Summary

  • AFRINIC reported 42 survey respondents after AFRINIC-10 in Cairo, held from 10 to 21 May 2009 with 135 event participants. The exact arithmetic is 42/135 = 14/45 = 0.311111... = 31.111111... percent, but this is only a crude respondent-count-to-event-participant-count ratio, not a demonstrated response rate or representative sample.
  • The report contains useful, mixed service evidence: 72.5 percent said their reasons for attending were fully fulfilled, while 95 percent said the website needed improvement. Those findings describe reported answers, not all attendees, members, operators, end users, Africa, or an authorising community.
  • A trustworthy denominator ledger would disclose who was eligible and invited, forms delivered, starts, unique and partial completions, item-specific denominators, blanks, exclusions, duplicate checks, wording, field dates, incentive, coding, and member/non-member cross-tabs. Without that chain, percentages cannot be converted into exact people or institutional mandate.
  • Participation is evidence, not authority. AFRINIC is a private bookkeeper and operational coordinator: survey satisfaction, supporter endorsement, consensus language, or attendance cannot give it sovereign, legislative, regulatory, police, prosecutorial, judicial, punitive, confiscatory, or ownership power.

Forty-two answers, bounded honestly

AFRINIC-10 took place jointly with AfNOG at the InterContinental Cairo Citystars Hotel in Cairo from 10 to 21 May 2009. AFRINIC's official meeting report recorded 135 event participants and later summarised a survey completed by 42 respondents. That disclosure creates the article's central tension. Forty-two direct accounts are substantial enough to reveal friction in registration, content, communications, sessions, and the website.

They are also far too weakly specified to become a verdict delivered by everyone who attended, much less by AFRINIC's members, the operators affected by its services, end users, or a politically meaningful African public.

The calculation should be stated without euphemism. Forty-two divided by 135 reduces to fourteen forty-fifths. In decimal form it is 0.311111..., or 31.111111... percent. The difference is 93: that many people lie outside the respondent count when the published event total is used as the comparison. Ninety-three divided by 135 is thirty-one forty-fifths, or 68.888888... percent. This is a crude coverage ratio between two published counts. It is useful because it gives readers a scale.

It is not a demonstrated response rate because the report, as retrieved, does not establish that the 135 and the 42 share one verified invitation and eligibility frame.

That qualification is not pedantry. A response rate normally requires a known denominator of eligible recipients, a record of invitations or forms delivered, and rules about unique, partial, invalid, or duplicate returns. Here, the report does not prove how many forms were distributed, whether every event participant was invited, whether every invitee was eligible, whether the respondent records were unique, or whether every respondent belonged to the 135-person event count. It does not document remote-participant treatment.

The ratio therefore compares a respondent count with an event-participant count while preserving uncertainty about the relationship between those sets.

This limitation helps rather than destroys the instrument. It prevents false precision from crowding out the real information the respondents supplied. Readers can say that AFRINIC collected 42 reported experiences and that those experiences included both satisfaction and criticism. They can use the answers to generate service hypotheses, identify areas for investigation, and design later measurements. What they cannot do is rename 31.111 percent as representative turnout, approval, consensus, mandate, or a population estimate. The denominator is a guardrail around useful evidence.

The categories must not slide into one another

The report gives a membership-status split: 22 respondents answered that they were members, 19 answered that they were not, and one response omitted that item. The sum is exactly 42. That clean addition does not resolve the harder identity questions. It does not establish 22 distinct member organisations. It does not prove that any respondent was an authorised legal representative of a member. It does not tell us whether several individuals came from one organisation, whether one person spoke in a personal or professional capacity, or how membership status was verified.

Eight categories therefore need to remain separate. A respondent is a person or record represented in the survey results. An event participant is someone counted in the meeting attendance figure. An AFRINIC member may be a legal or organisational party within AFRINIC's corporate arrangements. A member organisation is not necessarily identical to any individual attending or filling in a form. A voter is someone eligible and authorised under a defined decision rule. An affected operator may depend on registry services without attending the meeting or belonging to the respondent set.

