Summary

  • NANOG disclosed that a scholarship administration contractor supplied the Scholarship Committee with the top 12 applications in 2023, after which the committee selected four named recipients. Twelve is therefore the size of a shortlist, not the total number of applicants.
  • Public records identify historical eligibility rules, recipients, several award values, an administrator, a budget constraint and later organisational changes. They do not disclose the intake and eligibility denominators, shortlisting method, committee weights, aggregate recusals, a bridge between different accounting descriptions, or the post-2024 decision owner.
  • Candidate-level privacy is a strong reason not to publish files, references, financial circumstances, individual scores or rejection explanations. It is not a reason to withhold aggregate stage counts, public criteria, decision ownership, award totals and carefully designed outcome measures.
  • A privacy-preserving annual decision ledger would make the allocation chain legible across changes in administrators and committees without turning unsuccessful applicants into public subjects.

A denominator in the middle

The most revealing number in NANOG’s public scholarship record is not four. It is 12.

Four is an output. NANOG’s 2023 annual report names four students selected for scholarships: Nate Sales, Jason Zheng, Arisa Chue and Any Zelaya Delgado. The same report names the three Scholarship Committee members—Jeff Budney, Michael Costello and Tony Tauber—and describes the committee’s purpose as reviewing and selecting recipients of scholarships offered each year. Those details give the public a final count, a committee roster and recipient names. They are meaningful disclosures, but the roster alone does not establish each member’s voting role in every decision.

Twelve sits one step earlier. The annual report says the committee reviewed the “top 12” applications supplied by NANOG’s scholarship administration contractor. In one sentence, the report exposes a division of labour that a winner list alone cannot show. The contractor supplied the set the committee would see; the committee then selected four names from it.

But 12 is easy to misread. It is not the number of people who applied. It is not the number whose applications were complete. It is not the number found eligible. It is not evidence that 12 people met a published scoring threshold. It is the size of the contractor-provided shortlist. The reviewed report does not say how many submissions existed before that shortlist, how many were complete, how many were eligible, or what operation turned the earlier population into 12.

That distinction is more than statistical housekeeping. Every denominator answers a different question. Total submissions indicate demand. Complete applications show how many people navigated the process successfully enough to be reviewed. Eligible applications show how many satisfied the threshold rules. A shortlist count shows the pool presented to a later decision-maker. Recipients show the final allocation. Dividing four by 12 would describe the committee’s selection rate from the shortlist. It would not describe an applicant’s overall chance of receiving an award. Without the earlier counts, the latter cannot be calculated.

The 2023 record also leaves the nature of “top” undefined. A ranking might be based on a numerical score, a categorical assessment, an eligibility screen plus judgement, or another process entirely. The sources reviewed for this article do not establish which. They do not publish criteria and weights, the committee’s vote method, a tie procedure, aggregate conflict declarations or recusals. The annual report separately identifies Michael Costello as a Board member and the committee’s liaison, but it does not say whether the liaison voted in every selection or how any conflict would have been handled.

None of those omissions proves a flawed process. They define what an outside reader cannot reconstruct.

It is therefore inaccurate to compress the chain into the sentence “NANOG chose four students from 12 applicants.” The first half erases the contractor’s role; the second turns a shortlist into an intake population. A more faithful reconstruction is narrower: a scholarship administration contractor supplied 12 applications described as the top applications; a three-member committee reviewed that set and selected four named recipients. Before the handoff lies a screening or ranking stage whose method is undisclosed. After it lies the announcement and administration of awards.

In the public record, control is distributed even though the output bears one organisation’s name.

This is the missing middle. NANOG tells readers enough to see that the middle exists, but not enough to audit its rule. The issue is not whether the four recipients deserved their awards; the reviewed evidence offers no basis for ranking candidates anew and no justification for questioning any named recipient. The issue is whether an institution can describe a finite opportunity in a way that lets an outsider distinguish eligibility, screening, shortlisting and final selection without exposing the people who were not selected.

That is a modest accountability question. A student scholarship does not confer public authority, and a recipient does not become a representative of applicants, students or NANOG. Yet material support and institutional recognition are scarce goods. When an organisation says those goods advance educational access or community benefit, stage-specific information helps readers understand whether the operating process matches the stated purpose. The need for that information becomes sharper when responsibility moves from one contractor to another, or from one committee to an unidentified successor.

Predictability without pretending selection is automatic

Process legibility matters before an award is made, not only when an annual report looks back. An application asks a person to decide whether an opportunity is real for them and whether preparing a submission is worth the effort. Published eligibility rules answer part of that question. They can tell a student whether their course, enrolment status and academic standing meet a threshold. A deadline and intake cap can tell them when to act. An award value tells them what is at stake.

Comparative selection rules answer a different part. A person can be eligible yet have no idea what the institution considers when it chooses among eligible candidates. That uncertainty may be inevitable at the margins, because qualitative judgement cannot be fully predicted. It does not have to be total. A programme can identify the dimensions that matter, whether any are weighted, which body performs each review and what evidence applicants should provide.

The point is not to promise an outcome. Clear criteria should never be marketed as a formula that guarantees an award. Nor should stage statistics be converted into personalised odds when cycles differ and candidate circumstances are not interchangeable. Predictability here means that an applicant can understand the procedure and the kinds of judgement involved, not that the applicant can calculate a certain result.

The distinction is particularly important when an administrator and a committee divide work. If the administrator checks only completeness and eligibility, the committee may see the full eligible pool. If the administrator also ranks and shortlists, much of the comparative decision occurs before committee review. Both arrangements can be legitimate. They place discretion in different hands, and the public description should reflect that.

