Methodology
How Docu-Coach™ verifies restaurant statistics
The Restaurant Statistics 2026 page is a structured collection, not a blog roundup. A number is published only when we can point at the organization that produced it.
Primary-source preference
Government datasets come first: BLS CES, JOLTS, OEWS, and QCEW, plus EEOC enforcement releases when the case is restaurant-specific. Industry association PDFs, news roundups, and other statistics hubs can help us find a study. They are not the canonical source.
How statistics are collected
Each statistic is stored as data — identifier, statement, value, period, population, geography, source URL, and verification date — not as a sentence buried in page copy. The HTML hub, Markdown mirror, JSON feed, and CSV download all render from that catalog.
Restaurants vs hospitality
Food services and drinking places (NAICS 722) is the restaurant and bar series. Accommodation and food services (NAICS 72) adds lodging. Leisure and hospitality adds arts, entertainment, and recreation. When a source measures the wider bucket, the page says so in the first sentence.
Why there is no 75–150% annual turnover rate
That range is common in restaurant trade writing. It is not a BLS series. JOLTS publishes monthly quit and total-separation rates. We quote those rates and refuse to annualize them into a headline percentage the survey does not publish.
Update frequency
CES employment, hours, and earnings update monthly. JOLTS updates monthly. OEWS wages update annually. QCEW establishment counts update quarterly. A protected Vercel cron runs on the 15th of each month, hashes every watched source, and re-parses the official series. The HTML hub, JSON feed, and CSV download read the latest verified overlay. If the database is unavailable, the published TypeScript catalog is the fallback.
What “last verified” means
It means a successful check against the primary series — either a person opening the source or the monthly watcher confirming the parsed official value. It does not mean the underlying survey was run that day. The watcher never auto-publishes an AI-extracted guess. If a parser fails, a source hash changes without a safe parse, or a value moves more than 25%, the hub keeps the last verified figure and emails the review queue.
Geographic limits
Unless a statistic says otherwise, coverage is the United States. State wage bases and local pay look different; do not treat a national median as a schedule rate.
Estimates and revisions
BLS flags recent CES and JOLTS months as preliminary. QCEW 2025 establishment counts are preliminary. When a later release revises a number, the catalog and the live overlay keep the previous value in revision history rather than silently overwrite it. Settlement pages such as EEOC newsroom releases are hash-watched only; they are not treated as live numeric series.