Skip to content

Three ways to count the ranked sector, and what each is for

The board in README.md reads one ranking edition at one academic year. That is the strictly comparable view, and it is not the only one worth having. This note sets out the three, what each includes, and the one finding that only appears when you compare them.

Source for the second and third: extract_latest_per_college.sql, which takes each institute’s most recent observation carrying outcomes — latest ranking year, then latest academic year within it.


Population Measured at Use it for
A. One edition institutes in NIRF 2025 one academic year, same for all comparing disciplines against each other
B. Currently ranked, last report institutes in NIRF 2025 each at its own last report the ranked sector today, best coverage
C. Ever ranked, last report every institute NIRF has ranked each at its own last report the largest picture the data supports

B gains almost nothing over A. Across fifteen tracks it adds two institutes to Medical UG, one each to Dental and Law UG, and loses one each on Agriculture UG and Pharmacy PG — those two sit in the 2025 metrics table but not in the 2025 rankings table for their category, a small join mismatch worth knowing about. Rates move by under two points everywhere. NIRF’s 2025 edition is already near-complete, so “use each college’s latest data” buys very little within an edition.

C is where the information is.


C: every college NIRF has ever ranked, at its last report

Section titled “C: every college NIRF has ever ranked, at its last report”
Discipline Level Currently ranked Ever ranked × Graduates Placed % Higher study % Oldest vintage
Engineering UG 97 263 2.7× 222,407 70.95% 12.30% 2015-16
Dental UG 38 62 1.6× 4,527 55.31% 29.56% 2018-19
Pharmacy UG 87 128 1.5× 10,139 41.71% 40.64% 2017-18
Agriculture UG 37 47 1.3× 18,082 34.94% 48.74% 2021-22
Medical UG 47 61 1.3× 9,357 58.87% 35.66% 2017-18
Law UG 38 47 1.2× 6,902 63.68% 15.42% 2020-21
Architecture UG 40 45 1.1× 2,743 67.99% 15.17% 2017-18
All PG tracks PG unchanged 1.0× 2022-24

Engineering UG more than doubles: 97 colleges to 263, and 105,542 graduates to 222,407. Most of that is the 101–200 band NIRF published through 2022 and not after.

Every PG track is unchanged at 1.0×, because PG ranking tracks are recent — nobody has dropped out of them yet.


The finding: the current ranked set is positively selected

Section titled “The finding: the current ranked set is positively selected”

Compare each discipline’s placement rate under B and C — the same measure over the current set versus everyone ever ranked:

Discipline Currently ranked Ever ranked Difference
Dental UG 60.67% 55.31% −5.36
Law UG 68.50% 63.68% −4.82
Medical UG 61.23% 58.87% −2.36
Pharmacy UG 43.97% 41.71% −2.26
Agriculture UG 36.03% 34.94% −1.09
Architecture UG 68.00% 67.99% −0.01
Engineering UG 70.62% 70.95% +0.33

Colleges that fell out of the ranking placed worse than the ones that stayed. That is what you would expect — graduate outcomes feed the NIRF score, so weak placement pushes a college down and eventually out — but it means quoting the current ranked set overstates the ranked sector by two to five points in most disciplines.

Engineering is the exception, and instructive. Its 166 departed colleges place at essentially the same rate as the 97 that remain. They did not leave because their outcomes deteriorated; they left because NIRF stopped publishing ranks 101–200 for that track after 2022. A change in the ranking, not in the colleges — which is the cleanest evidence yet that the 2023 discontinuity is about publication depth rather than performance, whether it originated at NIRF or in our load.


Mixed vintage. Engineering UG under C reaches back to 2015-16. A composite rate over colleges measured up to nine years apart is not a snapshot of anything, and should never be quoted as “the 2025 figure”. The extract carries oldest_vintage, newest_vintage and pct_on_newest_vintage per group so the blend is visible; for currently-ranked tracks, 93.6–100% of institutes sit on the newest vintage, so B is nearly clean while C is not.

Survivorship runs the other way too. A college’s last report is often the year before it fell out — plausibly its worst. So C is not an unbiased picture of the ever-ranked sector either; it is the best available one, and its bias points down.

Neither view is “the” answer. A is what the board uses because comparing disciplines needs a common moment. C is the right base for any question about how many colleges and graduates the ranking has ever covered.