Home Platform Capabilities Methodology Calibration For Pricing FAQ About
Scroll
Calibration

Anchored to the historical record.

Every Svatantra output is grounded against confirmed historical outcomes. Real awards. Named recipients. Dated. Sourced.

53K+
Confirmed outcomes
72
Agencies
156
Sub-agencies
16
Markets
SOURCE USASpending.gov · FHWA · 2010 to 2024
247
comparables
$487M
median award
50% range$280M to $820M
80% range$145M to $1.94B
95% range$50M to $5.6B
Heavy tail · recent Brent Spence $1.635B · Cape Cod $1.06B
Historical comparable distribution

Every estimate has a shape.

For each signal we surface the distribution of comparable historical awards. The shape shows where the typical award lands, where the spread is wide, and where the heavy tail lives.

No point estimate. A distribution. Uncertainty made visible.

Resolution

Calibrated where the evidence is dense.

Calibration runs at multiple resolutions of the historical record. The system tries the finest cell first, then steps back when evidence at that resolution is thin.

No black box. No silent extrapolation.

Evidence framework

Seven mandatory layers. Every output.

Source attribution. Statistical transparency. Methodology disclosure. Prediction tracking. Confidence framework. Dissenting evidence. Audit trail.

Elements of the output. Not pages of fine print.

Suppression

When we don't know enough, we say so.

Estimates can be withheld. Ranges can be widened. Monetary claims can be suppressed entirely.

The reviewer always knows which case applies.

Institutional standard

Reviewed. Gated. Backtested.

Reviewed by an external statistical advisor. Gated by an acceptance protocol. Backtested against the calibration record on every recompute.

Take it further

Three ways to verify it yourself.