An end user may experience downstream effects while having no relationship to the survey. “Community” is a description of overlapping interests and activity, not a legal principal capable of silently delegating power. A population is a defined universe from which a sample can be assessed. None of these labels is interchangeable.

The temptation to merge them is easy to see. Survey prose can move from “respondents said” to “participants believed”; institutional prose can then move from participants to members, and from members to “the community.” With each change, a bounded observation appears to acquire a larger principal. Yet no additional authorisation record enters the chain. The people outside the respondent count do not become represented merely because the report uses percentages, and absent member organisations do not delegate corporate power by implication. The whole continent certainly does not speak through 42 forms.

This is the impact mechanism of denominator inflation. A count gathered for service learning is enlarged into a constituency; a constituency is enlarged into a collective will; collective language is then offered as support for control. The safeguard is equally mechanical: preserve the noun attached to every denominator. Say respondents when the evidence concerns respondents. Say event participants only when the attendance record concerns event participants. Say members or voters only where corporate records establish those statuses and their authority. A denominator should identify not merely how many, but how many of what.

A mixed service record, not a victory banner

AFRINIC's report said that 72.5 percent found their reasons for attending fully fulfilled. It also said respondents were very satisfied with the venue, registration, and content, and reported that 64 percent regarded the existing meeting format as more appropriate. Those findings matter to an event organiser. They point toward features worth maintaining and questions worth asking again. They do not prove a general institutional approval rating, because the survey's frame, item denominators, wording, and nonresponse pattern are not available in the record supplied here.

The digital-service findings are especially valuable because they resist a one-directional celebration. The report recorded 56 percent describing website content as very useful, while 58 percent described the site as somewhat navigable. Most strikingly, it said 95 percent thought improvement was needed. These numbers can coexist: respondents might value the information while finding the route to it difficult. For a registry and meeting organiser, that is actionable evidence.

It suggests separating content quality from information architecture, testing specific navigation paths, and asking what “improvement” meant instead of treating a high satisfaction label as a substitute for diagnosis.

The communications results were similarly layered. Newsletter readership was reported at 51 percent, and 50 percent described newsletter content as very useful. The report also said 75 percent considered communications enough. A 51 percent readership figure is not the same measure as a 75 percent judgment about communications, and the absence of exact item wording prevents confident interpretation of how respondents understood either one. Different items may also have drawn different numbers of answers. The right reading is that the report contains several communications signals, not one composite score.

Session-related figures offer another limited operational clue. The newcomer session was reported at 39 percent and the MyAFRINIC presentation at 33 percent. These published percentages may guide programme design: organisers could examine scheduling, discoverability, target audience, room capacity, or the clarity of session descriptions. But the report does not give the denominator for either item, and it does not establish whether a percentage described attendance, reported interest, eligible respondents, or another precisely worded concept beyond the report's summary.

The figures must remain attached to the source's wording and not be promoted into exact attendance counts.

Taken together, the service record is mixed in the productive sense. Respondents could value the venue, registration, content, and meeting format while still identifying substantial website problems. They could regard communications as sufficient while reporting only partial readership of one channel. Different people could also answer different questions.

A credible institutional reading does not average these signals into “satisfaction.” It asks which operational surface each item measures, which respondents reached it, what comparison is meaningful, and what follow-up evidence would distinguish a local inconvenience from a persistent service failure.

Why percentages cannot be reverse-engineered into people

Several reported percentages, when multiplied by 42, produce fractional results. That is consistent with rounding, with item-specific denominators smaller than 42, or with both. It proves neither. For example, treating 72.5 percent as though all 42 people answered would create an impossible fraction of a person. One could search for a smaller denominator that yields a neat fraction, but doing so would be an invented reconstruction. Many combinations of answered items, skipped items, and rounding rules can produce the same displayed percentage.

The correct unit of evidence is therefore the published percentage as a rounded report value. It should not be converted to an asserted integer count. Nor should all percentages be assumed to use 42 as their base. The report does not supply item-by-item totals, raw counts, or missingness. A reader cannot know whether a blank was excluded from a particular calculation, counted in a neutral category, or handled some other way. The one exact membership split is different because the report supplies three counts that add to 42. That precision cannot be borrowed by unrelated items.

