Methodology

Nine signals.
One score.

The Cook County Assessor values 1.8 million homes each cycle. Mistakes slip through — and the Board of Review reverses roughly 70% of the appeals that get filed. Censum re-scores every residential parcel against nine independent signals to flag the ones most likely to win at the Board. Calibrated against thousands of real BOR decisions, not modeled outcomes.

Production model · v3.1
Out-of-sample AUC 0.6845
Precision @ 0.80 threshold 86.1%
Properties scored 1.58M
Approach Proprietary · Calibrated
Calibration check

Calibration measures whether predicted probabilities match observed outcomes. A model that says "80% chance" should win 80% of the time.

0.8 – 1.0 bucketpred 0.869actual 0.861
0.6 – 0.8 bucketpred 0.692actual 0.677
0.4 – 0.6 bucketpred 0.491actual 0.502

The nine signals

01 Assessment spike

Detects assessment increases significantly above the township median.

02 Market divergence

Compares assessment growth to actual market sale-price trajectory in the surrounding area.

03 Permit reconciliation

Cross-references assessment increases with public building permit records.

04 Township ratio study

Measures inequity in assessment-to-sale-ratio across the township.

05 Prior BOR history

Weights the property's historical pattern of successful or denied appeals.

06 Comparable sales

Surfaces similar nearby sales assessed at materially lower ratios.

07 Equity (sister parcels)

Compares against assessor-defined sister parcels for uniformity violations.

08 Neighborhood appeal success

Encodes the success rate of recent BOR appeals within a 0.25-mile radius.

09 Dollar impact

Filters for cases where the dollar magnitude of expected savings justifies effort.

How tiers are assigned

Tier Conviction score Win probability Price
A ≥ 75 ≥ 0.70 $199
B ≥ 60 ≥ 0.60 $99
C ≥ 50 ≥ 0.50 $39

Built on hard data

Our analytics combine public assessment, sales, and appeal records with proprietary signal weighting and a calibrated machine learning layer trained on years of historical Board of Review outcomes.

Every input is sourced from the public record. Nothing is purchased from data brokers. We do not collect or store any personal information beyond what already exists in public assessment filings.

The specific blend of signals, weightings, training methodology, and refresh cadence is proprietary to Censum.

Run your own home through it.

Look up my PIN