFRINIC receives, it says, the more representative and useful the results will be for future actions and priorities.

That proposition is not yet supported by the public launch record. The same announcement asks readers to forward the link to colleagues and networks. The form was accessible without a member login at the evidence cutoff. Its welcome screen says responses are anonymous unless a participant chooses to identify at the end. Neither checked page defines the population that “representative” refers to or the rules that turn a response into evidence about that population.

This is a reporting gap, not proof of a rigged survey. AFRINIC may publish a full methodology when the exercise closes. LimeSurvey can support tokens, cookies, branching and server-side controls that are not visible on a welcome page. The research for this briefing advanced only to the first question, selected no answer and submitted no response. It therefore does not claim that duplicate answers are possible, that members cannot later be identified or that later questions fail to separate categories.

But the burden sits with the institution making the representation claim. “More” is a response count. “Representative” is a relationship between a defined sample and a defined population. One does not become the other through volume alone.

The launch combines populations that carry different authority

The survey title joins members and stakeholders. Its welcome text says AFRINIC is nothing without more than 2,500 members and tens of thousands of people helping the Internet in Africa. The first visible question asks for a primary professional role and offers choices including chief executive or director general, network engineer, systems engineer, project or product manager, business executive, researcher, student and other.

Those are all potentially useful voices. They are not interchangeable units of governance.

A student can identify a documentation problem. An engineer can report a routing-certificate or registry-service failure. A chief executive can describe investment uncertainty. A consultant can see patterns across clients. An authorised representative of an AFRINIC member can speak within the mandate granted by that organisation. The survey can listen to all of them without pretending that each answer carries the same corporate or constitutional weight.

The public record does not yet say whether AFRINIC will report those groups separately. It does not identify the unit of analysis: person, organisation, member account or completed form. It does not give a frozen member denominator, a stakeholder frame, response-rate formula or rule for several people from one company. It does not say whether forwarded responses are weighted equally with directly invited members.

Anonymity creates a real benefit. People may criticise a registry more honestly when their names are not attached, especially during a contested institutional transition. Yet anonymity and authenticated member mandate pull in different directions. A response can be protected as private advice, or it can be counted as proof that a legally defined member supports a decision. Without a disclosed authentication and separation method, it cannot safely do both.

Heng Lu's doctrine makes the boundary exact: a stakeholder is affected by a decision; a principal authorises one. Participation can supply evidence, expertise, warning and objection. It becomes a legitimacy device when officials use the presence of affected people as permission to bind the parties carrying the loss.

That distinction is particularly important at AFRINIC. NRS publicly calls the current directors the “Purported Board” and asks whether the Receiver, that group or some combination now lawfully speaks for the company. This briefing records NRS's position; it does not turn the label into a final court holding. A contested authority cannot cure the contest by controlling a consultation and later citing the responses as public approval.

AFRINIC's own 2019 report shows what disclosure looks like

AFRINIC does not need a new academic theory to repair the launch. Its archived 2019 Customer Satisfaction Survey report already provides a practical comparator.

That report says DCDM Research was commissioned for the work. It identifies Computer-Aided Web Interviews, a field period from 16 October 2018 to 7 February 2019 and invitations sent to all AFRINIC members, with periodic follow-up by AFRINIC. It also says a public website link was available to general stakeholders.

Crucially, the report calls the sample self-selected. It publishes a total of 68 responses: 54 in the member section and 14 from general stakeholders. It provides separate questionnaire flows. When it reaches the stakeholder results, it warns that the base is very low and the data should be interpreted with caution.

That earlier exercise was not statistically perfect. Sixty-eight self-selected answers cannot be made representative by a polished report, and a stakeholder base of 14 is plainly fragile. The useful precedent is disclosure. Readers could see the universe approached, the field dates, collection method, subgroup sizes and limitation. They could decide how much weight the result deserved.

The 2026 launch has not yet supplied the equivalent join. No closing date appeared on the checked announcement. No target response number, sampling frame, duplicate-control description, weighting plan, coding protocol or results timetable was visible. The welcome page displayed no survey-specific statement identifying data custody, retention, raw-data access or a publication commitment. AFRINIC has a general privacy policy elsewhere, and an absent notice on this screen is not proof that no internal or later-page rule exists. It is evidence only that the launch does not show the rule to the person being asked to trust the result.

The standard is therefore modest: match the candour of 2019 before making a stronger representativeness claim in 2026.

The people who control interpretation control the governance signal

A questionnaire looks like a set of questions. Institutionally, it is a sequence of gates.

Someone chooses the topics. Someone decides who receives a direct invitation and who sees a forwarded link. Someone defines “member” and “stakeholder,” handles repeat sessions, removes incomplete responses, codes open text, groups answers, selects percentages and writes the executive summary. Someone decides whether a result is merely a service-improvement clue or evidence that “the community” supports a policy, a strategic direction or the officials advancing it.

That is the agency problem. Respondents supply information; the institution controls classification and use. Members and operators bear the consequences if a claimed priority changes fees, registry processes, transfer treatment, RPKI, record continuity or the risk attached to number resources. A method published after officials have seen the answers would still help, but it would permit the rules of significance to follow the result. A dated pre-analysis and release plan is stronger because it constrains that discretion in advance.

The survey may never be used improperly. No reviewed evidence shows fabricated responses, suppressed answers, Receiver or Board interference, or a predetermined conclusion. The point of a method is not to allege those events. It is to make them unnecessary to trust against.

Supporters of the present arrangement should want the disclosure too. If the responses are broad, the categories separated and the analysis reproducible, a transparent method gives their preferred interpretation more weight. If the survey is exploratory, an advisory label protects it from being attacked as a failed election. The method is not an obstacle to the result. It is what makes the result usable.

What AFRINIC should publish before citing the outcome

AFRINIC should identify the survey owner, approving authority and decisions the results may inform. It should publish opening and closing dates, every recruitment channel and the complete questionnaire with branch logic. The target populations and frozen denominators for members and wider stakeholders should be stated separately.

The report should define whether the count represents people, organisations, member accounts or forms. It should explain authentication, duplicate and bot handling, repeat sessions and multiple respondents from one organisation. If authenticated members and open stakeholders are both included, their results should never be merged into one mandate-shaped percentage.

AFRINIC should publish subgroup sizes, response rates, weighting, excluded and incomplete counts, coding rules for open answers and the treatment of small cells. It should name who can access raw data, how long data are retained, how anonymity is protected and whether an independent reviewer can reproduce the published tables.

Finally, the result report should state what the survey cannot do. It cannot ratify the authority of a Receiver or Board, cure a contested corporate act, replace a valid member vote, waive a member's rights or convert a forwardable consultation into a mandate.

Readers who want the deeper denominator framework can use BTW's research on why the population selected for a governance percentage is itself a power decision. This briefing applies that discipline to a narrower, current event: the launch of an AFRINIC survey before its public method.

Sources