Reverse calculation would manufacture confidence in at least three places. First, it would pretend to know how many people answered a question. Second, it would pretend to know the rounding convention. Third, it would imply that the resulting people are unique and eligible within a defined frame. None follows from the rounded percentage. Publishing exact numerators and denominators beside every percentage would solve much of the problem. In their absence, disciplined analysis stops where the source stops.

This is also why a conventional margin of error or confidence interval would mislead. Such calculations need assumptions about sampling, population, response mechanism, and usually independence or weighting. The record documents no random sample, complete sample frame, or weighting method. A mathematical interval placed around 42 would look scientific while answering a question the instrument was not shown to have been designed to answer. The uncertainty here is structural before it is statistical.

The incentive is known; its effect is not

On the closing day, attendees were reminded to complete the survey for a chance to win an iPod Nano. The official agenda scheduled a “Lucky Draw-Survey” slot from 16:15 to 16:30 on 21 May. These details establish an incentive and an administrative moment. They do not establish how many forms were completed because of the prize, whether the prize changed who responded, or whether it changed the direction of any answer.

A completion prize can be a reasonable practical tool. Meeting attendees are busy, forms compete with travel and closing sessions, and a draw can make a request visible. The existence of the incentive is not evidence of bribery or impropriety. It belongs in the methods record because it may affect participation. But the possible effects run in more than one direction. A prize could bring in people who otherwise would not respond and thereby widen the respondent mix. It could disproportionately attract people motivated by the draw. It could have no meaningful effect at all. The record cannot choose among those possibilities.

Nor does the incentive tell us whether answers became more positive or negative. Respondents seeking entry in a draw might rush through questions, provide considered criticism, or behave exactly like respondents without an incentive. Without a comparison group, timing data, completion-quality measures, or respondent characteristics, direction and magnitude remain unknown. The honest statement is narrow: AFRINIC used or announced a chance to win an iPod Nano in connection with survey completion, and the available report does not reveal the effect.

Documenting the incentive is part of methodological transparency, not a moral verdict. A future report should say when the draw was announced, the eligibility rule, whether entry required a complete form, how duplicate entries were prevented, and whether responses could be separated from prize identification. Those details would help readers judge participation mechanics without making unsupported claims about respondents' motives.

The information chain from experience to institutional quotation

A survey answer does not travel directly from a person's experience into an institutional conclusion. It passes through a sequence of choices, each capable of preserving or losing information. The first link is the respondent's own experience and interest: which sessions they attended, which services they used, what expectations they brought, and what they cared enough to report. The second is the questionnaire: wording, response options, order, scales, language, accessibility, and the visibility of the incentive shape what can be expressed.

The third link is administration. Someone must define eligibility, distribute or deliver forms, accept returns, distinguish starts from completions, handle missing items, and decide whether records are unique. The fourth is AFRINIC's collection and validation process, including any duplicate checking and exclusions. The fifth is analysis: free-text answers may be coded; categories may be combined; item denominators selected; percentages rounded. The sixth is editorial: a meeting report cannot reproduce every answer, so staff select, compress, and frame results.

The seventh is later citation, when a sentence can be detached from the methodological context that once bounded it. The eighth is any claim that the findings support service quality, policy preference, community consensus, or authority.

Information can be lost at each transition without anyone acting in bad faith. A fixed response scale can omit an experience. A skipped item can disappear from a percentage. A rounded figure can conceal a small denominator. An editorial summary can preserve the headline but omit disagreement. A later speaker can remember the satisfaction number and forget that it came from respondents. The point of mapping the chain is not to allege manipulation, fabrication, or statistical misconduct. It is to show precisely why a conclusion cannot be stronger than the records that connect it to the underlying people.

The report gives readers the final summary and some aggregate figures, but it leaves much of this chain undocumented. The questionnaire and exact wording are not supplied in the sealed record. Survey language, accessibility, and response mode are unknown. The numbers of forms distributed and delivered are unknown. Eligibility and the sample frame are unknown. So are anonymity arrangements, identity linkage, unique-response checks, raw data, missingness, exclusions, free-text coding, survey-section authorship, and the exact post-meeting publication date. Those are not defects that must be filled by speculation.

