The Standard · v1.0.0
What the grade means
The 30 checkpoints, in 8 disciplines, that every automation is graded against. Pass all 30 and hold them green for 60 days to certify. Open any discipline to see what is inside.
- 01Row Count ReconciliationNo rows quietly go missing. A 240-unit rent roll comes out as 240 units, not 228.
- 02Schema Drift DetectionIf a source renames or moves a column, the run stops instead of putting rent in the expense field.
- 03Unit and Currency NormalizationMonthly versus annual and dollars versus thousands never get mixed up, so NOI is not off by 12x.
- 04Period and Date AlignmentA trailing-12 and a current rent roll are not blended across different dates into a wrong NOI.
- 05Null and Sentinel HandlingBlank or 'N/A' values are treated as missing, not silently turned into $0.
- 06Deduplication and IdentityThe same property or comp is not counted twice because the address was written two ways.
- 07Input ProvenanceEvery number can be traced back to the document and cell it came from when someone asks.
- 08Source-of-Truth Conflict ResolutionWhen the rent roll, T-12, and OM disagree, it flags the conflict instead of silently picking one.
- 09Document-Ingestion RobustnessIt handles messy real files (scans, phone photos, merged cells) or refuses them. It never guesses a number.
- 10Idempotency and Re-Run SafetyRunning the same job twice does not create duplicates or double-send anything.
- 11Scraper Layout-Change DetectionIf LoopNet or a county site changes its page, you get alerted instead of pulling garbage comps.
- 12Graceful Source DegradationOne source going down does not crash the whole run or store a blank as if it were real data.
- 13Partial-Failure IsolationOne bad file in a batch is set aside so the rest still finish.
- 14Timeout and Hang ProtectionA slow source cannot hang the whole job forever.
- 15Liveness AlertingIf the job stops running, you are told. Silence is not treated as everything-is-fine.
- 16Volume Anomaly DetectionIf a run processes 4 deals instead of the usual 80, it flags it instead of passing green.
- 17Data Freshness and Comp DecayStale data and six-month-old comps are flagged before you underwrite on them.
- 18Actionable Alert RoutingImportant alerts reach a person by text or Slack, not a log nobody reads.
- 19Numeric Source-TracingThe model cannot invent a cap rate or NOI that no document supports.
- 20Arithmetic in Code, Not ProseThe math runs in code, so the same deal always produces the same number.
- 21Regression Eval SetBefore any change ships, it is tested against known-good deals so accuracy does not quietly drop.
- 22Abstention on Unreadable InputOn a blurry scan it says 'I cannot read this' instead of confidently guessing.
- 23Output Range and Sanity BoundsImpossible numbers (140% occupancy, $4/SF office rent) get flagged, not passed downstream.
- 24Human Approval GatesNothing irreversible (an LOI, an offer, a tenant message) goes out without a person approving it.
- 25Model and Dependency PinningThe AI model is pinned, so an overnight provider change does not silently shift your outputs.
- 26Secret ManagementAPI keys and passwords live in a secure place, never hardcoded in the code or repo.
- 27Least-Privilege AccessThe tool only has the access it needs, read-only where it just reads, not admin keys to everything.
- 28PII and Untrusted-Input HandlingTenant SSNs and bank details are masked, and a malicious document cannot hijack the tool.
- 29Runbook, Ownership, and ConfigSettings live in one place a non-engineer can change, with a doc so it is not stuck in one person's head.
- 30Load Behavior and Cost CeilingsIt is tested at real volume and has spend caps, so a runaway loop cannot burn your budget overnight.