11.1Frequency as one observation
A count is interpretable only once the unit is named. In this small record there are six submitted messages, three labelled contributors, and two named concerns: A sends four messages about closing time and B sends one, while C sends one message describing a step at the north entrance. The three tallies should therefore be kept apart. Message volume answers how often an expression entered the archive; the contributor tally answers how many distinct labels entered it; the claim tally answers how many distinct matters were raised. Collapsing them would let repetition in one register stand in for evidence in another.
Applied to the record, five messages concern closing time: two contributors raise it, and it names one concern. The step names a second concern and arrives once, from one label. Nothing follows about its frequency in the wider population of service users, which is unknown, nor does the description establish that every entrance is inaccessible. But the archive's stated review task includes identifying entrance barriers. That requirement, not the fact that the concern is rare, is what makes C's message eligible for review.
Eligibility is not priority. Limited review time still demands an explicit priority rule, and the available counts do not supply one. Weighting by message volume would let one sender absorb attention; weighting by contributors would treat many labels as inherently more important than few; weighting by distinct claims would protect narrow concerns while ignoring how widely an issue is felt. Each is a choice with a foreseeable failure case, and the archive should say which one it adopts.
The strongest worry is that protecting every unusual statement invites manipulation and makes aggregate attention unmanageable. Keeping the counts separate does not dissolve that worry, but it shows when apparent support is only volume, so a weighting rule can be chosen knowingly rather than smuggled in by the tally.
Separating messages, labelled contributors and named concerns yields a clean first count for this record: six messages, three labelled contributors, two named concerns. The count is worth keeping, but it stops short of the inference one might want to draw from repetition. Repeated wording does not reveal whether messages are independent reports, a coordinated effort, or several submissions from one person. Here A's four submissions about closing time are stipulated to come from A; the repetition is a fact about the intake, not a detection of matching text, and it establishes neither coordination nor manipulation. The labelled contributors also do not exhaust the population of service users, which remains unknown.
That limitation on what repetition reveals cuts both ways when frequency is converted into attention. Weighting by message volume lets A's four submissions dominate the agenda. Weighting by distinct contributors largely erases A's additional submissions. Weighting by named concerns counts the closing-time issue once however often it is phrased. Each weighting is a choice with a characteristic failure case: sustained legitimate concern can be discounted, a single contributor can swamp the queue, and claims can be split or merged to alter their apparent weight. Frequency therefore records what arrived, while the weighting decides what the record is allowed to do.
C's single message about a step at the north entrance is eligible for review because the stated review task includes identifying entrance barriers, not because the message is rare. Eligibility is an entry condition, not a scheduling verdict. Limited review time still requires an explicit priority rule, and that rule should say what rarity may contribute and what it may not. The qualification that matters here is scope: the described step concerns the north entrance only, and it does not show that all entrances are inaccessible. Protecting an unusual concern need not mean promoting every one of them; it means refusing to let repetition alone set the terms of review.
11.2A rare but consequential distinction
Suppose the archive contains six invented messages with stable contributor labels: A sends four about closing time, B sends one about closing time, and C sends one about a step at the north entrance. A plain count records six messages, three labelled contributors, and two named concerns. Those figures describe the record; they do not settle what the review should address. Closing time dominates the tally, but the review task expressly includes identifying entrance barriers, and C's single message names one at the north entrance.
The inferential step is a stipulation for this case: take C's report as identifying a barrier that is essential to satisfying the declared access requirement. That stipulation does not treat rarity as evidence. A step at one entrance is not proof that all entrances are inaccessible, and one message does not establish how widely the barrier is experienced. The point is narrower. Once the record identifies a barrier bearing directly on a requirement the review has committed to meeting, its claim on attention cannot be discharged by counting expressions. It is eligible for review as a concern of a different kind.
