Hiring Climate — Washington DC, Technology

Roles newly posted each week across the tracked cohort — and why the slope is easier to misread than it looks.

Hatched weeks are reconstructed, not observed. They are rebuilt from job boards, which only list roles still open. Roles posted in that week and since filled have vanished from the board, so they are missing from the count.

That loss compounds with age: the further back a week is, the more it under-counts. Which means some of the rise toward the right of this chart is that artifact rather than real growth. A flat true series would still slope upward here.

Only the solid weeks were observed live, week by week, while the service was running. Those are counts, not reconstructions. The series becomes trustworthy as solid weeks accumulate and the hatched ones scroll off the left.

How many companies in this metro posted to the monthly Hacker News hiring thread in the last three months.

It is the only figure here that is a company saying so itself. Everything else on this page is observed — roles counted on a job board, filings lodged with the Department of Labor. This is stated intent, published by the company, on a date, in public.

It reaches a different population to the rest of the page. Hacker News posters skew to startups too small or too new to run an applicant-tracking system we poll, so this count moves independently of the job-board table — and a metro can score zero here while hiring heavily.

We publish the number and not the list. This is the one source permitted to create a company record from free text, which is what lets it see those startups at all, and equally what makes the names it produces the least reliable in the system.

How many positions this company's own public job board listed in this region on the day the page was built.

It is a count of what is on the board, not a count of hiring. A role can sit open for months, be filled the day after we counted it, or be a pipeline posting that never closes. The date at the top of the page is the day this was true; nothing about it updates until the next build.

Only companies whose job board we can read appear at all. We poll Greenhouse, Lever and Ashby. An employer on Workday, Taleo or an in-house portal is missing from this table entirely — in San Francisco that is most of the largest employers in the metro. So this is not a ranking of who hires most here, and a company absent from it has not been assessed and found wanting.

Roles are assigned to a region from the location text on each posting. “San Francisco, CA” is unambiguous; “Bay Area”, “Remote — US” and “Hybrid” are not, and some roles land in the wrong metro or in none.

Labor Condition Applications this employer had certified by the Department of Labor for technology roles in this metro, over the last 12 months.

A certified LCA is the step that precedes a visa petition. It records an employer's intent to sponsor a role at a stated wage and worksite — not a person hired. One filing can cover more than one position, and a filing may never become a petition at all.

Read it as evidence of a sponsorship practice, at roughly the scale the number suggests. It is the most checkable figure on this page: the filings are public, name the employer, and are linked from every row that carries a count.

A dash is not a zero. Counts are published only where they are above zero, and only alongside a positive tier. A dash covers three situations we do not separate in public — no filings found, filings that have lapsed, and employers no filing source has been checked for — so it is never evidence that a company will not sponsor.

Denials are not published at any threshold. Certification is close to a formality and denials are overwhelmingly clerical — a mis-keyed field, a missing signature — so a denial count beside a company's name invites the one inference the data cannot support.

Departments this company posted a role into in the last eight weeks, having posted none into them before.

A function opening for the first time is a stronger sign of expansion than backfilling one that already exists — a company hiring its first data-platform or its first APAC sales role is starting something, not replacing someone.

The departments are the company's own labels on its own postings. A board that gets reorganised, or a team that gets renamed, looks from the outside exactly like a new team. Treat this as directional.

It reads high for companies we have only recently started tracking. The measure compares the last eight weeks against everything before them, so where there is no “before”, every department is new. Companies added to the cohort recently will show more here than they have earned.

Only positive counts are published. “Opened no new teams” is a negative claim about a named employer, and a zero reads as a measurement rather than as an absence.

Whether this employer sponsors work visas, from US Department of Labor filings.

Every H-1B, H-1B1 and E-3 hire begins with a Labor Condition Application (LCA) filed with the Department of Labor. Those filings are public and name the employer, the worksite city, the role's occupation code and the wage. We count certified filings for technology occupations at worksites in this region.

These are tiers, not a score. Given the filings, sponsorship history is close to a matter of record, so a 0–100 number would invent precision that the data does not support.