For 2023, the phrase “top 12” suggests comparative ordering or filtering, but the reviewed source does not define it. It would go beyond the evidence to say which criteria the contractor applied or whether 12 was a fixed target. A future cycle could avoid that ambiguity by stating, before applications open, whether an administrator performs an eligibility screen, a scored review, an unweighted qualitative review or some combination. After selection, aggregate counts could show how the stated process operated.

Funnel figures can also help an institution diagnose its own process, though no single pattern proves a cause. A large difference between submissions and complete applications could trigger a review of instructions or required materials. A large difference between complete and eligible applications could trigger a review of how clearly threshold rules were communicated. A stable eligible pool combined with fewer awards might reflect a budget decision. Those are questions the data could trigger, not conclusions that can be applied retrospectively to NANOG’s undocumented stages.

The ledger should therefore resist the temptation to label every reduction “rejection.” An incomplete submission, an ineligible application, an administrator shortlist decision and a committee non-selection are different events. They may carry different notice rules and different opportunities for procedural improvement. Using one verb for all of them hides the division of control that the 2023 disclosure begins to reveal.

Applicant-facing information can remain concise. A public notice does not need to include internal discussion. It could state that eligibility is checked by the administrator, that eligible applications are reviewed against named dimensions, that a shortlist of a defined size or range is sent to a committee, and that the committee uses a stated decision method. The annual ledger would then report aggregate results against that advance description.

This model treats transparency as part of programme design rather than a response to controversy. It also reduces the pressure to interpret winner identities as clues. When the rule and stages are visible, applicants and readers can direct legitimate questions to the appropriate institutional owner. When they are not, the visible people at the end of the process can become proxies for an invisible procedure.

What the winner lists reveal—and what they cannot

Winner publication is real transparency. It tells a community that awards were made, supplies a count and, when values are included, gives the public a direct view of the material commitment. Across the records reviewed here, NANOG repeatedly named recipients rather than describing scholarships only in the abstract.

The 2019 annual report named Celine Irvene, Sydney Pugh, Daniel Albrecht and Chinasa Okolo. It placed the awards under the Abha Ahuja and John Postel scholarship names and framed the programme as an effort to engage more students, women and people of colour in scholarships and educational programmes. The 2021 report named Esu Ekeruche, Wendy Ruan, Juan Perez and Charly Guttierrez Jimenez and described the two scholarship types as USD 10,000 awards. The 2022 report named Katherine Coward and Angelina Xu, identified their fields and institutions, and said each received USD 10,000. The 2023 report again named four recipients.

These disclosures matter because they prevent the programme from becoming a promise without observable outputs. They also put human experience into a record otherwise dominated by organisational prose and financial categories. In 2021, Wendy Ruan said the award allowed her to replace an inadequate laptop and continue her studies with fewer worries. That testimony shows one concrete way an educational award can matter: an obstacle that could make study harder was reduced for one recipient.

It would be a mistake, however, to turn one statement into a programme-wide outcome estimate. It does not show how the average recipient used an award, how many recipients graduated, whether recipients later worked in network operations, or whether the programme caused a change in career direction. The public materials reviewed for this article do not supply those measures. Nor does a list of names reveal the demographic composition or financial circumstances of the applicant population. A stated purpose of engaging more women and people of colour is an organisational purpose claim; it is not, by itself, measured proof of demographic effect.

The same discipline applies to institutional recognition. Being named as a scholarship recipient does not make a student a delegate, member, speaker, volunteer or representative. The public record reviewed here does not establish that recipients possessed voting authority, owed political loyalty or were selected to speak for a constituency. Scholarship is access to support, not a transfer of mandate. Treating a recipient list as evidence of representation would burden awardees with a governance meaning the programme record does not give them.

Winner lists also cannot show who encountered the application process but did not reach the final stage. The absence is legitimate at the level of names. Unsuccessful applicants should not be turned into an involuntary public list. Yet the absence of names need not require the absence of counts. A programme could report, for example, that it received a certain number of submissions, found a certain number complete and eligible, shortlisted 12 and made four awards. That would allow a reader to see the funnel while protecting every person outside the winner list.

The distinction between output transparency and process legibility is central. Output transparency asks what was awarded and to whom. Process legibility asks how the candidate population moved between defined stages, who controlled each movement and what public rule applied. Neither substitutes for the other. A process can be well described without naming unsuccessful applicants, and named winners do not reveal how the process worked.

This distinction also protects recipients. When criteria and stage counts are absent, observers may be tempted to infer a selection rule from visible names, universities, majors or photographs. Those inferences are unreliable and unfair. They risk converting the identity of a successful applicant into speculation about the hidden process. A published criteria framework and aggregate funnel would direct scrutiny toward the institution’s procedure, where it belongs, rather than toward the individuals who received support.

The reviewed record gives no evidence that the 2023 recipients were chosen because of race, gender, school prestige, relationships or Board influence. It gives no basis for saying the contractor treated candidates unfairly. It also gives no demographic selection rates. A serious investigation does not fill those spaces with suspicion. It marks them as unknown and asks what information could make the process testable without creating a privacy cost greater than the accountability benefit.

From a launch announcement to historical rules

The programme’s public trail begins with a broad announcement. In May 2017, NANOG told its attendee mailing list that the Network Operators Scholarship programme had launched in April. The first round was scheduled to close on 2 June, with awards to be made in September. The message established a starting point and a timetable. It did not publish the eventual applicant count, shortlist, award value or selection reasons.

