Summary

  • APNIC’s 2026 Survey asked about an additional fee only after a respondent had rated a proposed service High Value or Essential. The 53% for premium customer support therefore describes a selected subset, not all 757 people who answered the preceding value question.
  • The report gives only a cross-service Q24 base range of 427 to 630. Neither the report nor the appendix attaches the exact eligible base to the premium-support percentage.
  • No amount, currency, billing unit, service entitlement or funded alternative was shown. A service-by-service price-and-denominator receipt would keep an exploratory preference from becoming a purchase forecast.

A price question without a price can still tell you something. It can show that a service is worth investigating. It cannot tell you what people will buy.

That is the useful boundary around one of the most eye-catching results in APNIC’s 2026 Survey. Among respondents who were asked the follow-up, 53% said they would consider premium customer support valuable enough to pay an additional fee. It is the highest Yes result across six proposed services. It is also attached to no amount and no exact published denominator.

The result is not wrong. It is unfinished.

The best case for the question

The strongest defence of APNIC’s presentation appears in the report itself. It describes willingness to pay as “more measured”. It says the question went only to people who had already rated a service High Value or Essential. It publishes No and Don’t Know alongside Yes. And the wording is conditional twice over: the service would first have to be successfully developed, and the respondent is asked whether they would “consider” paying.

That is a reasonable discovery question. Before spending money designing a detailed offer, an organisation may want to know whether the idea of an extra charge is immediately rejected. A directional answer can justify a priced follow-up without pretending to be that follow-up.

Nor does APNIC claim in the captured material that 53% of its Members will buy premium support. The report offers an interpretation instead. Premium support received the lowest broad value rating among the six services but the highest conditional willingness to pay. The report suggests that an individualised benefit may fit a user-pays model more readily than capabilities respondents see as benefiting the community.

That interpretation is plausible. It remains an interpretation. The published question cannot reveal the price imagined by each respondent, the service level they expected or whether they preferred an extra charge to collective funding.

There are two populations, not one

The sequence begins with Question 23. It was put to APNIC Members or Account Holders and asked them to rate six potential services: data mapping and network views, BGP analysis and insights, premium customer support, IP-abuse mitigation, further routing-security improvements, and advanced Internet-industry data and insights.

The main report gives the Q23 base as 757. On the combined High Value and Essential measure, routing-security improvements led at 86%. IP-abuse mitigation followed at 84%, BGP analysis at 83%, data mapping at 81%, advanced industry data at 76%, and premium support at 62%.

Question 24 then changed the population. It was asked for a service only when the respondent had placed that service in one of the top two value categories. Someone who thought premium support had Moderate, Low or No Value did not enter the premium-support willingness question. The same branching rule operated separately for every service.

The report says the resulting bases varied from 427 to 630. Its chart does not say which exact base belongs to which service. Appendix B is thinner still: it labels the base “various” and presents the Yes percentages.

This matters before any arithmetic. Fifty-three per cent is not a share of all 757 Q23 respondents. It is not a share of the survey’s 1,385 valid responses. And it is not a share of the APNIC membership. It is the share saying Yes inside one undisclosed service-specific group of prior favourable raters.

The report supplies the full displayed balance. Premium support records 53% Yes, 27% No and 19% Don’t Know. Routing-security improvements record 52%, 29% and 20%. Advanced data records 50%, 29% and 20%; IP-abuse mitigation 50%, 31% and 20%; BGP analysis 49%, 32% and 19%; and data mapping 45%, 35% and 20%. Rounding explains why displayed columns need not total exactly 100.

Those comparisons are useful. But because the six denominators differ, a percentage ranking is not a count ranking. The chart does not tell us whether premium support produced the largest number of Yes answers. It tells us that it produced the largest Yes share among the people routed into its own follow-up.

The offer exists only as a blank

The questionnaire asks whether a successfully developed service would be valuable enough to pay “an additional fee”. It does not supply an amount or a currency. It does not say whether the charge is per account, user, ticket, resource holding, month, year or service tier. It does not define the support entitlement, response time, hours of coverage, escalation route or boundary between ordinary and premium help.

It also offers no priced alternative. Respondents were not choosing between a collectively funded baseline and a paid enhancement, between two packages, or between different fee levels. A Yes can therefore encompass very different positions: a small annual fee for a defined escalation channel; a substantial subscription for an operational service level; willingness to pay only if ordinary support remains unchanged; or a general belief that direct beneficiaries should contribute something.

Likewise, a No does not prove that a respondent sees no value. Everyone in the branch had already assigned High Value or Essential. No may mean the capability belongs in APNIC’s core service, that the imagined price is too high, that the service boundary is unclear, or simply that the respondent declines an undefined charge. Don’t Know is almost the only literal answer available to someone unwilling to invent the missing offer.

This is why “willingness to pay” is a dangerous shorthand. In economics, a demand observation binds a quantity to a price and a defined good. Q24 binds none of those elements tightly enough to estimate take-up or revenue. It records an unpriced conditional preference.

Give the 53% a receipt

APNIC does not need to withdraw the figure. It needs to preserve the conditions around it if the figure travels into a budget, product or governance decision.

The appropriate control is a service-by-service price-and-denominator receipt. For each proposed service, the record should state the Q23 base and response counts, the Q24 branch rule, the exact eligible base, and the Yes, No and Don’t Know counts. It should disclose any exclusions, weighting and rounding. That would make clear whether the six percentage rankings are also meaningful in absolute terms.

A priced follow-up should then define the offer: service version, entitlement, response commitment, price, currency, charging unit, billing period and tier. It should show the funded status quo so that respondents can distinguish paying for an enhancement from paying to retain something they believe is already core. The receipt should name the decision for which the evidence is being used and carry correction or supersession links if the offer changes.

None of this requires personal data. Aggregated counts and a versioned questionnaire are enough. If no priced follow-up has taken place, the record can simply say so and label the 2026 result for what it is.

The distinction protects APNIC as much as its Members. A future proposal may find strong demand at a specified price. It may show that a premium tier can subsidise shared services. Or it may find that support disappears once cost and scope are made concrete. Each outcome is legitimate evidence. The current 53% cannot choose among them.

Sources