TOP 100

Top 100 — research, ranking criteria and publication plan

Prepared 22 November 2026. Audience: global, across all subjects. Target: https://top100.caqa.com

The publication

Top 100 helps readers explore notable people, places, ideas, organisations and achievements through clearly defined lists. The commissioned catalogue contains 500 enduring categories and 300 monthly categories. These are category briefs, not a claim that 80,000 individual positions have been researched. Every page distinguishes a published ranking from a research brief. A list is published only when its evidence and review are complete.

The website begins with a searchable category library, subject navigation, all-time/monthly filters, saved categories, linked source requirements, public methodology, editorial policy and two measured starter lists. It is an English-language global publication; country or regional editions must be scoped explicitly rather than described as worldwide.

Research foundation

The accompanying categories.json is the exact category register. Each row defines the title, subject family, cadence, method, metric, geographic scope, sources, eligibility and refresh policy. sources.json records primary-source routes and practical limitations. RESEARCH.md explains the selection and domain coverage. Source suitability at the family level is not proof that a source covers every candidate in a category; entity-level verification is a separate publication gate.

The OECD/JRC handbook on composite indicators informs the requirements for explicit weights, normalisation, uncertainty and sensitivity analysis: https://www.oecd.org/en/publications/handbook-on-constructing-composite-indicators_533411815016.html World Bank API documentation supports reproducible indicator extraction: https://datahelpdesk.worldbank.org/knowledgebase/articles/889392-about-the-indicators-api-documentation These sources support the method; neither endorses Top 100.

Category admission criteria

Every category must answer a distinct reader question, identify an eligible universe, have a defensible evidence route, and avoid duplicating another category with slightly different wording. A Top 100 requires at least 100 qualifying entries. If fewer can be verified, publish a clearly labelled shorter selection or retain the research brief; never fill the balance with invented entries. Do not split categories solely to reach the commissioned count. Review overlapping regional, genre and demographic categories for genuine value.

Enduring means historically or structurally meaningful, not permanently frozen. All-time musical, political or scientific achievement must allow older and non-English-language evidence. Historical coverage is uneven, so describe geographic and archival limitations. Monthly means reviewed on a monthly cycle; it does not automatically mean the underlying source publishes monthly.

Three ranking methods

Measured lists

Prefer one intelligible metric when it answers the question: population, land area, attendance, verified sales, award count, capacity or officially published performance. Define unit, direction, reference period, population and exclusions before collecting data. Do not mix cities proper, urban agglomerations and metropolitan regions. Do not mix company revenue, profit, valuation and market capitalisation. Do not compare currencies without a declared conversion date and source.

Sort raw unrounded values. Exact ties share the same competition rank (1, 2, 2, 4), with alphabetical order used only for display. A 100-row display may truncate a boundary tie; flag it, or expand the public list to include the complete tied group and explain why there are more than 100 entries. Missing data is unknown, not zero. Avoid a supposedly global ranking if the dataset covers only one market.

Editorial selections

Use this method for greatest singers, influential thinkers, landmark films or transformative political figures. Proposed default rubric: documented achievement 30%, lasting influence 30%, contribution or originality 20%, and recognition across independent sources 20%. Publish category-specific definitions and weights before scoring; these default weights are a proposal, not an independently validated scientific measure.

Rate each dimension on anchored 0–5 scales. A zero requires evidence of the defined lowest level; absence of evidence stays missing. Example influence anchors: 1 = documented local reach; 3 = sustained field or national influence with multiple independent records; 5 = sustained cross-generational international influence with corroborating evidence. Two suitably informed reviewers score independently, disclose conflicts and reconcile differences exceeding one point. Keep the rationale and evidence for each score. If qualified reviewers have not participated, do not claim an expert panel.

Calculate weighted score = sum(weight × dimension score / 5). Record reviewer ranges and test whether reasonable weight changes materially alter the order. Where rank differences are not defensible, use tied tiers or an unranked selection. Artistic merit, fame, influence and moral approval are different concepts. Political influence is not endorsement; explain both contribution and documented harm proportionately. Avoid ranking people by sensitive traits or subjective personal worth.

Attention charts

Use readership or platform engagement only for titles that say most read, most viewed or most discussed on the named platform. Fix the language edition, market, observation window and bot policy. Exclude disambiguation pages, automated traffic where measurable, duplicates and unrelated entities. A spike is attention, not quality or approval. Do not infer country of residence, political allegiance or demographic traits from platform activity.

Public voting is a later, separate feature, with eligibility rules, abuse controls, a minimum sample, opening/closing dates and visible sample size. Never mix unverified votes into editorial or measured rankings, and never seed votes, testimonials or comments.

Monthly operation

Maintain a register with category owner, source access, data frequency, retrieval method, last successful snapshot, reporting period, review date, next due date and status. Categories using annual filings or annual official statistics retain the latest reported period and show “reviewed this month; source data unchanged” when appropriate. No artificial movement arrows.

Proposed monthly cycle in Australia/Sydney time: day 1 collect the last complete calendar month; days 2–3 resolve identities and validate units/coverage; days 4–5 independent evidence checks; day 6 freeze approved snapshots; day 7 publish and verify. Allow source-specific lags. Each edition gets an immutable version, input manifest, source URLs, retrieval timestamps, method version and reviewer record. Retain the prior edition; correct through a documented new version.