By 2019, an archived programme page made several rules much more concrete. The page described an educational scholarship for current undergraduate and graduate students in specified computing, networking, electrical-engineering and telecommunications fields. It set a minimum grade-point average of 3.0 on a 4.0 scale, or the equivalent. It required at least six credits or full-time undergraduate or graduate study at an accredited institution for the relevant academic year.

The page also described a one-time USD 10,000 award, with up to four awards available. Past recipients could not reapply under those terms. Scholarship America administered the programme on NANOG’s behalf. The application deadline was 16 April 2019 or the first 50 submissions, whichever came first. The page said awards were made without regard to a list of protected and socioeconomic characteristics.

Those details are valuable because they show what explicitness can look like at the front of an allocation process. A prospective applicant could determine whether their field, enrolment and academic standing appeared to qualify. They could see the award value, the maximum number of awards, the one-time nature of the grant and the identity of the administrator. A deadline tied to an intake cap also told applicants that timing could matter.

But the “first 50 submissions” clause requires the same denominator discipline as the 2023 top 12. Fifty was an intake ceiling, not a reported result. The page does not prove that 50 submissions arrived. It does not show how many were complete or eligible, and it cannot be projected into later years. It would be wrong to use the 2019 ceiling as an estimate for 2023 demand or to assume that later cycles preserved the same cap.

The historical page also stops before the decisive middle. It does not publish a scoring framework, weights, reviewer identities for that year, a recusal process or an appeal route. Visible eligibility is not the same as visible comparative selection. The 3.0 floor could tell a student that an application was allowed into consideration, but it could not tell the student how academic record, field, experience, financial circumstances, references or other factors would be weighed after eligibility. The reviewed material does not establish what those factors were.

That is not an argument that every judgement must be mechanised. Scholarships often require qualitative assessment. A rigid score can create a false appearance of objectivity, and some legitimate considerations are difficult to reduce to a number. The accountability requirement is not necessarily a spreadsheet with mathematical precision. An institution can state that review is unweighted, explain the dimensions considered and describe how reviewers reach a decision. What matters is that “top” has a publicly intelligible meaning appropriate to the award.

Nor can the historical 2019 rules be presented as current policy. Archives preserve evidence of what an organisation once said, not a guarantee of present availability or terms. By 2026, the reviewed resources page does not state a current award value, application window, administrator, decision owner or annual applicant statistics. Any account of the programme must therefore date the rules it describes. “In 2019, the archived terms said” is accurate. “NANOG requires” would overstate the record.

The changing terminology makes dating even more important. The 2019 page describes educational awards to students. NANOG’s resources page groups educational programmes, scholarships and fellowships as mechanisms created to help people who lack resources or financial support attend a NANOG meeting. Those descriptions may concern different programme types or shorthand, so they cannot be collapsed into a single product without further evidence. Education funding, travel reimbursement and meeting attendance support can have different eligibility rules, decision owners, accounting treatment and intended outcomes.

Clear programme typing should therefore be part of the public record. Before publishing statistics, NANOG would need to say whether a cycle concerns an education award, a meeting-access fellowship, travel support or another form of assistance. Otherwise a total applications number or an expense line could combine unlike programmes, making an apparently precise disclosure less informative.

Continuity, benefit and a shrinking award count

The named-recipient record provides a partial view of continuity. Four people were named in 2019, four in 2021, two in 2022 and four in 2023. That sequence shows awards in multiple years, but it does not by itself explain changes in volume. A lower count could reflect budget, demand, eligibility, programme design, timing or another cause. Only where NANOG states a reason should a reader assign one.

For 2022, NANOG did provide a budget explanation in its federal information return. The Form 990 said awarded grants were lower that year because of budgetary constraints. That sentence connects the smaller grant activity to a stated institutional condition without requiring outsiders to infer motive from the recipient list. It is a useful example of disclosure that is both short and explanatory.

The annual report, meanwhile, gives the two visible outputs. Katherine Coward, studying computer engineering at Widener University, and Angelina Xu, studying computer science at MIT, were named as recipients of USD 10,000 each. Those facts show who was publicly recognised and the stated value of each award. They do not show whether the applicant pool shrank, whether selection standards changed or whether qualified candidates went unfunded. The Form 990’s budget explanation says grants were lower, not how many applicants existed at any stage.

This is where a stage ledger would add analytical value. If total and eligible applications remained stable while awards fell from four to two, the record would show a change in the allocation ratio associated with a budget limit. If demand also fell, the interpretation would differ. If rules changed before selection—as the filing says programme review and changes occurred annually—the ledger could identify which version governed the cycle. None of those scenarios can be chosen from the current sources, but they illustrate why a recipient count alone cannot carry the full explanatory load.

The impact side is similarly bounded. A USD 10,000 education award can be material, and the 2021 laptop account provides a vivid example. But the sources do not establish recipient income, employer support, institutional resources or financial-need distributions. They do not show the counterfactual—what a recipient would have done without the award. An institution can honestly state that a recipient reported replacing an inadequate laptop. It cannot turn that individual account into a causal claim that the programme transformed careers or built a workforce without a broader outcome design.

That restraint does not diminish the recipient’s testimony. It locates its evidentiary value correctly. Qualitative accounts can reveal mechanisms that aggregate figures miss: equipment can be a constraint, an award can reduce worry, and educational continuity can depend on practical resources. A good outcomes record could combine voluntary stories with aggregate, consent-based follow-up at defined intervals. It should not make continued publicity a condition of receiving help, nor should it assume silence means failure.