This creates a practical difficulty rather than abolishing one. Review time remains limited, so identifying a rare concern as requirement-relevant still demands an explicit priority rule. If every unusual statement receives top priority regardless of its connection to a declared requirement, exceptions can be manufactured and aggregate attention becomes unmanageable. The distinction therefore separates two tasks that a single tally collapses: counting expressions and evaluating stakes. The first reports how often concerns appear in the record; the second asks what follows from the requirement and the described barrier. Neither statistic substitutes for the other, and the second cannot be read off the first. What supplies the warrant for priority is the requirement and the evidence about the barrier — a matter the next move must make precise.
Applied to this archive, the access requirement supplies the reason to examine the step at the north entrance. C’s message is the only one naming it, while five concern closing time. The concern’s importance therefore follows from the stated task of identifying entrance barriers together with evidence of a particular obstacle—not from the fact that it appears once. Rarity alone establishes nothing: an isolated message may concern a trivial inconvenience, while a recurrent complaint may reflect a serious defect. What matters is the connection between the described barrier and a requirement the review is meant to serve.
The distinction between counting expressions and evaluating stakes produces a defensible priority rule under limited review time. A concern tied to a declared requirement cannot be displaced solely because other expressions are more numerous. Frequency remains relevant to questions of scope—how widely a concern is represented—but it does not settle stakes. Counting expressions and evaluating consequences are separate operations, and collapsing them gives A’s four stipulated messages four times the influence of B’s one without any evidence that closing time is four times more important.
The available alternatives each fail somewhere. Ranking by frequency buries the entrance barrier. Ranking by severity would invent comparisons, since severity has not been observed here and the record's counts do not measure frequency in the population. Giving every message equal weight confuses repeated expressions with independent evidence. Privileging rarity would merely reverse the error and make novelty a currency.
The governing objection is substantial: protecting every unusual statement invites manipulation and leaves aggregate attention unmanageable. The rule should therefore claim less. It authorizes bounded review of a requirement-linked barrier; it does not give every rare message automatic protection. A’s repetition is stipulated from stable labels and does not demonstrate coordinated manipulation. Likewise, C’s report establishes evidence of a step at the north entrance, not that every entrance is inaccessible. Nor can the six messages estimate prevalence, since the service-user population is unknown. The resulting judgment is local and revisable: investigate the identified barrier because the task requires it, while keeping the record’s evidential limits explicit.
11.3Repeated messages and coordinated input
The fictional archive holds six messages from three labelled contributors, naming two concerns. A sends four about closing time, B sends one about closing time, and C sends one about a step at the north entrance. A's repetition is stipulated here; it is not inferred from matching wording, so nothing in the record supports a conclusion about how the four arose. That leaves the task of separating two things the message count conflates.
Stable contributor labels distinguish repetition within one contributor's submissions from different contributors raising the same concern. They do not establish whether any repetition is deliberate, legitimate or abusive. A raises closing time four times; B raises it separately. Those forms can be treated differently without changing the underlying concern, and neither pattern establishes coordination. Labels record who sent each message, not why, and the archive offers no evidence of shared intent. Coordination would require something beyond resemblance or repetition within one label, and this record does not supply it. The limit is sharp: at most the record shows who repeated what.
C's message names a distinct concern, a step at the north entrance. Because the review task includes identifying entrance barriers, that concern is eligible for review; its importance follows from the stated requirement and from evidence about the barrier, not from its appearing once. The step does not establish that all entrances are inaccessible, so the concern merits examination without licensing a general claim.
Limited review time then forces an explicit priority rule. Ranking by raw count privileges A's four messages and allows repetition to amplify a concern. Ranking by rarity alone privileges whatever is least common, including a trivial item. The defensible rule records frequency and stakes separately, then states how they combine; the choice among weightings is a policy decision with a failure case, not a fact read off the six messages.
The archive's six-message record also limits what can be inferred from form alone. A's four complaints about closing time are stipulated repetitions; their similarity was not first detected in the text and then interpreted. B's message concerns the same topic under a different contributor label, and C's concerns a step at the north entrance. Resemblance among these messages is not evidence that any contributor or wider group coordinated to flood the archive. To call repetition manipulation at the point of intake is to add an accusation the record does not carry.
