How Zyphr decides what counts as a good recommendation.

1 · Universe construction

Every trading day at 19:30 IST we ingest the AMFI NAV file, reconcile it against our canonical scheme master, and build a clean fund universe. Direct plans only — regular plans are visible but not recommendable. Schemes younger than 36 months are surfaced with a freshness flag, never as a buy signal.

“A clean universe is the precondition for an honest recommendation. Most of the work is here, and most of it is invisible.”

2 · Fund scoring

Each fund receives a daily score on a 0–100 scale, composed of a weighted blend of risk-adjusted return (Sharpe, Sortino, Calmar), peer-rank consistency over rolling 1y / 3y / 5y windows, and a downside-capture penalty. Scores never reset — the previous fourteen days are always visible alongside the current value.

3 · Suitability gate

Every recommendation is checked against the client's recorded risk profile, time horizon, and current concentration. A high-scoring small-cap fund will not surface as a buy signal for a conservative client with a 3-year horizon, regardless of its score. The gate is conservative by design.

4 · Compliance rules

Eighteen coded SEBI rules run before any recommendation is ever shown — concentration limits, churning velocity, ELSS lock-in awareness, KYC currency, and fourteen others. Each rule writes a pass / fail / blocked record into the audit ledger, with the rule version and the input snapshot.

5 · Audit posture

Every advisor action — recommendation viewed, approval granted, export downloaded, override applied — is appended to a hash-chained ledger. The ledger is read-only by design; corrections are written as forward-dated reversals, never edits. This is the trail an inspection lands on.