Transparency
Methodology
How every deal is ranked. Everything below is rendered from the live profile configuration and real scored deals.
Loading methodology…
Transparency
How every deal is ranked. Everything below is rendered from the live profile configuration and real scored deals.
Loading methodology…
| Component | Max points | How it is computed |
|---|---|---|
| Sector fit | 30 | Gemini rates the deal's sector and sub-sector against your core and watch sectors (0–1), cached per sector × sub-sector × mandate, then × weight. |
| Stage fit | 20 | In your stages: full. Adjacent stage or one step away: half. Two steps: one fifth. Otherwise 0. Unnamed rounds score 0. |
| Cheque fit | 10 | Round size inside your band: full. Within ±50% of the band: half. Outside: 0. Undisclosed: 40% (neutral, never penalised). |
| Tracked investor | 12 | A watchlist investor leading: full. Participating: 7/12. Existing: 4/12. Best one counts. |
| Geography fit | 5 | Primary geography: full. Secondary: 60%. Another city: 20%. City not reported: 0. |
| Recency | 5 | Linear across the digest window: newest full, oldest 0. |
| Ecosystem significance | 18 | 0.5 × cheque percentile in its stage×sector bucket + 0.3 × investor activity percentile (from our own data) + 0.2 × min(sources ÷ 4, 1). Mandate-independent. |
Mandate: stages Pre-seed, Seed, Pre-Series A, Series A; round size $500K–$20M; core sectors Fintech, B2B SaaS, Supply Chain & Logistics, Healthtech, AI Applications, D2C & Commerce, Quick Commerce; primary geographies Bengaluru, Delhi-NCR, Mumbai.
The Pant Project · Series B · $11M · D2C & Commerce · 2026-10-19 · via BusinessLine — Companies
| Sector fit | 18 / 30 | D2C & Commerce → unlisted sub-sector premium menswear |
| Stage fit | 10 / 20 | Series B, adjacent to mandate |
| Cheque fit | 10 / 10 | $11M, within $500K–$20M |
| Tracked investor | 0 / 12 | No tracked investor |
| Geography fit | 0 / 5 | City not reported |
| Recency | 5 / 5 | Announced today |
| Ecosystem significance | 12 / 18 | 58th pct cheque for Series B × D2C & Commerce (n=6); investor tier 92th pct (Sorin Investments); 2 sources |
Raw 55 × confidence 1 → shown as 55 (relevant)
+40 when the company is on your watchlist; it always surfaces in the watchlist section.
+35 when it competes with a tracked company (same product, same customer); +25 when it is the same layer (different product, same customer and budget); +15 when it is a supplier, customer or complement.
+10 when a watchlist investor makes its first move into one of your sectors (only once the investor-rotation detector's four guards pass).
Hard filter: excluded sectors (Real-money gaming, Crypto/Web3 tokens, K-12 Edtech) and keywords score 0 and are archived before any scoring. Exclusion is a filter, not a negative weight.
× 0.85 when only one publication has reported the deal and it is unverified.
Scores are clamped to 0–100 for display; the raw value is kept in every deal's breakdown.
We deliberately surface the three most significant deals that did NOT match your mandate, because a perfect filter would recreate the blind spot this product exists to fix.
Selection: up to 3 deals in the current window whose mandate score (everything except ecosystem significance) is below 40, whose ecosystem significance is in the top decile, and whose sector is not excluded.
A detector fires only when all three gates pass: sample size ≥ n_min, effect size ≥ e_min and weeks of history ≥ b_min. A detector without enough history is shown as suppressed with its reason, never silently absent. Status for ISO week 2026-W42.
| Detector | n_min | e_min | b_min | Status | Reason |
|---|---|---|---|---|---|
| Sector momentum A sector's deal count this week vs its trailing 8-week mean, in standard deviations (both directions). σ is floored at 1 deal/week so a sector with a flat baseline cannot produce an infinite z-score (D-051). | 3 deals this week (accelerating) or baseline mean (cooling) | 1.5 σ | 4 wk | quiet | No sector moved ≥1.5σ with ≥3 deals (largest: Edtech at 1.5σ). |
| Stage shift Share of priced capital by stage this week vs the trailing 8-week share. | 5 priced deals this week | 10 percentage points | 4 wk | active (3) | 3 findings. |
| Cheque inflation Median cheque per stage this week vs the trailing 8-week median for that stage. | 4 priced deals at the stage this week | 25 % change | 6 wk | active (3) | 3 findings. |
| Investor rotation |
Sources: 16 live feeds, 2 weekly-roundup sources used for backfill, 15 disabled (reasons on /health).