Public purpose language deserves the same care. The 2019 annual report connected the programme to engaging students, women and people of colour. A decision ledger could help evaluate whether that purpose was operationalised, but only if the organisation can collect and publish relevant data lawfully, ethically and with sufficient group sizes. The reviewed record does not contain applicant demographics or selection rates, so it cannot establish whether the programme reached those groups proportionately. Names and photographs are not demographic data and should not be used as proxies.

The better question is not “Did the winners look like the purpose statement?” It is “What evidence did the institution choose to collect to test its purpose, and what can it responsibly disclose?” An organisation may decide that certain sensitive characteristics should not be collected at all. If it does collect them, small-cell suppression and consent rules matter. If it does not, it should avoid presenting a broad purpose claim as a measured outcome.

This approach keeps scrutiny proportionate to the decision. The award is meaningful, but it is not a public election. The institution need not expose every deliberation to demonstrate integrity. It can make the governing rules, aggregate pathway, annual constraints and outcome method visible. Those are institutional facts. They do not require a profile of the people who did not receive support.

The 2022 accounting bridge that the public record lacks

The 2022 annual report and the 2022 Form 990 describe the scholarship programme in different reporting units. Reading one as if it were a direct cash ledger for the other produces a false contradiction.

The annual report says two recipients each received USD 10,000. That is a programme-output description: two named people and an award value attached to each. The Form 990 describes USD 26,450 of scholarship programme-service expense, including grants of USD 10,000. Its Schedule I lists a USD 10,000 cash grant to Scholarship America for scholarships. The filing also says NANOG hired Scholarship America to manage the scholarship programme and monitor the use of grant funds.

It explains that awarded grants were lower in 2022 because of budgetary constraints and that the programme was reviewed annually, with changes agreed before the current year’s selection process.

These figures do not visibly reconcile on the face of the reviewed documents. But “do not visibly reconcile” is a description of the public presentation, not an allegation that either document is false. Programme-service expense can include categories and timing that differ from a recipient announcement. A payment to an administrator can reflect a grant, an administrative arrangement, timing across reporting periods or another accounting treatment. The sources reviewed here do not supply the detailed bridge needed to choose among those explanations.

It would therefore be wrong to say that the filing proves only one USD 10,000 award was paid. It would also be wrong to add the visible numbers mechanically or to infer that recipients were underpaid. The annual report and tax return answer different questions. One tells readers about selected recipients; the other reports organisational expenses and a transaction with an administrator under regulatory categories.

The gap is nevertheless relevant. NANOG is a 501(c)(3) identified by ProPublica’s Nonprofit Explorer under EIN 27-2534183, and its filings give the public an independent route into its financial history. The full filing, rather than an extracted database summary, controls precise claims. Readers following that route can find the Form 990 language, but they still need an explanatory note if they want to connect programme outputs with financial reporting.

A concise accounting bridge could solve the problem without publishing confidential detail. For each cycle, NANOG could report the number and stated face value of awards, the total authorised award value, the amount recognised as scholarship expense in the relevant financial year, and any material difference caused by timing, administration or classification. It could name the accounting period rather than assuming a programme year and a fiscal year align. If an administrator receives funds before or after recipient announcements, the note could state that at an aggregate level.

Such a bridge would also prevent the administrator’s role from being misunderstood. The filing says Scholarship America managed the programme and monitored grant funds in 2022. That is evidence of management and monitoring for that year. It does not establish that Scholarship America was the unnamed contractor that supplied the 2023 top 12. Although the 2023 annual report later records a Board decision to end the relationship with Scholarship America, the reviewed shortlist sentence does not itself identify the contractor. Chronological proximity is not textual identification.

This caution is especially important in a transition year. A programme can involve application processing, selection, award notification, disbursement, monitoring and accounting on different dates. A Board can decide to change an administrator while work associated with an earlier cycle is still underway. The 2023 report does not establish whether the four-recipient selection occurred before or after the October decision to end the Scholarship America relationship. A public ledger should attach dates and owners to stages rather than assigning an entire cycle to one undifferentiated year.

Accounting transparency and selection transparency are related but distinct. Financial statements can show that money was recognised; they do not reveal how candidates reached a shortlist. A selection record can show who decided; it does not necessarily show when expense was booked. Good accountability requires a bridge, not a merger. The institution should let each record keep its proper unit while explaining how the units connect.

A chain of institutional changes, not proof of a programme ending

The later record contains three facts that naturally invite a single story. In October 2023, according to the annual report’s Board-actions summary, the Board approved ending NANOG’s relationship with Scholarship America. In the October portion of the 2024 annual report’s Board timeline, the Board approved dissolving the Scholarship and Education Committees. The audited 2024 functional-expense schedule reports no scholarship expense, while its comparative 2023 schedule reports USD 40,000 of scholarship expense under other programmes.

Placed in sequence, the facts look like a winding-down: administrator relationship ends, committee dissolves, expense becomes zero. The sources do not establish that causal chain.

The 2023 Board-actions entry does not identify a replacement administrator, transition terms or the precise relationship between the October decision and that year’s selection. Ending one contractor relationship does not establish that administration ceased rather than moved. The 2024 committee-dissolution entry does not say the scholarship programme ended. A Board or staff team could assume a committee’s work; a new committee could be created; an award could be paused; or the function could remain unresolved. Those are possibilities, not findings. The source does not identify who, if anyone, inherited selection authority.

The audited zero line is similarly bounded. It establishes that the functional-expense schedule showed no scholarship expense for 2024. It does not establish that every kind of assistance stopped, that no award decision occurred, or that related activity could not appear in another account. It also does not explain why the line was zero. The comparative USD 40,000 for 2023 confirms a financial change in the schedule, not the organisational cause of that change.

