Methodology & data sources

What this page is

BirthdayLab is a data-and-tools site. This page is the originality defence behind it. It documents the editorial standard, the data pipelines, the per-source provenance, the per-page computations, and the build-time sanity gate that cross-checks each page's headline figures against verified dossier anchors before the site is allowed to ship.

The summary is plain: every figure on this site traces to a primary source recorded in the page's research dossier; if a fact can't be verified, it doesn't appear on the page. The sections below make that traceability inspectable rather than asking you to take it on faith.

The editorial standard

The one rule that overrides every other rule is verified-or-omitted. We never present an estimate as a fact, never fabricate a quote or credential, never write first-person experience that didn't happen, and never publish a figure we can't trace to a primary source recorded in the page's dossier. Where a useful number genuinely doesn't exist in the source record, we omit it.

Each page also runs against a content-quality bar inherited from our internal house style. Paragraphs are 3–4 sentences maximum, every figure is introduced in prose, no chart or table is published without explanatory text around it, and every page carries at least two genuinely depictive inline SVGs of its own data.

We hold ourselves to a "what we don't do" list as much as to a "what we do" list:

• No per-date pages. The tool answers any of 366 calendar dates; there are no 366 thin per-date pages built from a template.
• No per-year pages. Same principle for birth years.
• No AI-generated prose. Every page is written by a human editor working from a research dossier. Generative models are not in the writing loop.
• No scraping of competitor sites. Data comes from the primary sources listed in §4, not from downstream aggregators.
• No celebrity / famous-birthdays content. Deliberately off-strategy; not our angle.
• No third-party calculators or chart libraries. Every tool and every SVG on the site is hand-built for it.

The dossier-first workflow is structural: a research dossier is written for the page before any code is written. The dossier carries the verified anchor figures, the primary-source citations, and the page's claims. The page is built against the dossier, and the build refuses to ship if any computed value drifts beyond tolerance from a dossier anchor (see §3 and §6).

The build pipeline

Three concrete data pipelines run on the site. Each one starts at a vendored primary-source CSV inside the repository, passes through a build-time aggregation script, and produces a JSON file that the page renders from. The diagram below shows the flow end-to-end.

The rarity pipeline is the worked example. The two FiveThirtyEight CSVs (CDC/NCHS 1994–2003 + SSA 2004–2014) are vendored in data/raw/. The build script scripts/build-rarity-data.ts parses both CSVs, splices the overlapping years cleanly, aggregates by (month, day) and by weekday and by month, ranks the 365 non-leap dates by per-year average, and writes the result to data/rarity-by-date.json. That JSON is read by lib/rarity.ts and consumed by the homepage's rarity tool, the day-of-week page, the birth- month page, and the frequency multiplier on the birthday-twins page.

The same build script runs the sanity gate. A list of dossier-verified anchor figures (September 9 average per year, Christmas average per year, the Tuesday weekday peak, the 1994–2014 total, the February 29 share, the August total share, and several others) is compared against the computed values. The default tolerance is ±1% relative; anchors quoted as rounded percentages use an absolute percentage-point tolerance instead. If any anchor breaks tolerance, the script throws and the build halts — the site cannot be deployed in that state.

The same script also writes data/sanity-gate.json — the live anchor table this page renders in §6. That file is regenerated on every build, so what you see in §6 is the actual gate output from the build that produced this page, not a hand-curated snapshot.

Per-source dossiers

Every primary source we draw on is documented below in the same format: who publishes it, what it actually captures, what it does not capture (honest limits matter more than headline figures), licensing, coverage, when we last verified it, the pages on the site that consume it, and the on-page attribution string we use verbatim.

Where a canonical source URL could be verified at build time, it appears inline. Where it couldn't (the sandbox cannot probe USNO, NIST, McCrindle, or the Slovenian Information Commissioner), we cite the source by name only — a name without a link is more defensible than a guessed URL.

Per-page computation

For each Phase-1 page, the table below states the headline figure the page leads with, the literal arithmetic behind that figure, and the sanity-gate anchor (if any) that verifies it. Pages with no data dependency — pure calendar arithmetic or pure probability — are listed with the computation but no anchor.

Live sanity-gate output

The table below is generated on every build by scripts/build-rarity-data.ts. Each row is a dossier-verified anchor figure, the value computed at build, the tolerance applied, the resulting drift, and a pass/fail status. If any row fails, the build halts.

As of the last build of this page, 38 of 38 anchors passed. The worst observed drift was 0.38pp.

Generated 2026-06-25. Source: data/sanity-gate.json (rebuilt on every site build).

