How we count.
What a job is
One piece of work on the water that needed somewhere to stand, stage equipment, or put people beside it. One response is one job. A person can count as many jobs as they have done.
What we ask
Nine taps: the kind of job, the town, when, how much ground, for how long, the heaviest thing on it, what they stood on instead, what not having good ground cost them, and what a 20 × 10 ft steel platform would be worth to them per month. Every answer is a bucket with a fixed id. Nothing is typed except an optional phone or email on the last screen, which is stored separately and never published.
How square-foot-days are computed
Each size bucket and each duration bucket has a midpoint: about a pickup truck is 200 sq ft, a small yard 600, a big yard 1,000, a parking lot 2,000, more than that 4,000; a day is 1, a week 7, a month 30, a season 90, a year 365. A job's square-foot-days are size midpoint × duration midpoint. Jobs that never got done, and jobs whose size is not sure, are counted as jobs but excluded from the square-foot-day total.
What is published
Totals: jobs and square-foot-days. Breakdowns: by region, by job type, by what people stood on, by what the absence of ground cost, and region × job type. Never dollars: what people said a platform is worth stays private. Never a single row.
The suppression rule
Any cell with fewer than five responses shows as counting… and is left out of the bars. The totals still include it. In a town of 1,400, one row is a person; this rule is what makes no name, no company true.
What is kept, and what is not
Responses are anonymous and append-only. We keep no name, no company, no exact place, and no IP address; a salted hash of the address is held for one hour to limit repeats, then gone. Exact duplicate answers within ten minutes are dropped. In-person counts, tapped together with someone on a dock, carry a different source flag from web counts so the two can be compared.
Verifying the number
The tally is published as plain JSON and CSV at /v0/census.json and /v0/census.csv, with the time it was computed. Anyone can load the CSV into a spreadsheet and reproduce the bars from it. If you cite the number, cite the timestamp with it.