The public record therefore supports a transition sequence but not a termination conclusion. It would be inaccurate to treat these three facts as proof of a 2024 termination, to make the committee’s dissolution the established cause of the zero, or to treat that expense line as a census of every form of financial assistance. Each interpretation adds a proposition the evidence does not provide.

This is where decision ownership becomes more than an organisational-chart question. In 2023, the disclosed chain contained an administration contractor and a Scholarship Committee. The Board made decisions affecting the contractor relationship and later the committee itself. If the award remained available, who defined eligibility, accepted applications, screened them, made final selections, authorised payments and monitored funds after those changes? If the award paused, who made that decision, and under what programme authority? If it changed type, how should earlier student-education awards be distinguished from meeting-access support?

A durable ledger would preserve answers even when the organisational form changes. It would record the legal or organisational owner of the programme, the outside administrator for a given cycle, the body or officer responsible for final selection and the owner after any mid-cycle change. The record would not assume that a committee name equals practical control. It would specify what each actor did.

This matters because responsibility can otherwise disappear between accurate fragments. The Board can truthfully report that it ended a contractor relationship. The annual report can truthfully report that a committee was dissolved. The audit can truthfully report a zero expense. Yet no single fragment answers the reader’s operational question: who owned the next decision? A stable record should make continuity—or a deliberate pause—visible without requiring outsiders to infer it from adjacent events.

The record should also distinguish a decision date from an accounting date. A 2024 expense line might reflect decisions made in another period, and a decision made in 2024 might affect a later financial year. Recording opening, closing, selection, authorisation, disbursement and recognition dates at an aggregate level would prevent a calendar sequence from being mistaken for causation.

Institutional change is not inherently suspicious. Ending a vendor relationship can improve a programme. Dissolving a committee can reduce duplication or assign work more effectively. A zero expense can reflect a lawful, deliberate budget choice. The critique is not that change occurred. It is that the reviewed public record does not connect the changes to a named successor owner or a clear programme status. A decision ledger would allow NANOG to explain change in its own terms rather than leave readers to assemble a story from gaps.

The strongest case for the existing level of disclosure

Any proposal for more transparency should begin with the reasons an institution might resist it. Scholarship administration handles sensitive material. Applications may contain academic records, references, statements of need, personal histories and information about family or financial circumstances. Publishing such material could harm unsuccessful applicants, discourage applications and create legal or ethical risks.

Selection is also unavoidable when funds are finite. A programme offering up to four awards cannot support every eligible applicant if demand exceeds supply. The existence of discretion is not evidence of abuse; it is often the mechanism by which different strengths and circumstances are considered. A committee may need room for contextual judgement that a rigid formula cannot capture.

An outside administrator can serve legitimate purposes. It can provide specialised systems, consistent processing, privacy controls, eligibility review, tax handling and grant monitoring. It can reduce the burden on volunteer committee members and separate some administrative work from final judgement. NANOG’s 2022 Form 990 says Scholarship America managed the programme and monitored grant funds, functions that can support integrity rather than obscure it.

The historical rules gave applicants several useful facts: a grade threshold, enrolment requirement, relevant fields, award amount, maximum award count, one-time status, application timing and administrator identity. Annual reports named recipients and published values in several years. The Form 990 acknowledged a budget constraint. Annual reports disclosed the Board’s administrator and committee decisions. The audited statements placed scholarship expense into a comparative schedule. This is not a record of complete silence.

Candidate-level disclosure would add little institutional accountability while imposing large privacy costs. Publishing individual scores could let readers identify unsuccessful applicants in a small field. Publishing references would violate the expectations under which they were supplied. Publishing financial circumstances could stigmatise applicants. Publishing rejection reasons could expose protected information, invite harmful comparison and freeze a contextual judgement into a permanent public label. Committee minutes that discuss individuals could be equally intrusive.

There are also limits to outcome measurement. Contacting former recipients repeatedly can turn a one-time award into a lasting reporting obligation. Employment and educational paths reflect many influences, so simple claims of programme causation can be misleading. Small recipient cohorts make aggregate demographic reporting prone to re-identification. An institution may reasonably decide not to collect some sensitive data.

These are strong arguments. They establish the boundary of a proportionate remedy. The public does not need application files, individual scores, personal finances, references, protected characteristics, named conflict disclosures or candidate-specific rejection reasons. It does not need to reverse-engineer why one student was chosen over another. It needs a view of the institution’s rule and aggregate pathway.

Aggregate disclosure can be designed around the defence rather than against it. Counts can be suppressed when a category is too small. Criteria can be described without publishing an applicant’s response. A committee can report its size and aggregate recusal count without naming the person who stepped aside or the candidate involved. Outcomes can be voluntary and reported only in sufficiently large groups. The institution can publish what it will never disclose, making privacy a visible rule rather than a generic explanation for every gap.

The strongest defence also suggests a role for clear uncertainty statements. If a legal, contractual or privacy constraint prevents publication of a particular aggregate, NANOG could name the class of constraint and explain why it applies. That would materially weaken the critique. A reader does not need to assume bad faith when an institution identifies a specific boundary. What creates unnecessary ambiguity is treating all stages as if candidate privacy forbids even non-identifying totals and rules.

The goal is not radical transparency. It is reliable, bounded transparency: enough to reconstruct institutional control and stage movement, not enough to reconstruct an individual application.

