About

Who publishes this site

BabyNames is written, built and maintained by one person, publishing as Milo Hart. That is a pen name, and saying so is the point: a name you can hold the work to is worth having, and pretending a pen name is something else is not.

What that name is accountable for is concrete. The same person writes every article on this site, maintains the data pipeline that produces the name pages, decides the publication threshold, and answers correction reports. There is no editorial team, no contributor network and no syndicated content — if something here is wrong, it is one person’s mistake to fix.

This page claims no academic post, professional credential, employer or research history, because the project has none to claim and inventing them is exactly the kind of authority signal that is worth nothing. The claim this site does make is a method, and the method is published in full on the methodology page: which records are excluded, how spellings are grouped, how trends are derived and where the approach can be wrong. Anyone who disagrees can check it against the same public SSA files.

Editorial and privacy contact: milo.hart@babynames.epage.me. Corrections and the fastest way to get a grouping error fixed are on the contact page.

Why this site exists

A list of names is easy to publish. The useful questions are harder: how common is a name across an entire lifetime rather than in one annual ranking, did it rise steadily or arrive in one short wave, how much popularity is hidden across alternative spellings, and does a supposedly unisex name actually have substantial use on both sides of the source data?

BabyNames is built to answer those questions from reproducible records. It keeps the searchable database, but the database is the input rather than the finished product: the site adds pronunciation grouping, longitudinal curves, derived trend measures, comparisons, relationship tools and editorial explanations of what those numbers do and do not mean.

Where the data comes from

The US Social Security Administration publishes annual counts of names in its records going back to 1880. A name/sex combination is included for a year only when at least five records exist, which protects privacy but also makes the source incomplete for rare names. National data and the SSA territory files are processed together here.

The site is not affiliated with the SSA. The source is public; the grouping, analysis, publishing threshold, tools and explanations are this project’s own layer on top of it.

What is different about the analysis

The biggest transformation is spelling normalization. Official counts treat Jonathan, Johnathan and other spellings independently. Where spellings can be grouped safely by pronunciation, this site combines them so popularity is not fragmented simply because parents chose different letters for the same sound.

Group membership is intentionally conservative. Curated overrides and the CMU Pronouncing Dictionary are used first, with dictionary-based subword matching for unresolved names. A neural grapheme-to-phoneme model can improve the pronunciation shown on a page, but its prediction is not allowed to merge names automatically because unfamiliar and non-English names are exactly where an English-focused model can make damaging guesses.

The methodology page documents the full pipeline, including exclusions, ranking, peak-year rules, trend derivation and known limitations.

What readers can do here

• Inspect a complete historical curve for a published name rather than a single current-year rank.
• Compare names on the same axes to see whether similar totals hide very different histories.
• Filter by measurable properties such as syllables, length, peak era, first letter and recorded unisex share.
• See grouped spellings and nickname relationships instead of treating each string as an unrelated entry.
• Read editorial analysis explaining common traps in rankings and how the underlying data should be interpreted.

What this site is not

• Not a source of invented meanings. Historical popularity cannot prove a name’s etymology, cultural origin or personality traits, so those claims are not generated from the dataset.
• Not a popularity forecast. Recent movement is measured as history, not presented as certainty about what parents will do next year.
• Not global. The records describe the United States and included territories. A name may have a completely different history elsewhere.
• Not a complete census of rare names. The SSA privacy threshold removes small annual counts before this project ever receives them.

Corrections

Pronunciation grouping is useful precisely because spelling alone is misleading, but it is also the part most likely to need human correction. If two names have been merged that should be separate, or the reverse, the contact page explains how to report it. Corrections are added to explicit override data and flow through the next regeneration, so the same rule applies consistently everywhere.