Movement is prior rank minus current rank only when scope and method are comparable. Label new entrants NEW and missing prior entries UNRANKED, never invent a previous position. A source outage leaves the previous verified edition visible with an overdue notice. Changes in methodology break the comparison series and must be labelled. Automation may collect and validate evidence, but must not silently turn draft briefs into published rankings.

The initial site does not claim 300 live monthly pipelines. Activate each source adapter only after a successful dry run, permission/licensing review and independent acceptance. API keys, paid datasets and scraping permissions must be resolved before dependent categories can publish. No subscription or credit purchase is included.

Evidence and editorial safeguards

Use official datasets, original award bodies, recognised archives, official sporting federations, museums, scientific repositories and audited company filings. Secondary scholarship can interpret historical influence; social posts and commercial listicles are leads, not sufficient proof. For each material claim retain the exact URL, reporting period, retrieval date, relevant excerpt or field, and allowable reuse basis. Two independent sources are desirable for contested qualitative claims; a single authoritative numerical dataset can support a clearly attributed measured table.

Before publication check source completeness, duplicate identities, territorial naming, unit consistency, ties, missingness, rank reproducibility, broken citations, image licences and conflicts. Do not republish third-party rankings in full merely because they are accessible. Attribution is not a licence. Use original synthesis and permitted data, and link to restricted material without copying it. Photographs need an asset register with creator, licence, source and credit requirements.

Editorial choices are separate from paid placement. Sponsorship cannot buy a rank. CAQA ownership and relevant commercial relationships must be disclosed. Offer a corrections route with a reproducible evidence requirement, log material changes and allow responses to disputed factual claims. Do not invent editor identities, credentials, testing labs or review panels. Health, finance and safety categories require suitable specialist review and clearly bounded comparative claims before release.

Information architecture

Home: publication identity, browse/search, genuine published lists, featured research categories and methodology entry. Category library: search title/family, cadence filters, subject filters, saved-only view and pagination. Category detail: title, status, scope, qualification, method, metrics, sources and update policy; published tables additionally show rank, entity, raw value, unit, data period and row-level source. Methodology: methods, ties, missingness and monthly policy. Editorial policy: ownership, conflicts, corrections and evidence. Plan: this publication and operational plan. Future edition archive: immutable monthly snapshots with meaningful rank movements.

Reader controls must work without an account. Saved categories live in that browser only and can be cleared. No subscription or nomination form should appear to submit unless a real receiving system is implemented. Search pages and navigation should work on phones and keyboards; use semantic headings, visible focus, meaningful link names, readable contrast and reduced-motion support.

Technical approach

Use a small static generator and plain HTML/CSS/JavaScript on the existing CAQA cPanel server. This suits an evidence-led publication and avoids a large application or database before the editorial workflow exists. Generate durable category URLs, pre-rendered details, a compact search index, per-page descriptions, canonical URLs, and a sitemap limited to substantive published pages. Mark research briefs noindex until they contain complete original rankings. Do not assert Review/AggregateRating structured data without genuine reviews.

Keep raw source snapshots and build scripts outside the public document root. Deploy only the generated site. Stage a release archive, validate local links and record checksums, back up any existing target, install domain-specific files, obtain valid TLS and verify public URLs. Do not modify other CAQA sites. A rollback restores the prior target release or removes only newly created Top100 content; DNS/vhost removal requires deliberate owner review.

Launch sequence and acceptance

Foundation: deliver exactly 500 enduring and 300 monthly category briefs, source register, methodology, plan and tested public website. Seed release: two complete World Bank measured Top 100 tables with fixed reporting years and explicit inclusion of territories. Broad editorial release: select 20 priority categories spanning singers, people, cities, politics, film, sport, science and companies, build candidate pools, perform independent review and publish only accepted lists. Expand in batches of 25, retaining a visible completion register. Monthly launch: validate a small representative set of adapters and two consecutive comparable editions before expanding to all 300.

For each list require 100 verified eligible records or an explicitly shorter list; full row-level source coverage; declared reporting period and method; zero unexplained duplicates; reproducible numerical sort; resolved review findings; usable phone and keyboard views; working source links and a public live check. The category catalogue count is not the completed-ranking count.

Capacity and commercial plan

A mature 800-list publication can contain about 80,000 ranked positions. Even five minutes per initial position is roughly 6,667 hours before review; shared entities and permitted APIs reduce duplication but do not remove the evidence requirement. Three hundred monthly lists at even 30 minutes per list require 150 hours per month before acquisition and substantive editorial work. These are planning illustrations, not quotes or measured productivity. Assign accountable editors and budget source access before committing a public delivery date for the full library.

Start with free permitted data and the existing hosting allowance. Later revenue options include clearly labelled advertising, sponsorship outside ranking decisions, optional reader memberships and licensed original research. Do not introduce paid ranking placement. Measure return readership, successful searches, source-link use, correction rate, overdue editions and verified-list completion; avoid celebrating page count alone.

Immediate next production work

Finish verification of the two measured starter lists and website deployment. Audit the full category register for source feasibility and overlapping topics. Choose the first editorial candidate pools, obtain category-specific reviewer acceptance, and implement the monthly data manifest/adapter interface. No future list is described as complete until that work is actually done.