Designing a privacy-preserving decision ledger

A decision ledger should be an annual, structured account of the programme rather than a narrative assembled after questions arise. Its value would come from consistent fields, stable definitions and dates. The design should survive a vendor change, committee dissolution, budget reduction or change in award type.

1. Name the programme and its purpose

The ledger should begin by identifying the award and stating whether it supports education, travel, meeting attendance or another purpose. This prevents student scholarships and meeting fellowships from being combined merely because both broaden access. It should record the award’s stated public purpose without presenting the purpose as an achieved outcome.

If NANOG offers more than one scholarship—such as awards bearing the Abha Ahuja and John Postel names—the record should explain whether they share one application pool and selection process or operate separately. Counts mean different things in a combined pool and in two distinct competitions.

2. Identify the owner and administrator

For each cycle, the ledger should name the organisational owner, outside administrator and final decision-maker. It should describe functions: who opened the application, checked completeness, tested eligibility, produced a shortlist, selected recipients, authorised funds, disbursed or monitored grants, and published results.

If the Board changes a contractor or dissolves a committee, the same record should name the effective date and successor owner for each affected function. If no successor exists because the programme is paused, it should say so. This would address the current post-2024 ownership gap without requiring speculation about why the change occurred.

3. Publish the calendar and version of the rules

Opening date, deadline, any intake cap, selection date, notification date and relevant accounting period should be recorded separately. A version identifier for the eligibility and selection rules would show which rules applied. The 2019 “first 50 submissions” clause demonstrates why an intake cap must be distinguished from actual intake.

If rules change annually, as the 2022 Form 990 says, the ledger should record what changed before selection. The purpose is not to expose legal advice or internal drafting. It is to ensure applicants and later readers know which public criteria governed a cycle.

4. Count each stage with stable definitions

At minimum, the ledger should report total submissions, complete applications, eligible applications, shortlist size and award count. Definitions should state, for example, whether a person who began but did not submit is included, how duplicate submissions are handled and what makes an application complete.

The counts should be displayed as a funnel, not one denominator. For 2023, the known entries would include a shortlist of 12 and four recipients. The unknown earlier cells should remain explicitly unavailable unless NANOG can recover them. No one should backfill total applications with 12.

5. Describe criteria and decision method

The public record should list the dimensions reviewers consider and any weights. If review is deliberately unweighted, it should say that and explain the method used to reach a decision. If the administrator ranks applicants before committee review, the record should say whether the ranking is numerical, categorical or based on another rule.

The ledger need not publish candidate scores. It should provide enough information for an applicant to understand what “top” means and for a reader to know whether the committee applied the same or a different judgement at the final stage.

6. Record committee structure and aggregate recusals

Committee size, appointment source and decision method can be reported without candidate detail. The ledger could state whether decisions were made by majority vote, consensus or delegated authority. It could report the number of recusals in aggregate and whether a recusal reduced the voting body or triggered a replacement reviewer.

If the Board liaison has a different function from ordinary committee members, that distinction should be explicit. The current record names a liaison but does not establish voting status in each decision. A role description would prevent readers from inferring authority from title alone.

7. Connect awards to accounting

The record should state the number of awards, face value per award, total authorised award value, amount recognised as scholarship expense in the named period and any material timing or classification difference. It should separate payments to recipients from transfers to an administrator where the accounting treatment differs.

This is the missing bridge in the 2022 presentation. A short note could explain why a programme announcement and a Form 990 line use different units without disclosing recipient banking or tax information.

8. Set a restrained outcome method

Outcomes should be tied to programme purpose and collected with consent. Possible aggregate fields might include award acceptance, continued enrolment or completion at a defined interval, but only if those measures are appropriate, feasible and not coercive. NANOG should avoid using employment at a network operator, meeting attendance or volunteer service as a presumed measure unless the programme actually states those objectives.

Qualitative accounts can be included voluntarily, as the 2021 laptop statement was, but should not be presented as representative. The ledger should state response rates and avoid implying that non-response is a negative outcome.

9. Publish suppression and retention rules

Small groups create re-identification risk. The ledger should define a minimum reporting cell and combine or suppress categories below it. It should state that candidate files, individual scores, references, financial circumstances and candidate-specific reasons will not be published. Data retention should be limited to legitimate administrative and legal needs.

This privacy section would give applicants an affirmative assurance. It would also stop privacy from being invoked vaguely, because the line between aggregate institutional data and personal candidate data would be written down.

10. Preserve corrections and unknowns

The ledger should distinguish “zero” from “not collected,” “not available” and “suppressed.” These states are not interchangeable. A zero means a measured absence; not available means the institution cannot establish the value; suppression means it knows the value but protects it.

Corrections should preserve earlier versions and explain the change. If NANOG cannot reconstruct historical intake counts, the honest entry is “not available,” not an estimate built from the shortlist or a past intake cap.

Together, these fields would produce a record that is useful precisely because it is unglamorous. It would not adjudicate the merits of individual applicants. It would let the public see how a scarce institutional benefit moved from an announced purpose through rules, screening, judgement and accounting.

Do not turn the ledger into a league table

Structured data creates its own risk: figures designed to make one programme legible can be repurposed to rank institutions or declare a cycle fair. The ledger should include enough definitions and context to resist that misuse.

An application-to-award rate is not a fairness score. A low rate may reflect strong demand, a small budget or a narrow programme. A high rate may reflect lower demand, broader funding or strict self-selection before submission. Neither rate shows whether criteria were appropriate or consistently applied. Even within one programme, changes in outreach, award value, eligibility or intake caps can make annual rates incomparable.

