Public Methodology
How FlowLedger calculates and validates insights
FlowLedger tracks historical institutional capital movement from public mutual fund disclosures. It does not provide predictions or investment advice.
FlowLedger is the evidence ledger for institutional capital movement — not a screener, not a chatbot.
1. Data source and ingestion
FlowLedger ingests public mutual fund holdings disclosures under the SEBI reporting regime, normalizes each fund’s positions into comparable instrument records via ISIN, and timestamps every batch. Reporting date (the disclosed holdings snapshot) and detected date (when FlowLedger processed it) are kept as separate fields, never merged.
2. Detection logic
Deterministic rule sets inspect period-over-period changes in holdings and capital exposure — accumulation, distribution, sector rotation, multi-fund consensus. Rule thresholds are period-bounded and inspectable inside the product. The rules explain what happened in the disclosed data; they do not infer why, or what should happen next.
3. Backtesting
Once a pattern is detected, FlowLedger checks how the same type of pattern behaved historically across three standard windows — 1M, 3M, 6M — and records a win rate (the share of prior occurrences that moved in the flagged direction) and a median return for each window.
4. Ledger entry
A signal that clears the backtest thresholds is recorded as an immutable, timestamped ledger entry with a unique ID. It is never edited in place — if a correction is needed, a new entry is added and the old one stays on the record.
About the “Backtested Win Rate” metric
The headline percentage is a directional win rate, not a prediction-accuracy score: it's the share of backtested signals that moved in the flagged direction within the window, against a 50% baseline (an unbiased coin flip). It is derived from historical backtest outcomes recorded with insights, evaluated across standard windows (1M, 3M, 6M) and aggregated into a single figure for public display.
This number is descriptive of historical behavior. It is not a promise of future returns.
On sample size: FlowLedger is in alpha and the recorded history of pattern occurrences is still growing. A win rate built on a small sample of prior occurrences can move meaningfully with the next disclosure cycle — treat it as directional context for a signal, not a statistically stable estimate, until the underlying sample grows. Each signal in the product shows the sample it was computed from so you can judge that for yourself, rather than take the headline number on faith.
Worked example
One signal, traced from disclosure to ledger entry
Disclosure. Seven mutual funds file their monthly SEBI holdings disclosure. Across the three months covered, each fund's position in HDFC Bank (INE040A01034) increased.
Rule trigger. The Multi-month Accumulation rule detects that all seven funds moved in the same direction within the 3M window, for a combined net inflow of +₹243 Cr.
Backtest. This pattern type is checked against its own history: 68% win rate over 1M, 74% over 3M, 61% over 6M. The 3M result clears the pass threshold, so the signal proceeds.
Ledger entry. An immutable entry is recorded — ld_a7f3c291e84…, timestamped, fund names and capital shift attached — visible to any authenticated user with the full fund-by-fund breakdown behind it.
This is the same illustrative example shown on the homepage — the format is real, this specific walkthrough is for explanation rather than a live signal.
What FlowLedger does not do
- No predictions or price targets.
- No personalized investment advice or suitability logic.
- No trade execution or brokerage workflow.
Glossary preview
Five terms worth knowing before you look at a signal
Pulled word-for-word from the full glossary inside the product — 21 terms in total, covering metrics, core concepts, and technical fields.
Win Rate
The percentage of historical occurrences where the pattern's flagged direction played out — price rose after an accumulation signal, or fell after a distribution signal. Above 50% means the pattern beat a coin-flip baseline historically; it is not a prediction of the next occurrence.
Net Value Delta
The total change in capital, measured in ₹ lakhs, flowing into or out of a stock across mutual funds. Positive means accumulation. Negative means distribution.
Ledger ID
The unique identifier for each recorded insight. It lets users trace evidence and raw disclosures for that exact entry.
Multi-Fund Consensus
When two or more funds show directional agreement, all buying or all selling, on the same stock within the same period.
Backtest Window (e.g. 3M)
The historical time period used to evaluate how similar patterns performed in the past. 3M window = last 3 months.
See all 21 terms →
Full glossary, searchable by category (login required)
What this page is
This is the public methodology — enough to evaluate FlowLedger, cite it, or decide whether to request access, without logging in. The full glossary, per-signal readiness flags, and evidence drill-down for every live signal are available inside the authenticated product.