For administrators, rankings are not valid measures of their Uni’s performance relative to others. Don't view overall or component scores as ‘key performance indicators’, but develop and validate own metrics of research and teaching quality. Interpret ranking data with extreme caution! (8/8)

It suggests QS maintains a ‘pecking order’ - the data are manipulated. What are the implications? Students should do their own research on Unis they are interested in, and ought not to view changes in scores or rankings as necessarily reflecting changes in quality. (7/8)

The overall and component scores are not measures of Aus/NZ universities in isolation – there is a ‘global’ effect on the scores as a group. What does this mean? QS uses a proprietary methodology so it’s impossible to say for sure. However… (6/8)

Now, there were changes in QS methodology that can explain some of the shared variance (in 2024 QS added component scores that Aus&NZ did well on, and the 2026 data were rescaled). But if you look at the component scores, a similar pattern emerges – the data are not fully independent (5/8)

The year-on-year changes are highly correlated! Scores for almost all Unis increased from 2023 to 2024 and then again from 2025 to 2026. Overall the % of shared variance is 83.3% - much more than would be expected be if these scores were independent. (4/8)

There is a stable pecking order – some Unis (e.g. Melbourne & UNSW) are generally at the top, others are consistently in the middle and some are at the bottom. Let's remove the pecking order by mean-centering the scores, so that each Uni’s average is zero: (3/8)

QS compute rankings from an Overall Score which is determined by component scores (Academic and Employer Reputation, Citations/Faculty, Faculty/Student ratio…). Here are the Overall Scores for 29 Australian/NZ universities that were ranked in each of those years (2/8):