Benchmark Methodology

How Benchmarks Are Calculated

Every contract reviewed through LegalStack generates a set of anonymized metadata signals — clause types detected, risk levels assigned, term durations, and structural patterns. These signals are aggregated across all reviewed contracts to produce market-standard benchmarks.

Benchmarks are computed as simple distributions: for each contract type + clause type combination, we calculate the percentage of contracts that include the clause, average numeric values (e.g. term lengths, cap amounts), and risk level distributions.

Privacy Protections

• Zero contract text stored: We never save any actual contract language in our benchmarks database.
• No party names or identifiers: Contract reviews are anonymized before any signals are extracted.
• Aggregate-only reporting: All benchmark outputs are population statistics — no individual contract can be reverse-engineered.
• Minimum sample threshold: We require at least 10 data points before reporting any benchmark for a clause/contract combination.

Confidence Levels

Freshness & Weighting

All contracts contribute equally to benchmark calculations regardless of when they were reviewed. As the dataset grows, older patterns are naturally diluted by new data. We plan to introduce time-weighted rolling windows (prioritizing the most recent 6 months) once the dataset exceeds 1,000 reviews per contract type.