GMAC published an updated percentile table yesterday. The new figures are based on the population of test-takers between July 2019 and June 2024 — the first refresh that includes a full year of Focus Edition data, and the first one large enough to materially move the band at the top of the curve.
Most of the early coverage has framed this as "705 is no longer 99th percentile." That sentence is true and not especially useful. Here is what actually changed, what it means for a candidate sitting the test this fall, and what we updated.
The headline shift
On the August 2024 table, a 705 maps to the 98.2nd percentile. On the previous (early-2024) table, the same score read as 99th. To recover a 99th-percentile reading under the new figures, a candidate needs roughly a 745 — a 40-point difference, on a scale that increments by ten.
The middle of the curve barely moved. Scores between 555 and 635 sit within one percentile point of where they sat in February. The lower bands compressed slightly downward, meaning the same scaled score is now slightly more competitive at the lower end. The entire shift is concentrated above 695.
Why this happened
Two forces. First, the Focus Edition has been running long enough that the rolling five-year window now contains a meaningful slice of Focus data, which scales differently than legacy data and therefore re-anchors the percentile baseline. Second, and more interesting, the population of test-takers has shifted in shape over the same window. Verbal performance is slightly weaker on average than it was in 2019. Quant is significantly stronger.
Verbal scaled mean (rolling, 2019-2023): ~80.2Verbal scaled mean (rolling, 2019-2024): ~79.8Quant scaled mean (rolling, 2019-2023): ~80.4Quant scaled mean (rolling, 2019-2024): ~81.6Verbal percentiles rose by roughly one point at most bands. Quant percentiles fell by one to two points across most of the curve. The reason most commonly cited — and consistent with the population data GMAC has published — is the rising share of international applicants whose quant performance is on average stronger than the historical baseline.
Whatever the cause, the consequence for a candidate is the same. The same Quant scaled score buys less percentile ranking than it did three months ago. The same Verbal scaled score buys slightly more.
What this means for a target score
If a candidate has a percentile target — say, "I want to be at the 95th percentile" — the new table requires a higher scaled score than it did under the previous one. We have seen application consultants quietly raise the de-facto "M7-competitive" floor over the last week. The publishable version of the consensus, as best we read it, is that a candidate who would have aimed at 705 in early 2024 should now be aiming at 715 to 725 against the new table.
If a candidate has a scaled-score target — "I want a 705" — the score itself has not gotten harder to obtain. The underlying item parameters have not been re-fitted. What has changed is the company that 705 puts you in.
Most candidates we work with operate under some mixture of the two — a scaled-score target tied to a percentile benchmark. The mixed target is the one that has moved.
What we updated
Three things changed in the Brightroom engine this week.
The predicted-band model now reports against the August 2024 table. Every diagnostic and mock-exam result that surfaces a percentile alongside a scaled score uses the new percentiles. Older sessions are re-percentile-coded on read, not on write, so the historical scaled scores remain stable while the percentile column updates everywhere it appears.
The internal target-score recommender has been re-anchored. Candidates who set a percentile-based target (e.g. "top 1%") get the recommendation against the current table. We do not retroactively change a candidate's scaled-score target, because that is a contract; we do surface a small note when the percentile equivalent of their target has moved by more than half a band.
The guarantee bands are unchanged. Our guarantee is written in scaled-score points, not in percentile ranks, precisely because percentiles drift and scaled scores do not. A candidate who was promised a +70 lift to 705 is still promised a +70 lift to 705. The external meaning of 705 has moved; the internal contract has not.
What we did not change
We did not re-calibrate item parameters in response to the percentile table. The (a, b, c) values on every item are estimated from response data, not from population percentiles. A change in how scaled scores map onto percentiles is downstream of item calibration, not upstream of it.
We mention this because at least two competing platforms have published "we have re-calibrated our adaptive engine in response to the percentile change" claims this week. That sentence does not describe a coherent psychometric operation. If the underlying item difficulties did not move, the engine has nothing to re-calibrate. The percentile column on the report should change. The engine should not.
What to do if you are sitting in the next sixty days
Two adjustments.
Re-read your target. If you set your target score earlier this year against percentile-anchored advice (admissions consultants, school medians published before July), the percentile equivalent of that target has likely moved. Adjust the scaled-score target up by 10 to 20 points if the percentile match matters to you. Leave it alone if the scaled score was always the target.
Do not chase the rebase. Two weeks of intense Quant grinding to "make up" the percentile slip is the worst possible response to a population-level statistical adjustment. The marginal cost of pulling a Quant scaled score from 82 to 84 has not changed. What has changed is what 82 means. Train the skill, not the statistic.
Closing note
GMAC publishes an updated table every August. We will publish a brief like this each time. The next one will be in roughly twelve months, and barring a structural change to the test itself, it will move a smaller distance than this one did.
— Brightroom Research