Inputs
What data is needed?
The product does not need tick data for the MVP. It needs clean daily market data, same-day public events, issuer/listing mappings, peer and sector context, and curated exposure tags so the system can explain what is known and what remains uncertain.
Minimum viable dataset
| Dataset | Required fields | Why it matters | MVP source type |
|---|---|---|---|
| Issuer master | issuer_id, legal name, display name, ISIN, sector, industry, market-cap bucket | Prevents wrong-company incidents and powers company pages. | Manual seed plus exchange security masters |
| Listing master | listing_id, exchange, symbol, series, ISIN, F&O flag, active status | Separates the company from its exchange-specific traded listing. | NSE/BSE security files and manual verification |
| Daily OHLCV | Date, open, high, low, close, previous close, volume, value traded | Computes returns, gap, range, volume anomaly, and move diagnostics. | Official bhavcopy/report files |
| Delivery data | Delivery quantity and delivery percentage | Adds quality of participation context to large moves. | Official exchange reports where available |
| Index and sector moves | Nifty return, sector index return, breadth, top movers | Separates stock-specific moves from market-wide or sector-wide moves. | Exchange index reports |
| Corporate filings | Event title, timestamp, category, summary, source URL, document hash, raw text | Gives direct public evidence for filing explained and earnings reaction cards. | NSE/BSE corporate announcements |
| Financial results | Revenue, profit, margin, segment details, period, source document | Explains earnings reactions without inventing beat/miss language. | Exchange filings/XBRL/manual extraction |
| Peer map | Company-to-peer relationships with confidence and source | Detects peer divergence and sector read-throughs. | Manual curated knowledge graph |
| Sensitivity tags | Commodity, FX, rates, policy, import/export exposure edges | Lets the system connect macro or commodity events to sectors and stocks. | Manual curated knowledge graph |
| Macro/commodity context | Crude, USD/INR, yields, rates, policy events, timestamp, source | Powers sector read-through cards and contextual hypotheses. | RBI/official sources and licensed commodity/FX feeds later |
Nice-to-have later
Consensus estimates
Required before the system can safely use beat or miss language in earnings cards.
FII/DII and flow data
Useful for context, but not enough by itself to claim causality.
Options and futures data
Useful for positioning cards after the trust layer is proven.
Intraday bars
Useful for delayed or event-time cards, but not necessary for the post-market MVP.
Data quality rules
- Every price bar must include source ID, date, listing ID, and previous close.
- Every filing or document must include a source URL, source timestamp, hash, and parse status.
- Every derived claim must reference evidence IDs.
- Missing values should be explicit. For example, order margin and execution timeline should be listed when absent.
- Commercial redistribution rights must be reviewed before public launch.