The weekly brief, the searchable case database, the monthly intelligence reports, the early access, the full archive. The Pro tier is the foundation; the Diligence tier adds the model layer on top.
Submit a case (district, framework, agent firm, owner type, neighborhood, year) and receive a calibrated probability of a recommended-on-consent outcome at first hearing — the single most actionable signal in the dataset. The probabilities are calibrated against held-out test cases: when the model says 70% it means roughly 70% of similar cases historically went on consent.
The same call returns expected postponement count and time-to-disposition as range estimates (with honest confidence intervals; the underlying models are usable but not pinpoint).
A propensity-adjusted score for every named agent firm in the dataset. The index compares each firm's empirical consent-rate against the model's prediction for that firm's actual case mix. Firms scoring above 0.50 outperform what their case mix predicts; below 0.50 underperform. The index is the closest thing in the market to an apples-to-apples comparison of land use representation effectiveness.
An empirical scoring of each neighborhood planning area on opposition intensity — derived from contact-team letter patterns, public-comment volume per hearing, NPA amendment frequency, and the postponement-chain distribution of cases in that NPA. Used to gut-check a project's likely community-engagement burden before filing.
The same prediction, similarity, and index endpoints used by the web portal are available as a JSON API with a personal API key. Useful for integrating ADW signals into a CRE underwriting model, a deal-flow CRM, or a custom dashboard.
# Predict outcome for a case $ curl -H "Authorization: Bearer $ADW_TOKEN" \ -H "Content-Type: application/json" \ -d '{"district":"D3","framework":"DB-90","agent":"Drenner Group","owner_type":"Shell LLC"}' \ https://adw-api-proxy.mmsebenzi-oracle.workers.dev/predict # → response { "p_consent_first_hearing": 0.42, "p_consent_ci_low": 0.31, "p_consent_ci_high": 0.55, "expected_postponements": 2.6, "days_to_disposition_p50": 128, "days_to_disposition_p90": 340, "model_version": "v5.0c", "trained_on_cases": 2552 }
One 30-minute Zoom per quarter with Michael — Diligence subscribers walk through what the model is seeing in the current cycle, what's changed in the District Friction Index, what to expect in the next quarter. Four briefings per year.
Drenner Report. State of Austin Land Use quarterly. DB-90 at Two Years. The upcoming Armbrust & Brown, Husch Blackwell, and Thrower Design firm reports. Diligence subscribers get every paid PDF without separate purchase. Across the 2026 product slate, that alone exceeds the subscription delta vs Pro.
The Case Outcome Predictor is a gradient-boosting model (CatBoost primary, LightGBM control) trained on the full 10-year case set. Features include district, framework (DB-90, ETOD, PUD, NPA, CO Modification, VMU, etc.), agent firm (canonicalized; top firms one-hot encoded), owner type derived from owner-string parsing, opposition signal density, repeat-player indicators, framework-age in days at filing, and the case's commission body.
Evaluation uses chronological 80/10/10 split — the model never sees future cases during training, and metrics are reported on the most recent ten percent of the dataset. This is the legal-prediction best practice for avoiding temporal leakage. Bootstrap confidence intervals are reported on every headline metric. Target 1 (recommended-on-consent at first hearing) is well-calibrated; targets 3 and 4 (postponement count, time-to-disposition) are usable as ranges, not as point estimates, and are reported as such.
The model retrains monthly as new cases land. Diligence subscribers see the version number on every API response and the calibration-curve report in the dashboard.
CRE underwriters pricing entitlement timeline risk on Austin acquisitions. Institutional capital partners deploying capital into Austin development funds. Repeat-player land use attorneys benchmarking their case mix against the model's expectations. Developer principals running portfolios with five-plus active entitlement files at any time. Anyone whose decisions on Austin entitlement risk benefit from a calibrated probabilistic forecast rather than gut estimation.