Method version 1.0
How we score promises
Every promise gets a whole-number rating from 0 to 10 using one fixed rubric. A manifesto's score is the importance-weighted average of its ratings, times ten. The same rules apply to every party, and every verdict links to its sources.
This page runs in serious mode: no jokes, just the rules.
What counts as a promise
- Measurable: has a number, a named law or scheme, or a clear yes/no outcome (“build 2 crore houses”, “pass a law on X”). Scored.
- Directional: a clear direction with no number (“increase spending on health”). Scored only if a reasonable public indicator exists.
- Vague: “empower youth”, “make India a global leader”. Not scored. Counted in the manifesto's Gol Gol Index: the share of tracked promises that are vague (lower is better).
Deadlines are the ones the manifesto states. Where none is stated, the deadline is the end of the term. Importance: headline (weight 3: cover, summary or a main campaign plank), major (weight 2: specific, funded or legislative), minor (weight 1).
The rubric
| Status | Also shown as | Rule → rating |
|---|---|---|
| Kept | Vaada Poora | Kept: achieved as promised, on time or within 6 months of the deadline → 10 Kept with a small dilution → 9 |
| Kept late | Der Aaye Durust Aaye | Kept late: achieved 6 months to 2 years after the deadline → 8 Kept late: achieved more than 2 years after the deadline → 7 |
| Partially kept | Aadha Adhura | Partially kept: 75%+ of the target met, or a law passed in diluted form → 6 Partially kept: 50–74% of the target met → 5 Partially kept: yes/no promise only partly done → 5 Partially kept: 25–49% of the target met → 4 Partially kept: under 25% of the target met → 3 |
| In progress | Kaam Chalu Hai | In progress: deadline not reached; rating is provisional and not counted in the final score |
| Stalled | Tareekh pe Tareekh | Stalled: deadline passed or 2+ years idle, but funded or a bill was introduced → 2 Stalled: deadline passed or 2+ years idle, only announced → 1 |
| Not started | Coming Soon™ | Not started: no action found after the deadline → 0 |
| Broken | Hawa Hawai | Broken: abandoned, reversed or officially dropped → 0 |
| Not scorable | Gol Gol Baatein | Not scorable: too vague to measure; counts toward the Gol Gol Index |
A partially kept promise is rated by the share of its target met: 75%+ (or a law passed in a diluted form) → 6; 50–74% → 5; 25–49% → 4; under 25% → 3. A yes/no promise that is only partly done gets 5. Meme names describe the state of the work, never anyone's motive; none of them means “lie”.
The Manifesto Score
Manifesto Score = 10 × Σ(weight × rating) ÷ Σ(weight)
- Final score: after the term ends, over every promise with a non-provisional rating. It changes only through an accepted correction, and each change is logged.
- Live score: during a running term, in-progress promises count at their provisional rating. Labelled “Live” everywhere.
- Only a governing party's manifesto is scored. Opposition manifestos are shown as promise profiles with a Gol Gol Index.
- Grades: 81–100 A (Vaada Pakka) · 61–80 B (Dum Hai) · 41–60 C (Theek-Thaak) · 21–40 D (Kaagzi Sher) · 0–20 F (Hawa Mahal).
- Party leaderboard score: the same formula pooled over all of a party's scored manifestos within one era and level (Centre or states).
Coverage tiers and comparability
| Tier | Elections | Depth |
|---|---|---|
| T1 | Lok Sabha 2014, 2019, 2024 and current state cycles | Full tracking |
| T2 | Lok Sabha 2004, 2009, UPA Common Minimum Programme | Full tracking, thinner evidence |
| T3 | Lok Sabha 1977–1999 | Headline promises only, mostly low-confidence |
| T4 | Lok Sabha 1951–1971 | Archive: text and context, never scored |
- Leaderboards never mix eras: modern (T1/T2) and historical (T3) are separate.
- The number of rated promises is shown next to every score. A 90 on 12 promises is not a 70 on 400.
- When more than 20% of a manifesto's verdicts are low-confidence, the score is flagged as a rough band.
- Current data is a seed set: headline and major promises, not every line of every manifesto. Coverage notes on each manifesto say how much is tracked.
How a verdict is made
- The manifesto is archived and its text extracted page by page.
- Promises are extracted (AI-assisted) and each quote is checked against the manifesto text.
- Evidence is gathered, preferring official sources (PIB, Parliament answers, budgets, CAG, gazette, ministry dashboards), backed by independent reporting.
- A verdict is drafted: status, rating, rubric line, confidence and a plain-language reason.
- A first editor reviews it; a second, different editor approves it. Headline promises need a senior editor. Nobody reviews their own draft.
Current state (beta): most verdicts are AI drafts: researched from cited sources by AI agents, not yet signed off by two human editors. Every such verdict carries a “draft” label and share cards carry a draft banner. AI never sets a final status.
Corrections
Anyone can submit a correction with proof. Only evidence changes a verdict, never the number of complaints. Every decision is in the public log, and rejected tickets can be appealed once.
Method changelog
| Version | Date | Change |
|---|---|---|
| 1.0 | 2026-10-04 | First published method: 0–10 promise ratings, weighted Manifesto Score out of 100, four coverage tiers. |
A method change recomputes every score and is recorded here.