Accuracy & verification — trust markers.
Every fact in a Mama brief carries a verify pill. The pill tells you when we last confirmed the fact, from what source, and what to do if it's wrong. This page is the full reference for the trust-marker system.
TL;DR
5 pill states: Verified <24h (green) · Verified 1-7d (green dim) · Verified 7-30d (amber) · Inferred (grey) · Stale >30d (grey dashed). Click any pill to report incorrect. Reports route to verification queue and corrections propagate within 6h. Accuracy SLAs per category are public — see below.
01The 5 pill states
| Pill | Color | Means |
|---|---|---|
| Verified · <24h | Solid green | Confirmed by a primary source in the last day. Highest confidence. |
| Verified · 1-7d | Green (dimmer) | Confirmed within a week. Fresh enough for outreach. |
| Verified · 7-30d | Amber | Confirmed within a month. Use, but re-verify if stakes are high. |
| Inferred | Grey | Constructed from patterns, not observed directly. Common for emails when company hasn't published the pattern. |
| Stale >30d | Grey dashed | Last verified more than 30 days ago. Re-verify before outreach. |
02Source attribution per pill
Hover any pill for source details. Examples:
Verified · 2h · Crunchbase— funding amount cross-checked from CrunchbaseVerified · 4d · LinkedIn— role title pulled from public profileInferred · email pattern: firstname.lastname@— constructed by domain ruleStale · 47d · Mama crawl— last seen on company /team page 47 days ago
03Report-incorrect workflow
Click any pill → "Report incorrect" → optional one-line note → submit. Reports route to our verification queue.
| What happens | Latency |
|---|---|
| Report received, queue updated | Instant |
| Our verifier reviews + classifies (false-positive, true-error, ambiguous) | < 6 hours business |
| If true-error: fact corrected, all instances across workspaces update | < 6 hours business |
| If ambiguous: investigation, may take 24-48 hours | 1-2 business days |
| You get email confirmation with outcome | At resolution |
04Accuracy SLAs per data category
Our internal accuracy targets, sampled monthly via blind audit. These are commitments, not aspirations.
| Category | Accuracy target | Q1 2026 actual |
|---|---|---|
| Funding signal (amount, date, round type) | > 99% | 99.4% |
| Hiring signal (job posted) | > 98% | 98.7% |
| Exec move signal | > 95% | 96.1% |
| Tech change detection | > 90% | 91.2% |
| Firmographic (industry, employees, geo) | > 95% | 95.8% |
| Email verification (SMTP-checked) | > 97% deliverable | 97.6% |
| Inferred emails (pattern construction) | ~85% deliverable | 87.3% |
The numbers are public in /sources, updated quarterly.
05What Mama suppresses
If a fact's confidence is below threshold, Mama omits it rather than guessing. Examples we explicitly suppress:
- Decision-maker emails when neither verification nor pattern is available — we don't make up emails
- Signals with single-source provenance below the freshness threshold
- Tech detections with only one weak signal (e.g., a single careers mention with no JS confirmation)
Omission is honest. Inferred is honest. Made up is the failure mode we never accept.