Stage rates are more diagnostic but still incomplete. A high ineligibility rate could reflect confusing public rules, but it could also reflect applications submitted despite clear requirements. A shortlist rate could be determined in advance rather than emerge from scoring. A committee selection rate from a fixed shortlist says little about the earlier administrator decision. The ledger should publish raw counts, definitions and rule changes alongside any calculated percentage.

Outcome fields also need restraint. Continued enrolment, graduation or later employment can be influenced by many conditions beyond an award. A recipient account may illuminate how support was used without proving that the programme caused the later result. Comparisons between recipients and unsuccessful applicants would raise much greater privacy, consent and causal-design questions than the aggregate operating ledger proposed here.

The safest purpose statement for the ledger is operational: it shows what the programme offered, how applications moved through defined stages, who controlled the decisions, what was awarded and how the institution reported the expense. It may support evaluation, but it should not manufacture a performance grade from a few ratios.

This boundary also keeps the proposal proportionate to NANOG. The record does not establish a mandate to solve educational inequality or to represent an applicant population. NANOG can be held to the purposes it states and the procedure it operates without being made responsible for every later educational or labour-market outcome. A ledger should test consistency between promise, rule and action—not inflate a scholarship into a universal social programme.

Give the ledger an owner and a publication rhythm

A disclosure instrument can fail even when its fields are well designed if no one is responsible for completing it. The Board should therefore identify an organisational owner for the ledger itself. That owner need not be the selection committee. Indeed, separating record stewardship from candidate deliberation may make the privacy boundary easier to enforce. The steward’s task would be to collect aggregate stage information from the administrator and decision body, confirm definitions, connect award figures with the appropriate accounting period and publish the completed record on a predictable schedule.

The schedule should follow the decision closely enough to be useful while allowing time for notifications and privacy review. An initial entry could publish the programme purpose, rules, administrator and dates when applications open. A post-selection entry could add the stage counts, committee structure, aggregate recusals, award count and authorised value after recipients have been notified. A financial note could later connect those outputs to the relevant reporting period. Version dates would make clear that the three entries describe one cycle at different moments rather than three conflicting accounts.

Responsibility for a missing field should also be visible. If an outside administrator cannot provide a total because its system counted only completed submissions, the ledger should say which denominator is available and how it is defined. If a committee did not use numerical weights, it should say so instead of leaving an empty “weights” cell. If a value is awaiting the audited financial statements, the provisional record should identify the expected update rather than silently remaining incomplete.

This maintenance rule would be particularly useful during institutional transition. Ending an administrator relationship should trigger an export of aggregate cycle data and definitions before access is lost. Dissolving a committee should trigger a handover identifying where decision authority and record stewardship move. Neither action requires publication of candidate files. Both require the organisation to preserve enough institutional memory to explain its own programme.

Finally, the ledger should be easy to find from both programme and financial-reporting pages. Discoverability is part of accountability: a technically public table buried in an unrelated archive does little to connect applicants, recipients and financial readers to the same authoritative record. Stable annual entries, linked rather than overwritten, would allow future readers to distinguish a genuine rule change from inconsistent reporting.

How the ledger would change the reading of 2017–2024

Applying the proposed structure retrospectively shows both its utility and its limits.

For 2017, the ledger could record the April launch, the 2 June closing date and the planned September awards. The applicant, eligibility, shortlist and recipient fields would remain unavailable from the reviewed announcement. That explicit absence would be more accurate than leaving readers to assume no data existed or to search later winner lists for a denominator.

For 2019, the ledger could capture the historical academic and enrolment thresholds, applicable fields, USD 10,000 one-time award, maximum of four awards, Scholarship America’s administration and the deadline-or-first-50 intake rule. It could record four named recipients at the public-output stage. It could not honestly fill the actual submissions, eligibility or shortlisting fields from the sources reviewed here. The number 50 would remain labelled as a cap.

For 2021, it could record four recipients and the stated USD 10,000 scholarship value. It could link voluntary qualitative testimony to one recipient without turning it into a cohort metric. Earlier stage counts would remain unknown.

For 2022, it could record two named awards of USD 10,000 each, NANOG’s statement that grants were lower because of budgetary constraints, Scholarship America’s programme-management and grant-monitoring role, the USD 26,450 programme-service expense description and the Schedule I grant. The accounting bridge would require an explanation that the current documents do not provide. A retrospective ledger should not invent one.

For 2023, it could show the contractor-created shortlist of 12, the three-member committee and four recipients. It could identify Michael Costello’s separately reported Board-liaison role while marking the voting and recusal method unavailable. It could record the October Board decision to end the Scholarship America relationship but should not identify Scholarship America as the shortlist contractor without explicit evidence. Intake, complete and eligible counts would remain unknown.

For 2024, it could record the October committee dissolution and the audited zero scholarship-expense line, while marking programme status, decision owner and reason for the zero as unavailable. It would explicitly decline to infer termination or causation.

This retrospective exercise is not a substitute for contemporaneous reporting. Some fields may never be recoverable, and memories of process are weaker than records made at the time. Its value is to demonstrate that accountability does not require filling every cell. A ledger can be useful when it records knowns and unknowns honestly.

It would also make changes comparable without suggesting that stability is always preferable. A smaller award count can be explained by budget constraints. A new administrator can be evaluated by whether stage definitions and reporting remain stable. A dissolved committee can be followed by a clearly named successor. The ledger would measure continuity of purpose and control, not institutional sameness.

Most importantly, it would prevent one number from doing the work of another. The 2019 cap would not become an applicant count. The 2023 shortlist would not become the intake denominator. The 2024 zero expense would not become proof that all aid ended. Each fact would stay attached to its stage and reporting unit.

