How is sensitive data protected?

All documents are encrypted with AES-256-GCM at rest and TLS 1.3 in transit. A GDPR anonymization layer removes PII before any AI processing. Your data never leaves our SOC 2 certified infrastructure.

How long does integration take?

Less than 30 minutes. Upload documents directly, connect via OAuth for Slack, or use our REST API for bulk processing.

How does outcome tracking work?

Outcomes are detected automatically from follow-up documents, Slack messages, and web intelligence. You confirm with one click. Each reported outcome makes your future analyses more accurate.

How is Decision Intel different from ChatGPT?

ChatGPT gives one opinion from one model. Decision Intel uses 3 independent judges for noise measurement, a 20×20 bias interaction matrix for compound scoring, 31 domain-specific biases, and an outcome flywheel that gets smarter with every decision.

All case studies

FailureTechnology2021

Zillow

Zillow iBuying Program Shutdown

Estimated impact: $881M write-down; 2,000 layoffs; program shuttered

F30

Decision Quality Index

Critical — 5 biases detected, 2 toxic combinations

Zillow's algorithmic home-buying program (Zillow Offers) purchased 27,000 homes using ML price predictions that systematically overpaid, resulting in a $881M write-down and 2,000 layoffs.

Decision context

Zillow's Zestimate algorithm was repurposed for buying decisions despite known accuracy limitations. When the model consistently overpaid, leadership increased purchase volume to hit growth targets rather than recalibrating the model. Internal data scientists raised concerns that were deprioritized.

Deep hindsight-stripped analysis coming soon. This case is part of our ongoing backfill to reconstruct what the platform would have flagged at decision time. The summary and bias profile below are drawn from primary sources.

Biases present in the decision

★ Overconfidence BiasAnchoring BiasConfirmation BiasPlanning FallacySunk Cost Fallacy

3D Decision Knowledge Graph

Zillow — 5 biases, 2 toxic combinations

Drag to rotate · Scroll to zoom · Click to explore

Decision

Critical

High

Medium

Toxic

Bias Intensity Profile

Severity × confidence across 5 detected biases

Toxic combinations

Optimism Trap
Sunk Ship

Reference class base rates

Across all 146 curated case studies in our library:

90%

base failure rate

10%

base success rate

Lessons learned

ML model accuracy for estimation ≠ accuracy for buying decisions (asymmetric loss)
Growth targets must not override model recalibration signals
Real estate markets have latency that algorithmic models underestimate

Source: Zillow Q3 2021 Earnings Call; SEC Filing 10-Q 2021; Bloomberg investigation (SEC Filing)

Pilot slots available