Data Availability Commitment
County GIS endpoints go stale, move, or quietly stop answering queries. Most parcel-data vendors serve a cached copy and hope you don't notice. We check all 155 county endpoints across 50 states every two hours and publish the result.
01, What we commit to
- Probe every two hours. A full automated check runs in four shards, testing not just whether each layer loads, but whether it still answers real queries and exposes the fields we publish. Endpoints already marked degraded are re-checked every 30 minutes.
- Flag immediately. A degraded or down endpoint is marked in every API response (
endpoint_status) and on the public status page while it's being fixed. - Repair or retire within 7 business days. We find the replacement layer and restore it. If we can't, the county is marked down and its atlas page says so. No silent dead links.
02. How the monitoring works
The full check runs every two hours in four shards, with 30-minute re-checks for endpoints already marked degraded. A daily Vercel run remains as a backstop. For each county endpoint the probe reads the layer metadata and runs a live query: an actual request to the county's server, not a reachability ping. Government ArcGIS servers stall for a second or two often enough that the check retries a slow response before counting it as a failure. What reaches the status page is a real problem.
An endpoint that fails three consecutive checks doesn't sit in a backlog waiting for someone to notice. If a field we publish has disappeared upstream, the system drafts the correction on its own. If the endpoint moved or went dead, it files a repair task with the diagnosis attached: last healthy date, failure type, the dead URL. A person confirms each change before it goes live, inside the seven-business-day window. You can watch it on the status page.
03, No data lock-in
The atlas points at public county ArcGIS REST endpoints. Those are government servers anyone can reach. We don't wrap that data in a format you can only read through us, and we don't hold it back to keep you subscribed. If you leave, the endpoint directory is still yours: which county sits at which URL, and what its fields are called, are public facts, and the open atlas package ships them under an MIT license.
What you pay for is the work of keeping all 155 entries currently true: the automated verification, the self-healing, and the repair commitment. A copy taken today starts drifting the day a county moves a layer; the maintained atlas does not. You're paying for the upkeep, and the data underneath stays public.
04. What it is not
This is not an application-uptime SLA. It covers the third-party county endpoints the atlas indexes, not our own servers. We don't run the county's ArcGIS server and can't promise it stays up, only that a scheduled check will catch the failure and we'll be honest about it. For website and API uptime (99.5%, with service credits on Team and Enterprise), see the Enterprise SLA.
05, Check it yourself
The status page is public and live, no account, no sales call. See which endpoints are healthy right now.
View current endpoint health →06, By plan
| Plan | Data Availability | Application uptime SLA |
|---|---|---|
| Free · Individual · Pro | Two-hour full sweep, 30-minute degraded re-check, public status, repair commitment | — |
| Team · Enterprise | Same + named contact for repair escalation | 99.5% + service credits |
Enterprise service credits: 10% for missing 99.5%, 25% for 95–98.9%, 50% below 95%, capped at 50% of the monthly fee, claimed within 30 days of the incident.
07, Questions
What exactly do you commit to?
We schedule a full four-shard check of every county REST endpoint every two hours, and re-check endpoints already marked degraded every 30 minutes. If one stops answering queries, we flag it in the API response and on the public status page after a completed check, then investigate or replace it within 7 business days. If we can't restore it in that window, we mark the county down and update its atlas page so you're never quietly handed a dead endpoint.
How does a repair actually happen?
The automated probe counts consecutive failed checks. After three, the system steps in: for a field that changed upstream it drafts the correction, and for a moved or dead endpoint it files a repair task with the diagnosis attached. A person reviews and ships each one. Nothing rewrites what your queries return without that review.
Is this an uptime SLA?
No. This is about the county data endpoints we index, not our own servers. Application uptime, the website and the Parcel Data API, is a separate 99.5% SLA available to Team and Enterprise plans, with service credits. The Data Availability Commitment covers the third-party county layers the atlas points at.
What if the county takes their own server down?
We don't control the county's ArcGIS server, and we don't pretend to. The commitment is that a scheduled check will surface the failure in the API response and here, instead of serving a cached or broken layer as if it were live. The repair clock starts when a completed check detects the problem.
Which endpoints does it cover?
Endpoints in the atlas that returned a healthy result on the previous check. A county we've never been able to verify is listed honestly as such; the commitment applies to keeping working endpoints working, and to flagging the ones that break.
What happens to my data if I stop paying?
Nothing of yours is locked up. The county endpoint directory is public fact, and the open atlas package is MIT-licensed, so the URLs and field mappings stay usable. What you lose by leaving is the maintenance: the recurring checks and repairs that keep the list current. Without them, the copy you keep is accurate the day you export it and a little less so every week after.