C's concern reaches review by a different route. The stated task includes identifying entrance barriers, so a reported step at the north entrance is eligible because it falls inside that requirement, not because rarity entitles any single message to special protection. Eligibility still does not settle how much attention the concern receives, and the report supports no claim about entrances generally: a named step is a named step. The archive can register it without converting one account into a finding about every entrance. Equally, the stipulated repetition from A establishes nothing about the service-user population, so it cannot be used to argue that this record is typical or inflated.
Priority is therefore the remaining problem, and limited review time forces an explicit rule. Frequency can enter such a rule, but as a chosen weighting rather than a measure of importance: what gets counted determines whether A's repetitions, B's separate contribution, or C's distinct concern presses the stronger claim. Each weighting has a failure case, and none is compelled by the record itself. Selecting among them is the next question; the limit here is that no resemblance between messages, and no rarity among them, decides the matter in advance.
11.4Weights chosen for a purpose
The small record produces three different counts: six messages, three labelled contributors, and two named concerns. A sent four messages about closing time, B sent one about closing time, and C sent one about a step at the north entrance. None of these numbers substitutes for the others.
Each count answers a different question. The message count measures how much input arrived and may matter when estimating handling work. The contributor count measures how many labelled voices appeared in the record. The concern count separates problems rather than communications: five messages concern closing time, while one concerns the entrance step. A's repetition is stipulated rather than inferred from matching language, and it does not establish coordinated manipulation.
The appropriate weighting follows from the review question. If the task is to identify entrance barriers, C's message is eligible because it describes one, not because it is rare or because rarity proves importance. Closing-time messages remain eligible to the extent that the task covers them, but their larger tally does not by itself settle priority. A weighting that gives the entrance concern special protection is therefore a choice about how scarce review attention should be used, not a conclusion extracted from the counts.
That choice has a boundary. The population of service users is unknown, so these six messages cannot show what share of users encountered either problem. If a rate is wanted, the eligible population and its denominator must be specified separately; the message, contributor, and case counts cannot be treated as interchangeable denominators. Nor do the facts establish that closing time is more severe than the entrance step, or the reverse. Limited review time still requires an explicit priority rule, but that rule should declare what it is protecting and what evidence it needs, rather than disguising its purpose as a simple tally.
Having separated messages, contributors and cases, the remaining choice is which of them the weighting is meant to describe. In this invented archive, counting each labelled contributor once yields three contributors: A and B both raised closing time, and C raised the step at the north entrance. That figure answers a question about the range of labelled voices represented in the file. It does not answer how many service users share a concern, because the population of service users is unknown and the stable labels do not establish who else uses the archive.
A proportion can therefore be stated only once its scope is fixed. Five of the six messages concern closing time; two of the three contributors raise it; the two named concerns are closing time and the entrance step. Each comparison is bounded by the items counted, and none of them is a rate of concern among service users, for which no denominator is available here. Treating the six messages as that denominator would quietly assume a correspondence between messages and users that this record does not provide. The same caution applies to A's four messages: their repetition is stipulated, not detected from matching text, so its concentration in this file is not evidence of coordinated manipulation.
Eligibility must also be settled before priority is assigned. The stated review task includes identifying entrance barriers, which is what makes C's report eligible despite appearing once; rarity alone confers no claim. The described step supports only the obstacle reported at that entrance, not a finding that every entrance is inaccessible. Limited review time consequently leaves an explicit priority rule to be chosen, rather than inherited from whichever count is most convenient. Frequency is stipulated within this constructed record, while severity remains unspecified. The priority rule records a choice about attention; it does not estimate prevalence or severity among service users.
11.5Minimum coverage and its costs
A minimum-coverage rule gives every distinct eligible concern one place in the record, regardless of how many messages mention it. In the six-message archive, that means two retained items: the closing-time concern and the step at the north entrance. Five messages mention closing time, but coverage is not proportional attention. The rule answers whether each eligible concern reaches review, not how often it was raised.