What we can verify, what we can't

An honest methodology page names its limits explicitly. The ones that matter on this site:

• Day-of-month U.S. natality stops at 2014. NCHS holds day-of-month in restricted-use files; the public series exposes year, month, and weekday only. The full reasoning is in §8.
• Leapling counts are estimates. The "~5 million leaplings worldwide" figure on /leap-year and /birthday-twins is a commonly cited estimate, not a measured census. It's labelled as an estimate on every page that uses it.
• Generation ranges are research conventions. The Pew ranges through Gen Z are widely adopted but explicitly framed by Pew (May 2023) as comparison tools rather than fixed categories. Gen Alpha and Gen Beta are McCrindle Research conventions, not Pew, not government. The /generation-calculator surfaces the attribution visibly on every result.
• Some canonical URLs aren't linked. Where the build couldn't verify a canonical URL at build time — USNO's leap-year page, NIST's tropical-year statement, McCrindle Research, the Slovenian Information Commissioner — we cite the authority by name and omit the link. The standing rule is that a name without a link is more defensible than a guessed URL.
• U.S.-centric. Our daily resolution comes from U.S. sources (CDC/NCHS + SSA). Global figures (UN WPP, world population) are coarser — annual, country-level. The pages name which scope each figure covers.
• Period vs cohort life expectancy. The life-expectancy module on /age-calculator uses the SSA period life table — i.e. it assumes 2023 mortality holds going forward. It's a population average on a period basis, framed explicitly as not a personal prediction.

Why day-resolution stops at 2014

The same moat sentence appears on the homepage capsule, and this section expands it. Exact date-of-birth counts aren't available in CDC's public natality data — they're held in restricted-use files for confidentiality. CDC WONDER's public natality interface exposes year, month, and day-of-week, but not day-of-month. The matrix below makes the per-variable status visible at a glance.

This is a policy stance, not a temporary outage. Day-of- month natality lives in NCHS's restricted-use files, which require institutional review board approval to access and cannot be redistributed. The 1994–2014 daily series we vendor (via FiveThirtyEight, CC BY 4.0) captured day-of- month before the current public-use boundary stabilised. That's why the daily series ends in 2014: there is no public successor to extend it from.

The relative rankings the rarity calculator produces — September peaks, holiday dips, the Tuesday weekday spike, the weekend gap — are structurally stable across decades. They reflect scheduled-delivery practice and seasonal conception, both of which have changed slowly. If NCHS ever republishes day-resolution natality in a public file, we'll extend the series; until then, the 1994–2014 window is the best public answer to the question the page asks.

Refresh cadence

Each source we read from has its own publication cadence, and each page that depends on it gets reviewed and bumped when the source refreshes.

Derived data we publish

BirthdayLab publishes a small set of derived datasets for free re-use. Each one inherits its license from the upstream source; please cite as listed.

Site changelog

An honest, dated record of major site events. One entry per user-visible shipping commit.

• 2026-06-14

  Site initial publication: /, /age-calculator, /birthday-paradox, /birth-statistics, /methodology.

• 2026-06-15

  Added /birth-rate-trends (1909–2025 series + downloadable CSV).

• 2026-06-16

  Added /day-of-week-born and /birth-month (cluster spokes); shared WeekdayBarsSVG and the per-day↔total toggle.

• 2026-06-17

  Added /leap-year (Template 1 tool with live Feb 29 countdown) and /generation-calculator (attributed Pew + McCrindle ranges, 2010–2012 overlap shown).

• 2026-06-18

  Added /birthday-countdown (live birthday timer) and /birthday-twins (the Phase-1 synthesis hub); Phase 1 complete.

• 2026-06-19

  Added /privacy and /about (trust pages); registered business name (Moving Data Systems, d.o.o.) on the footer and in Organization schema.

• 2026-06-20

  Pinned SITE_URL to production; preview deploys serve robots Disallow:/.

• 2026-06-21

  Upgraded the homepage rarity reveal (hero count-up, choreographed line→marker, interpretive verdict, scale-first capsule with restricted-use-files moat); profile mesh wired to real Phase-1 siblings.

• 2026-06-15

  Methodology page rewritten with full per-source dossiers, per-page computation table, and live sanity-gate output.

References

Canonical citations referenced across the site, gathered. Where a stable URL exists and could be verified at build time, it's linked. Where it couldn't, the citation is by name only.

• Martin JA, Hamilton BE, Osterman MJK. Recent declines in nonmarital cesarean delivery and related changes in U.S. births by day of week. NCHS Data Brief No. 200, May 2015.
• National Center for Health Statistics. Births: 2024 Final Data. NCHS Data Brief No. 535, July 2025.
• National Center for Health Statistics. Vital Statistics Rapid Release No. 43, April 2026 (provisional 2025 U.S. natality).
• U.S. Social Security Administration. Period Life Table, 2023, in the 2026 OASDI Trustees Report.
• United Nations Department of Economic and Social Affairs, Population Division. World Population Prospects, 2024 Revision.
• Pew Research Center, May 2023 statement on stepping back from rigid generational framing (cited by name; no inline URL verified at build).
• U.S. Naval Observatory — statement of the Gregorian leap-year rule (cited by name).
• U.S. National Institute of Standards and Technology — Earth's tropical year length (cited by name).