Methodology
Exitr turns dated public facts into signals and two scores. Missing data never adds points. Every claim links to its source.
Measured quality
Accuracy: each week a separate checker re-fetches a random sample of records straight from the source and compares them field by field. Target: 97% or better.
Inference precision: each week a sample of inferences is re-checked against their evidence. Target: 95% or better for every inference type; a type below the bar is hidden from customers until it passes again.
Freshness: how long ago each record was last confirmed at its source. Target: under 30 days (status and acquisition fields under 7 days).
Readiness score (0 to 100)
| Part | Max | Rule |
|---|---|---|
| Tenure | 15 | 0 under 8 years since formation or first license, rising to 15 at 20+ years. |
| Growth stall | 20 | Permits or headcount flat or down over 12 to 24 months. |
| Founder dependence | 20 | No second-in-command roles posted in 24 months; single officer. |
| Exit preparation | 15 | First finance hire, leadership page refresh, new certification, site refresh after 24+ stale months. |
| Staleness | 10 | Website not meaningfully changed in 24+ months (archive-dated). |
| Market in play | 10 | Acquisitions within 50 miles in the last 12 months. |
| Corporate change | 10 | Officer, agent or address change in the last 12 months. |
Event signals fade with a 180-day half-life; standing facts do not. Confidence is low when fewer than 4 parts have data. Weights start as expert values; once 50+ real outcomes exist in a niche, new weights ship only if they beat the current ones on held-out outcomes.
Fit score and rank
Hard filters first: your states, size, exclusions. Acquired, PE-backed and closed companies never appear. Then service overlap (40%), distance to your nearest office (25%: full within 25 miles, none beyond 150), industry overlap (20%) and tool match (15%). Parts that don't apply are left out and the rest re-weighted. Rank = 60% readiness + 40% fit (your Owner can set 40% to 80%).
Inferences
An inference combines sourced facts into a plain conclusion, like "Permits show 6 data center or utility jobs since March 2025, across 4 quarters." Each shows its evidence, a confidence label and what would change the view. A validator rejects any sentence with a number we did not see, more than two sentences, or any guess about a person.
Sources
We use a source only after checking its terms. We follow robots.txt and never log in to anything. We do not store owner names, personal phones, personal emails or home addresses.
| Source | What it feeds | Last success | Terms reviewed |
|---|
Refresh cadence
- License files: daily. Permits: daily. Deal news: every 6 hours.
- Formation records and enrichment: weekly. Scores, inferences and feeds: weekly (Sunday), after the integrity gate passes.
- Alerts on saved targets: within 1 hour of a new signal (Pro and Scale).
What is not included
No owner ages, personal social profiles, home addresses, personal emails or phones, and nothing about individuals beyond what a company publishes about its own leadership. See why.