How we grade, and why you can argue with it
Make India Best is a curated belief map, not ground truth. Everything here can be inspected, challenged, and improved. Below is how we assemble the claims and how we mark our own uncertainty.
Honesty principle
We do not present curation or community opinion as evidence. Evidence grades come from a published rubric and the sources behind it. If you see something wrong, the goal is to make it easy to challenge and correct.
The problem graph
The map shows 28 problems and 204 directed edges. An edge A → B means "A causes or worsens B" in the current curation.
Each node is placed in a tier from 0 (Given conditions) to 4 (Outcomes). Tiers encode human judgement about how far upstream a problem sits. They are visible everywhere so you can disagree with them.
Solution evidence rubric
Every solution is tagged with one of four grades. The grade is a statement about the quality of evidence, not about whether the idea is good or bad.
- Strong
- Replicated evaluations or large-scale rollouts show the effect. Disagreement is about design, not direction.
- Mixed
- Real evidence exists on both sides, or results swing hard with implementation quality and context.
- Thin
- Plausible mechanism, but little rigorous evaluation at Indian scale. Treat as a bet, not a finding.
- Contested
- Serious people disagree about whether this helps at all. Listed because it is argued for seriously, not because it is endorsed.
Voting
Signed-in accounts can upvote or downvote problems, solutions, and causal edges. Each account gets one vote per item. Votes are stored server-side so they cannot be inflated by refreshing the page.
Tallies are hidden until at least 10 votes have been cast. This keeps early impressions from looking like settled community opinion.
Rootness formula
The "model score" you see on problem pages combines tier judgement with structural reach. It is one lens, not the answer.
structure = Katz-style damped walk (α=0.85, depth 12)
with in/out share-normalised edges
tierScore = (1 − tier/4) × 100
rootness = normalise(0.6 × tierScore + 0.4 × structure) → 0..100
Confident pair: |rootness(a)−rootness(b)| ≥ 18 AND tier(a) ≠ tier(b)Known biases
- The graph is one strongly connected component — every problem reaches almost every other. There is no natural upstream order readable from structure alone.
- Out-degree reflects curation attention, not causal certainty. Some problems have many outgoing edges because the curators reached for them often as an explanation. That is why they score high in the model — and why we show the reasoning, not just the number.
- Evidence grades are based on what has been measured, mostly in economics, education, and public-health literatures. Some domains are harder to evaluate than others, and a "thin" grade often means "not studied enough" rather than "does not work."
Disagree?
Good. The site is built to make disagreement cheap. Vote on the specific item you disagree with, check the sources, and send feedback on LinkedIn if you have a correction or a better source.