They are boundaries around the conclusions.

A rigorous account also distinguishes an unknown from a negative finding. “The report does not establish that all 135 were invited” does not mean some were excluded. “Unique respondents are not documented” does not mean duplicates existed. “The coding rules are unavailable” does not mean comments were distorted. Each statement locates missing evidence; none supplies the missing fact. This discipline matters because an audit can become as misleading as promotional prose if it converts documentary silence into accusation.

Selection uncertainty points both ways

Voluntary completion permits self-selection. People who were highly satisfied might have wanted to thank the organisers. People who were highly dissatisfied might have been more motivated to document problems. Prize-motivated people might have answered regardless of their experience. People with more time, greater comfort in the survey language, stronger institutional engagement, or easier access to the form might have been over-represented. None of these plausible mechanisms is demonstrated as dominant.

The uncertainty begins even before the survey. Travel funding, professional role, language, timing, and the ability to attend an event in Cairo shaped the set of people counted as participants. Session attendance then shaped which services or presentations a person could evaluate. The report provides no cross-tabs by these dimensions. It therefore cannot support weighting from respondents to all event participants, and still less from meeting participants to AFRINIC membership, affected operators, end users, Africa, or any other population.

Nonresponse is not a hidden ballot box. The 93-person difference between the published event count and respondent count does not consist of verified nonrespondents under a shared survey frame, because universal eligibility and invitation are not established. Even if it did, silence would not disclose belief. Those outside the respondent count might have agreed, disagreed, lacked time, missed the request, declined the prize, completed only an uncounted portion, or not been eligible. The evidence does not identify them or their reasons. Their absence widens uncertainty; it does not cast 93 imaginary votes.

This two-directional uncertainty is a reason to retain the respondent data, not discard it. If selection had an unknown direction, the results remain observations from actual respondents. They can reveal possible problems and support questions for a later instrument. What they cannot do is estimate the prevalence of those views in a broader population. The report is strongest when read as a map of issues encountered by a bounded group, weakest when read as a census of collective will.

The strongest defence deserves full credit

The best defence of AFRINIC's survey is straightforward: it was plausibly a practical event-management exercise, not an election, referendum, or scientific poll. Meeting organisers routinely ask attendees about venue, registration, content, formats, communications, and sessions. Forty-two direct responses can be useful even when they are not representative. The presence of sharp website criticism shows the published summary was not merely applause. The exact 22/19/one membership-status split shows some attention to respondent composition. A prize draw can reasonably raise completion in a busy closing schedule.

That defence succeeds on operational value. If a service team learns that 95 percent of people answering an item saw a need for website improvement, it should investigate. If respondents report that their attendance goals were fulfilled and approve of a format, organisers should preserve what worked while testing whether the finding recurs. It would be wasteful to reject those signals simply because the exercise cannot estimate a continental population. Service improvement often begins with imperfect but concrete feedback.

The defence fails only if it asks operational value to carry constitutional weight. A useful survey does not become an authorisation instrument through praise, repetition, or time. Forty-two respondents can describe their experiences. They cannot authorise on behalf of people who did not answer, organisations whose representatives were not verified, operators outside the meeting, or an undefined community. Even unanimous satisfaction among all 42—something the report does not claim—would not supply the separate legal authority required for governance, custody, expenditure, coercion, or control.

This is not an argument against participation. Participation supplies evidence, expertise, warnings, and practical knowledge. Consensus can express technical judgment among people engaged in a process. Neither is sovereignty. “Community” can describe a network of people and institutions, but it is not itself a principal capable of owning assets, delegating coercive power, or binding absent persons without a defined legal mechanism. The survey's proper dignity lies in being what it is: a service-feedback instrument with human observations, not a miniature constitution.

The bookkeeper boundary

AFRINIC's operational role can matter greatly without becoming sovereign. A registry records and coordinates number-resource administration. Its databases, processes, and recognised records can influence how organisations plan and maintain infrastructure. That influence rests on service, coordination, recognition, and trust. It does not amount to control over routing reality, ownership of the Internet, or inherent public authority.

