A Correction: We Measured Our Connectors and Called It a Metro

Boston is the row that gives the game away. It was published as the diversified counter-example, and the reason it looks diversified is that we do not run a direct connector for any employer with a Boston hub. The spread between Columbus and Boston is not a spread between concentrated and diversified labour markets. It is the difference between a metro where we read an employer's own portal and one where we do not.

Source: Seeker job corpus, 257,534 active job postings, snapshot 2026-08-16. seekerscore.com/insights/one-employer-metro-job-markets-2026

Correction, 2026-08-16

What this page claimed: 'Columbus is 396 postings of which 363 (91.7%) are JPMorgan Chase, across 11 distinct employers,' with Philadelphia at 71.3%, Seattle 62.5% Amazon and San Diego 47.6% Apple, presented as evidence that a metro's posting count hides how few employers stand behind it. What is true: of the 465 active postings we hold with a Columbus metro tag, 421 (90.5%) come from a single source, and that source is `jpmc.fa.oraclecloud.com`, JPMorgan's own careers portal. 418 of those 421 are JPMorgan Chase. The concentration figure and the connector are the same fact stated twice. Every largest-employer in the published table, JPMorgan, Amazon, Apple and Citi, is an employer we ingest directly rather than one that emerged from a survey of the metro.

Why this took three tries to see

This page was itself a correction. An earlier version ranked metros from a 30,000-row sample and got the ordering badly wrong, and this page was written to fix it by moving to the full corpus. Moving to the full corpus fixed the sampling error and left the deeper one untouched, because the full corpus has the same shape as the sample in the way that mattered: it contains whichever employers we built connectors for. A bigger denominator does not fix a biased numerator. The number that should have stopped us was printed on the page in its own column: 11 distinct employers in a metro of 465 postings is a statement about coverage, not about concentration, and we published it as a finding rather than reading it as a warning.

What stands

Two things. First, the arithmetic: the counts, the per-metro shares and the 55,867 metro-tagged denominator were all computed correctly over the corpus. Nothing here is a calculation error. Second, and more usefully, the general warning the page was trying to make is still true and is now better supported: a metro's posting count on any job board is not a census of that metro, and the employer mix behind it is worth checking before you read anything into the volume. We are simply the demonstration rather than the observer. If a job board tells you a city has a certain number of openings, that number is a function of which employers that board reads.

What is withdrawn

Every per-metro concentration figure, the ranking of metros by concentration, the Columbus and Philadelphia headline claims, and the Boston-as-diversified counter-example. We are not replacing them with corrected concentration figures, because we cannot compute an honest one: measuring how many employers are behind a metro requires knowing that we see a representative set of that metro's employers, and we demonstrably do not. Fixing this needs different data, not better arithmetic on the same data.

How we measured it (and what we did not claim)

The correction is a source decomposition of the same corpus rows the original page used: all active postings carrying a Columbus metro tag, grouped by source and by company, on 2026-08-16. That is one query, and it is the query that should have run before publication. We are NOT claiming Columbus is not a JPMorgan town, which it may well be, nor that Seattle is not Amazon-heavy, which it plainly is. We are claiming that our data cannot establish either, because the employers we see most are the employers we chose to read. We are also NOT claiming the metro tagging is wrong; 22.6% of the corpus carries a metro and that scoping was stated correctly throughout. This is the third correction today with the same root cause, which is a claim built by comparing our own copy of the world against itself. The others are on the salary page and the posting-length page.

Frequently asked questions

Related Data

Guides

• Hidden Job Markets →
• Apply on the Company's Own Site →