Brain learns, measures, and shows you the math
Every recommendation is measured against real SEO data at +7 / +14 / +30 days. The numbers below are live aggregates across every customer (anonymized).
Outcome-measured recommendations
When you mark a Brain rec done, we snapshot your current metric (audit score, top-10 keywords, citation rate, etc.). Then 7, 14, and 30 days later we measure the same metric. The delta is the proof.
A/B tested prompts + features
Forge ships every customer-facing prompt as a versioned shadow A/B test. We only promote when the shadow beats production on quality + cost. The /changelog page shows what survived.
No hidden quotas, no dark patterns
Credit costs are documented and capped. Auto-publishing requires explicit approval. Sensitive actions emit an audit-log row visible in your account.
Privacy-preserving aggregation
Cross-customer signals (e.g. 'verified prospects from your niche') only surface with an n≥5 cohort gate, no user IDs stored. Public stats are anonymized rollups — no per-customer data ever rendered here.
How we measure
1. Baseline at mark-done
The moment you click Done, we record the current value of the affected metric (audit score, citation rate, keywords in top 10, etc.) into the outcome ledger.
2. Re-measure at +7/+14/+30d
A scheduled job re-reads the same metric on each anniversary day. The delta is stored — that's the lift, however positive or negative.
3. Aggregate honestly
The average lift, win rate, and per-category breakdown above are computed from real outcome rows. We surface losses too — that's the only way the number stays meaningful.
What this number is NOT
- — Not survivorship bias. Every measured outcome counts — including negatives.
- — Not "users who finished onboarding". Anyone who marked a rec done is included.
- — Not cherry-picked metrics. Brain picks the metric at mark-done based on the rule_key — we don't switch metrics later.
- — Not a promise of YOUR result. SEO is noisy. We surface the cohort average; your project may vary.