The governing boundary is categorical. AFRINIC remains a private bookkeeper and coordinator without sovereign, legislative, regulatory, police, prosecutorial, judicial, punitive, confiscatory, or ownership power. It cannot manufacture those powers by calling a responding group “the community,” by describing a policy as consensus-based, or by collecting evidence of satisfaction. A private ledger may be operationally important; importance is not sovereignty. A coordinator may convene expertise; convening is not authorisation over absent principals.

Official AFRINIC sources can prove what AFRINIC recorded, said, counted, or did. They cannot self-prove community representation, consensus, mandate, legitimacy, ownership, or legal authority merely by using those terms. This evidentiary rule applies to the 2009 report itself. The report proves the stated event count, respondent count, split, and published percentages. It does not validate the larger institutional meanings that later speakers might attach to them.

The distinction protects both users and the registry. Users receive clearer claims about who actually expressed a view. The institution avoids building decisions on an imaginary principal. Staff can act on service feedback within delegated operational roles while keeping questions of corporate authority, legal power, and asset control attached to their proper instruments. A survey need not be belittled because it cannot do a court order's, member resolution's, contract's, or authorised vote's work.

A denominator ledger that would make the survey stronger

The missing record can be described concretely. Begin with registered participants, then distinguish people actually present by day and, where relevant, by session. Record the number of survey invitations or forms issued, the number delivered, the eligible recipient set, delivery failures, starts, unique complete responses, partial responses, exclusions, and duplicate checks. For every item, publish the exact number who answered, the number blank, the numerator for each option, and the denominator used for the displayed percentage.

The ledger should also state the questionnaire wording and response options, field dates, survey mode, available languages, accessibility arrangements, and the incentive rules. It should explain whether membership status was self-declared or verified, how one missing membership answer was treated, and whether privacy-safe member/non-member cross-tabs were produced. Free-text coding rules should identify how comments became themes, whether more than one coder reviewed them, and how staff selected quotations or summaries. None of this requires exposing respondents' identities.

Raw aggregate counts should sit beside rounded percentages. A line such as “numerator/answered item, with blanks shown separately” would prevent readers from assuming that every percentage used 42. Publishing the blank questionnaire would let later analysts distinguish between a broad satisfaction question and a precise service question. A clear methods note would say that findings apply among respondents and that the instrument was not designed to prove a governance mandate.

Where privacy-safe attendance information exists, organisers could compare respondents and nonrespondents on limited, relevant dimensions such as day of attendance or broad participant type. Such a comparison should never be invented where records do not exist, and it should not expose individual behaviour. Its purpose would be to detect obvious coverage differences, not to declare representativeness. If no comparison is possible, the report should say so.

Finally, service satisfaction and governance authority should be separated at design stage. An event survey can ask whether registration worked. It should not be asked to ratify policy, corporate control, or legal power. If an authority-sensitive question must be studied, independent administration, a defined eligible principal, identity and delegation rules, an auditable denominator, and a separate legal basis are required. The authority chain must be published independently because no satisfaction percentage can supply it.

What the 2009 report actually establishes

The defensible conclusion is narrower and more useful than either celebration or dismissal. AFRINIC recorded an event with 135 participants and summarised responses from 42 survey respondents. The crude count ratio is 31.111111... percent; 93 people, or 68.888888... percent of the event count, fall outside the respondent count comparison. The report recorded 22 member answers, 19 non-member answers, and one omission. It published several positive meeting signals and several service concerns, including a very strong reported desire for website improvement.

The report also records a completion incentive and scheduled draw. It does not establish the survey frame, universal invitation, unique response status, item denominators, raw counts, question wording, missingness rules, respondent representativeness, or direction of self-selection. Those unknowns mean the instrument cannot prove a response rate in the technical sense, a community verdict, or a governance mandate. They do not erase the respondents' experiences.

The lasting lesson is methodological. Preserve the denominator and the noun it counts. Carry that identity through collection, coding, reporting, and later citation. Use service feedback to improve service. Demand a separate, documented authority chain for control. In that disciplined form, 42 answers are not embarrassingly few or magically many. They are 42 answers—capable of directing attention, incapable of appointing themselves as everyone else.