What would weaken this critique

The case made here is falsifiable. It would weaken materially if NANOG publishes a stable record for the years at issue—or for future cycles—showing total applications, complete and eligible counts, the shortlist method, public criteria and weights, aggregate recusal information, award totals, the accountable decision owner and appropriately designed outcomes.

It would also weaken if NANOG identifies a specific privacy, legal or contractual constraint that prevents release of a particular aggregate. Some small cohorts may make even a count identifying when combined with public information. A clear explanation and a consistent suppression rule would be stronger than a demand for disclosure at any cost.

The critique would need to be revised if the apparent gaps are already filled in public material outside the record reviewed here. The proper response would be to incorporate that evidence, not defend an obsolete conclusion. Likewise, if the 2022 programme and financial descriptions can be reconciled by a published timing or classification note, the accounting section should give that bridge full weight.

This matters because institutional accountability should not be a presumption of wrongdoing. The record does not show discriminatory selection, capture, unlawful conduct, recipient underpayment or misuse of funds. It neither identifies Scholarship America as the unnamed supplier of the 2023 shortlist nor establishes programme termination. The argument is about the legibility of a process, not the motives or integrity of the people inside it.

The analytical principle behind the inquiry is therefore deliberately narrowed. A visible winner list cannot stand in for an applicant denominator. A benevolent purpose does not by itself demonstrate equitable effect. Accountability follows practical control when a Board, contractor, committee and staff divide work. But this is a finite student award, not a constitutional mandate or a claim to represent global users. The remedy must match the institution and the risk.

The discipline of not knowing

Institutional reporting often rewards confident narrative. Annual reports are built to tell readers what an organisation accomplished. Winner photographs, names and testimonials fit that form; incomplete denominators and role transitions do not. Yet the credibility of an accountability record depends as much on the precision of its unknowns as on the visibility of its achievements.

For NANOG’s scholarship record, the unknowns are specific. The reviewed sources do not establish the total applications in any year. They do not establish completeness or eligibility counts. They do not show how an administrator ranked or filtered candidates, whether the 2023 top 12 followed a score, or how the committee weighted criteria. They do not provide aggregate conflict or recusal information. They do not reveal applicant demographics, income, employer support or institutional resources.

They do not establish current availability, the administrator after the 2023 relationship change, or the successor decision authority after committee dissolution. They do not explain the 2024 zero expense. They do not provide a cohort measure of graduation, employment, network-operations work, meeting attendance, speaking or volunteering.

Naming those absences does not accuse NANOG of possessing and withholding every data point. Some may never have been collected. Some may exist only in confidential files. Some may be unavailable because systems or vendors changed. A decision ledger should distinguish those situations wherever possible, but an outside reader cannot assume which explanation applies.

Nor should missing data be replaced by proxies. Recipient universities do not describe the applicant pool. Recipient names do not establish protected characteristics. A historical intake cap does not estimate demand. A contractor relationship in one year does not identify an unnamed actor in another sentence. An expense line does not describe application screening. A committee’s dissolution does not reveal who inherited its authority.

This discipline protects institutions as well as applicants. It stops a chronological sequence from becoming a causal allegation. It prevents accounting differences from becoming claims of falsehood. It keeps a governance critique from drifting into personal suspicion. The result may feel less dramatic, but it is more useful: it identifies the exact evidence that would resolve each question.

The top 12 are a particularly good lesson because the figure is both informative and incomplete. NANOG did not merely announce four winners; it disclosed the set the committee reviewed. That deserves recognition. It also allows a reader to see that another decision occurred before the committee’s. Once that handoff is visible, stage ownership and stage denominators become legitimate questions.

The next disclosure need not be a dossier. It can be a row of aggregate counts, a short criteria guide, an aggregate recusal number and a named owner. The better the institution defines those fields in advance, the less it will need to defend itself against conclusions drawn from silence.

Conclusion: put the process, not the applicants, on the record

NANOG’s scholarship record contains more substance than a simple award announcement. It contains a launch, historical eligibility rules, award amounts, named recipients, an administrator’s management and monitoring role, a stated budget constraint, a contractor-to-committee handoff, Board decisions and audited expense lines. Taken together, those records show a real institutional programme and real material support.

They also stop at crucial joins. The 2023 top 12 do not reveal how many people applied or how the shortlist was created. The four names do not reveal the committee’s criteria or aggregate recusal practice. The 2022 annual report and Form 990 do not supply a visible bridge between programme outputs and accounting descriptions. The 2023 and 2024 structural changes do not identify a successor decision owner. The 2024 zero expense does not explain itself.

None of those gaps justifies exposing unsuccessful applicants. Privacy is not an obstacle to the proposed accountability; it is a design requirement. The institution should keep personal files, references, financial circumstances, individual scores and rejection reasons confidential. It should suppress small cells and make outcome participation voluntary.

What it should publish is the institutional pathway: purpose and award type, owner and administrator, calendar and intake cap, total and stage counts, criteria and weights or an unweighted-method statement, committee size and decision method, aggregate recusals, award number and value, accounting period and bridge, successor ownership after organisational change, and bounded outcomes.

Such a ledger would not tell readers whether they would have chosen the same four people. It would tell them what process existed, which actor controlled each stage and what the public evidence can support. That is the proper level of scrutiny for a scarce educational opportunity.

The four names are not the problem. The 12 are not the answer. They are the point at which NANOG’s own reporting opens a window onto the machinery between application and award. A durable, privacy-preserving record could open that window far enough to make the process legible without putting any applicant on display.

Sources