Coverage and freshness

History accrues forward, it does not reach back

A change exists in this dataset only if two snapshots were taken around it. Detection began when we started snapshotting each source, so no plan returns events from before that date. The lookback window on a plan is a ceiling on what that plan may request, never a promise that the history exists yet. Setting a wider date range on the change feed does not return older events.

The newest change in the dataset was observed Aug 24, 2026. Per-state dates are in the table above, and they differ by months.

Some states are a snapshot, not a feed

Virginia, Texas, Alaska, Utah, GA, MS, NY, AL, TN, NC, OH, MI, IL, PA, SC, NJ, LA, IN, MO, CA, MN, KY, MA, MD, WI, CO, KS, WA, OK, IA, NV, NH, NE, WV, MT, SD, DE, ID, AR, RI, NM, HI, ME, ND have changes in the dataset but no scheduled refresh. Those states came from bulk files that are downloaded and applied by hand, some of them several gigabytes, so their records stand still between runs and an alert on them cannot fire in the meantime. Florida, Oregon, Connecticut, Arizona are on a schedule.

Coverage is uneven by design, and by source

Association records come from public sources, and those sources differ enormously by state and county. A state that publishes a corporate registry, a parcel roll and an officer roll produces a deep record. One that publishes only a corporate registry produces a name and an address. The 56 jurisdictions are not comparable to each other and this page does not average them into a single number.

Document search is metadata on the free tier, full text on a paid one

Two bodies of association documents are searchable: files harvested from associations’ own websites, and filings recorded with the county. Without a subscription, a search matches document titles, filenames and the county’s own category labels, across recorded governing documents: declarations, amendments, bylaws, articles, other governing documents and legal descriptions. That answers whether we hold a document and what it is, which is what most people arrive asking.

A paid plan searches the text inside the documents, returns the matched passage, opens every category both bodies hold, including budgets, financials, insurance certificates, capital projects and structural reports, and links to the file itself. The county corpus has no extracted text for anyone, so it is searched by filing name at every plan; which of its categories are visible follows the same line.

Every plan has a monthly allowance of documents and of requests, counted per account rather than per search, so the same allowance applies however a search is sliced. The allowance and how much of it is left are shown on the document search itself.

No personal contact data, and no free text we have not read

A personal email address or phone number for an individual never appears here, on any surface, in any export or through the API. That is enforced in the database rather than in this application, so it holds regardless of what any page or API route asks for.

Board members and officers do appear by name, title and term, from the state officer rolls and annual report filings those names are already public in. What does not appear is any way to contact them personally. A management firm, an insurance carrier, an engineering firm or a vendor is a business, and its name and professional details are shown.

Some of what we hold is free text read out of documents an association published: minutes, budgets, inspection reports. Those values are shown only for the fields we have reviewed and classified as carrying no individual's details. Everything else is marked as withheld rather than printed, and a new kind of extracted value stays withheld until somebody classifies it.