1 · What starts the run
A watcher checks the department's extracts and picks up new data the moment it lands, whichever day of the week the state chooses to publish.
The department says its records refresh weekly and never says which day, so we watch for the data to arrive rather than set a timer against a schedule the state has not published. That is our problem to manage, not yours.
What it means for you is a fixed day. Whatever the state does midweek, your list is built and sent on Monday. The delivery day is ours, so you can plan your week around it.
2 · Pull and compare
The inspection extracts are cumulative. Each district file is a full rebuild covering July 1st to the run date, and there is no "records added this week" feed anywhere in the product. Without comparing each new file against everything that came before, every weekly list would re-deliver the entire fiscal year.
So every weekly snapshot is retained permanently, and the new file is compared against the archive rather than against last week's copy alone. Three things depend on that archive holding:
- The fiscal-year rollover. On July 1st the district files reset to near-empty and rebuild.
- Prior citations. Counting how many separate times an establishment has been cited requires history the current file no longer contains.
- Your delivery history. What has already been sent to you is kept separately, where the source files cannot reset it. You do not get the same row twice.
3 · Filter to your service area
Rows are narrowed to your geography and to the categories that describe your trade's work. That second part is less obvious than it sounds: the same category means different things depending on what is being inspected, so a reader that ignores the difference produces confident nonsense in both directions. Ours refuses to read a category at all until it knows what kind of property it is looking at, which is why a lodging citation never reaches you mislabeled as a food one.
4 · Resolving contacts
The inspection record names a business and an address. It does not tell you who to call. Closing that gap is the actual work of the product.
Each establishment on your list is paired with a license number and a business phone number, taken from public business sources. Where a field cannot be resolved it is marked as unresolved rather than left ambiguously blank, so you always know the difference between "no number on file" and "nobody looked".
A record is attached to a business on an exact match or not at all. There is no similarity scoring and no confidence threshold anywhere in it, because a tunable threshold is a guessed match with a number attached — and the realistic way to harm someone with this product is to attach a record to the wrong business.
5 · Getting it to you
Your list is checked, built and sent every Monday, so you start the week with the records already in front of you.
It arrives as an XLSX spreadsheet attached to an email. No portal, no login, nothing to install, and no shared file quietly going stale between weeks. Open it, sort it, work it, or paste a block straight into your CRM — the citation travels on every row, so the record goes with it.
Volume is best-efforts with no minimum, because it depends entirely on what the state publishes in a given week. A record that cannot be reported cleanly is left off rather than padded in, and a week with nothing in it gets a short written note naming what was searched and the dates covered — never an empty attachment, and never silence.