Deals by ingestion mode: backfill 250 · live 80. History before live operation comes from weekly roundups, not continuous collection.
Deduplication: four layers (domain → normalised name → company embedding plus name agreement → Gemini adjudication), within ±14 days and a compatible stage. Merged deals keep every source.
Field confidence: corroborated = two or more publications agree; single-source = one publication; disputed = publications disagree (amounts by more than 10%, or different named stages). A deal is verified when two or more publications report it and no field is disputed.
Mandate inferred from the problem statement. All values user-editable. Sectors, stages, cheque size, geographies, portfolio and watchlist are assumed defaults, editable in settings.
Simulated portfolio. Aarav is a fictional persona from the assigned problem statement. The companies listed as portfolio holdings are real Indian startups, used so that competitive-adjacency detection can be verified against companies a reader recognises. No investment, advisory or commercial relationship between this project, the persona and these companies is implied or asserted. Watchlist entries are companies and funds being monitored, which likewise implies no relationship.
| 3 deals in the database for the investor |
| — |
| 8 wk |
| active (13) |
| 13 first moves passed all four guards. |
| Thesis clustering Three or more deals this week whose companies form one tight embedding cluster. Cohesion must also exceed the corpus's 99th-percentile pairwise similarity: with this embedding model the median pair already scores ≈0.77, so 0.75 alone would call any three deals a thesis (D-051). | 3 deals in the cluster | 0.75 cohesion (mean pairwise cosine) | 2 wk | quiet | No group of ≥3 deals reached cohesion 0.91 (corpus 99th percentile); tightest triple RapidCanvas, AdGraam, Shamrock AI at 0.87. |
| Syndicate formation An investor pair co-investing repeatedly, including this week. | 3 co-investments | 2 lift (observed ÷ expected) | 8 wk | active (1) | 1 finding. |
| Follow-on compression Median months between a company's rounds, recent 4 weeks vs earlier history. Round pairs come from 'has raised X to date / previous round' mentions plus consecutive observed rounds. | 4 recent round pairs | 20 % shorter | 8 wk | suppressed | Only 19 recent and 0 earlier round pairs; needs 4 of each. |
| Capital concentration Top-3 deals' share of the week's priced capital vs the trailing 8-week mean share. | — | 15 percentage points | 4 wk | active (1) | 1 finding. |
| Geographic shift A city's share of deals this week vs the trailing 8-week share (deals with a reported city). Also needs ≥20 baseline deals with a reported city; weekly roundups rarely name cities, so this stays suppressed until live history accumulates (D-051). | 4 deals in the city this week | 10 percentage points | 4 wk | active (1) | 1 finding. |
| Sector silence A mandate core sector with zero deals for 3 or more consecutive weeks. | — | 3 consecutive silent weeks | 6 wk | quiet | Every core sector had a deal within the last 3 weeks. |
| Outlier deal A deal more than 2σ from its stage × sector distribution (log scale), falling back to stage-only. Amounts are compared on a log10 scale: round sizes are heavy-tailed, so a linear σ is dominated by one mega-round (D-051). | 8 priced deals in the bucket | 2 σ | 6 wk | active (1) | 1 finding. |
| Repeat founderbest effort Founders of two or more of this week's funded startups share a former employer. Best effort: founder backgrounds are rarely in funding news, so this fires rarely by design. | 2 deals | — | 4 wk | quiet | Best effort: 18 founder background(s) reported this week, no shared former employer. |