The eligibility of the entrance concern follows from the stated review task, which includes identifying entrance barriers. Rarity does not make the concern eligible; the task does. Had the step at the north entrance fallen outside the review requirement, minimum coverage would not compel its inclusion.
The cost appears in what the retained records no longer show. One item for closing time does not distinguish four messages from A and one from B. Nor does the archive reveal prevalence: the population of service users is unknown, so neither 5/6 of messages nor 2/3 of contributors licenses a claim about how widespread either concern is. A's repetition is stipulated, not evidence of coordinated manipulation, and should neither earn extra standing nor trigger exclusion.
With limited review time, even two records require a priority rule, but that is a decision about effort, not about coverage. If a later rule gave the entrance step explicit priority, it would express the task's importance weighting; it would not show that the concern appeared often or rarely.
Minimum coverage therefore buys a bounded guarantee—no distinct eligible concern is lost to repetition—while surrendering frequency as an internal ranking signal. The qualification that matters is precise: coverage concerns eligibility, not prevalence. A record that preserves distinct concerns can still fail to support proportional claims, and that limitation should remain visible when interpreting the archive.
After minimum coverage retains one record of each distinct eligible concern, the question becomes which concerns are eligible at all. Eligibility should track the stated review task. The archive's task includes identifying entrance barriers, so C's concern about a step at the north entrance enters review despite appearing once. Rarity supplies no reason to include or exclude it. Nor does the described step support a wider claim: it establishes one barrier at one entrance, not inaccessibility at every entrance.
Scope is also a cost control. If every unusual statement created an open-ended review obligation, the queue could grow faster than review time allows. The six messages here carry three labelled contributors and two named concerns: four from A about closing time, one from B about the same concern, and C's single entrance concern. The record does not establish coordinated manipulation. Manipulation therefore cannot supply the eligibility test; the boundary must rest on scope. Merging duplicates keeps both named concerns while stopping message volume from setting the review burden.
The boundary needs two provisions. Exclusions should state a reason—outside the task, or duplicating a retained concern—so the decision concerns classification rather than the merits of a suppressed claim. Contributors need a route to challenge an exclusion by showing that the concern falls within the task, as C's does through the requirement to identify entrance barriers. That challenge can correct misclassification, but it cannot create review capacity. Limited time still calls for an explicit priority rule among eligible concerns. Eligibility itself says nothing about how frequent or severe any concern is, and the record's message counts are not measures of either.
11.6Uncertain populations
The six messages are a known collection, but they are not the eligible population. A sent four about closing time, B one about the same subject, and C one about a step at the north entrance. Those are six messages, three labelled contributors, and two named concerns in this record. The number of service users who could have written, or who face the step without writing, is unknown and stays unknown.
The gap between the six recorded messages and the unknown eligible population matters for the counting question. A's four messages express one concern four times; C's message expresses one concern once. Counting messages therefore allocates attention to A by four to one. Counting contributors gives closing time more standing than the entrance step, two labelled contributors to one. Counting distinct named concerns gives the two concerns equal standing, though it would discard the fact that closing time was raised more often. These are alternative weightings, not competing discoveries: each rule answers a slightly different question, and each fails when treated as a measure of stakes. Message counts reward repetition; contributor counts treat a lone contributor and a hundred quiet users alike; concern counts cannot distinguish a widespread complaint from a narrow one without information this archive does not contain.
Repetition by A is stipulated in the example and was not detected from matching text, so it does not establish coordinated manipulation. C's report of a step establishes one entrance barrier, not that every entrance is inaccessible. The review task names entrance barriers as an object of review; that requirement, not the rarity of C's message, makes the concern eligible. Rarity alone would be a poor boundary, since it would let almost any outlier claim attention. With review time limited, some priority rule must still rank the eligible items, and the choice among these weightings should be stated openly rather than hidden inside the count.
