Ranking the water transparency of every country on earth sounds like a tidy data project. It was not. The Reservoir Transparency Index went through four methodology versions in its first month, mis-graded India for weeks, and forced us to invent a category for countries that have no reservoirs at all. This is the honest post-mortem of how the H1 2026 edition was built — and what we got wrong before we got it right.
The shape of the job
We evaluated 194 countries. For each one, a research pass had to answer the same questions: which reservoirs exist above the size threshold, who operates them, what is published, in what format, how often, how far back, and under what licence. Every claim had to be backed by a real URL — a portal, a bulletin, a dead link that should have worked.
That produced 167 graded countries and 27 marked not-applicable, each with a per-dimension scorecard and an evidence trail. The country profiles you see on the site are the cleaned-up front of a much messier research process behind it.
Where it went wrong: the India problem
India is the case that broke — and then fixed — our definition of "coverage."
India's Central Water Commission publishes a Weekly Reservoir Bulletin covering roughly 155 strategic reservoirs. Sounds modest. But here is the trap: India has on the order of 1,400 reservoirs above the size threshold. So what is India's coverage?
- If you measure share of reservoirs by count, it is about 155 / 1,400 ≈ 11%. A failing number.
- If you measure share of national storage by volume, those 155 strategic reservoirs hold roughly 71% of India's live storage. A strong number.
In our first methodology, coverage was scored as share of capacity, and India landed near 65. Then a v1.2 revision redefined coverage as share of reservoir count — and India collapsed to 11. For a couple of weeks the index was telling two opposite stories about the same country, depending on which denominator we'd most recently argued ourselves into.
India forced the call. A transparency index should answer the question a citizen actually asks — "what share of my country's water can I see?" — and that question is about volume, not headcount. The long tail of ~1,245 small state-managed reservoirs matters, but it is not where the water is. So v1.3 made coverage capacity-weighted: the percentage of total national storage capacity that has public data. India settled at a coverage of 78 and a final grade of C+ (rank 35) — neither the flattering 65 nor the punitive 11, but the honest figure.
The four versions
The changelog is public, and deliberately so:
- v1.0 — initial release. Seven dimensions, 194 countries.
- v1.1 — coherence patch. Coverage made strictly linear; quality dimensions scoped to the covered subset (you don't get credit for accuracy on reservoirs you don't publish); the typical reservoir assessed at the median, not the best showcase example.
- v1.2 — weights rebalanced. Coverage raised from 15% to 30%, because it bounds the meaningfulness of every other dimension.
- v1.3 — coverage redefined from count-based to capacity-weighted, eliminating the arbitrary ">10 hm³" threshold and resolving both the India problem and the small-country bias in one move.
Four revisions in a month is not a sign the methodology was sloppy. It is a sign we let real countries stress-test it and changed the rules when a country exposed a flaw — rather than defending a number we knew was wrong.
Two design decisions worth defending
Median, not best. Early on it was tempting to grade a country on its most impressive reservoir. That rewards showcasing. A nation with one gorgeous real-time dashboard for a flagship dam and silence everywhere else is not transparent — it is performing transparency. Scoring the median covered reservoir kills that incentive.
Not-applicable is not zero. Twenty-seven countries — Bahrain, the Bahamas, Barbados, Cape Verde, Comoros and others — have no qualifying reservoirs at all. They are desert islands, coral atolls and karst limestone states that run on desalination and groundwater. Scoring them zero would punish geology and pollute the rankings with fake failures. We mark them structurally N/A, with a documented reason and water source for each, and exclude them from the graded table. A transparency index has to be honest about the difference between "hiding the data" and "there is no data to hide."
What this enables
Because every score decomposes into seven documented dimensions with source URLs, the RTI is not a black box you have to trust. You can open any country profile, see exactly why it scored what it did, and challenge it. We have already re-graded countries on reader corrections, and we will again.
That is the deal: a global ranking is only as credible as its willingness to show its work and fix its mistakes in public. India was our most public mistake. It made the index better. If you find the next one, the methodology page shows you exactly where to push.
For the headline results this all produced, see Introducing the RTI — H1 2026.