This chapter shows what an unknown denominator blocks in the six-message record. Within this stipulated archive, A sends four messages about closing time, B sends one on that concern, and C sends one about a step at the north entrance: six entries, three labelled contributors, two named concerns. Yet the population of service users is unknown. Nothing in the entries licenses the claim that these contributors represent everyone behind the service, so no frequency claim about users can be recovered from these counts. The repetition by A is stipulated, not inferred from matching text, and so supplies no basis for a charge of coordinated manipulation. Nor does C's single report show that all entrances are inaccessible.
The application point separates eligibility from priority. Because the stated review task includes identifying entrance barriers, C's concern is eligible on that basis, not because rarity itself confers automatic protection. C's entry is rare within the record, but "rare" here describes the entries, not the users. Limited review time still requires an explicit priority rule; without observed frequency data, that rule is a choice rather than a measurement read off the archive.
The archive bounds what has been documented; it does not bound who stands behind the service. That gap is not repaired by repeated entries, stable contributor labels, or the internal coherence of the record itself. The strongest objection—that protecting every unusual statement invites manipulation and makes aggregate attention unmanageable—therefore cannot be answered by inflating this record into a population estimate. If further users behind the archive were identified and their concerns observed, frequency estimates could become available; that is a conditional possibility, not a result here. Allocation must be defended as an allocation choice, not disguised as a finding about prevalence.
11.7A transparent allocation dispute
The record supports precise counts and little more. Six messages arrive from three labelled contributors: four from A and one from B concern closing time, while one from C concerns a step at the north entrance. There are therefore two named concerns in this small file. The population of service users remains unknown, so the five-to-one split among messages cannot be read as a prevalence estimate. A's repetition is stipulated rather than inferred from matching text, and it is not evidence of coordination or manipulation; equally, the reported step does not establish that every entrance is inaccessible.
Scarcity makes the allocation explicit. The review task includes identifying entrance barriers, and that requirement—not the rarity of C's message—makes the north-entrance concern eligible for attention. A workable rule therefore gives the accessibility concern a defined share of review time first, because the task assigns it a stake, and allocates the remaining time across the closing-time submissions in proportion to their number. This rule is a choice, not a deduction from the counts.
Each weighting fails somewhere. Pure frequency would crowd out the accessibility concern indefinitely whenever the archive receives many routine complaints, even though the stated review task requires such barriers to be identified. Pure stakes weighting, applied without a cap, invites the opposite failure: if any concern can claim unlimited priority by being framed as severe, aggregate attention becomes unmanageable and repeated or strategic claims gain leverage. The chosen rule accepts both costs in limited form—minority concerns are not displaced by volume alone, and no concern consumes the whole review budget.
The qualification to retain is narrow: the north-entrance message earns review time because the task names entrance barriers, not because one occurrence is proof of importance. Once the rule is stated in this way, disagreement need not appeal to hidden facts.
Once the competing priorities are stated, the remaining disagreement cannot be settled by counting again. In the stipulated record there are six messages from three labelled contributors: five concern closing time (four from A, one from B), and one concerns a step at the north entrance (C). All of those totals are correct, and nothing turns on disputing them.
What the totals license is narrower. They count messages, not the prevalence of underlying concerns in the population of service users, which is unknown. A's four messages are stipulated to be repetitions, not detected from matching text, and their number does not establish coordinated manipulation. Likewise, the described step establishes a barrier at that entrance; it does not establish that all entrances are inaccessible, and it supplies no measure of how severe the barrier is. So neither frequency in the affected population nor severity is observed here.
The allocation choices therefore remain choices. A frequency-weighted rule gives closing time more review time because more messages express it, but its failure case is clear: the single entrance concern can receive almost no attention even though the stated review task includes identifying entrance barriers. An importance-weighted rule gives C's concern review time because the task makes it eligible, and not because rarity itself is valuable. Its failure case is equally clear: if every unusual statement is protected, attention becomes unmanageable and repeated messaging can crowd out others. The qualification that keeps this manageable is that the weighting must be explicit and bounded by the task, with each concern treated as one entry rather than scaled by its senders' persistence.
Reasonable reviewers can still divide scarce time differently here, giving priority either to the concerns voiced most often or to the barrier the task requires them to find. The counts do not decide between them; the priority rule does.