The year-end Letter promised a longer note in January on what the first Ultra cohort taught us. The first Ultra candidates signed up in May 2025, ran the six-month window, and exited in November. This is the read of that cohort from one quarter on, with the figures anonymised and the patterns aggregated. The next read will be in April, when the spring cohort closes.
The cohort, in figures
One hundred and forty-two candidates signed up to Ultra during the May–June launch window. Of those, one hundred and twenty-eight completed the six-month run in good standing on eligibility — the diagnostic, the three Mock Exams spaced out by the required six weeks, the sixty hours of adaptive practice. Fourteen did not complete eligibility and therefore did not trigger any contract, in either direction. That is what "eligibility" is for; it filters the cohort against whom the floor is being measured down to candidates who used the method as designed.
Of the one hundred and twenty-eight who did complete eligibility, one hundred and sixteen sat the official test inside or shortly after the Ultra window and reported their score back. The 715+ floor applies to every Ultra customer, regardless of starting band — no diagnostic threshold, no exclusions inside the cohort.
Of the one hundred and sixteen who reported their live test back, one hundred and fourteen hit 715 or above. Two did not. Both had their Ultra windows renewed for another six months at no charge, under the contract-as-written remedy. The first Ultra cohort therefore closed at 98 percent against the floor — a figure we are reporting because it is the figure the contract was written to produce, not because we think it will hold at the same level forever.
What the cohort that hit the floor did differently
Three patterns separated the candidates who cleared the floor from the two who did not.
Adaptive practice density inside the window. The floor-hitters averaged 78 hours of adaptive practice across the 180-day window. The two who missed averaged 61 — barely clearing the eligibility floor of 60. The cohort-wide pattern across the rest of the one hundred and fourteen was that 70+ hours of practice predicted floor-clearance roughly four times as well as a diagnostic-band-only model did. Eligibility met is not the same thing as practice completed.
Section Analytics engagement. Section Analytics shipped in private alpha in April; Ultra candidates were the first cohort outside the original fifty to receive it. The candidates who opened the Section Analytics panel at least twice a month — to read the predicted band, the time-per-question drift, the per-section percentile — cleared the floor at a materially higher rate than the candidates who did not. One hundred and twelve of the one hundred and fourteen floor-clearers opened the panel regularly. The two who missed did not.
Mock Exam spacing. The contract requires three Mock Exams across the window with at least six weeks between consecutive ones. Candidates who completed the three Mocks evenly spaced — roughly forty days apart — hit the floor at a higher rate than candidates who completed the three Mocks at minimum spacing and stacked toward the end. Both groups cleared eligibility. Only one group cleared the floor as cleanly.
The two candidates who missed
Both candidates who missed the 715+ floor under eligibility completed the diagnostic, the three Mocks, and the sixty practice hours. We are not naming the candidates or surfacing identifying details; what is worth recording is what they had in common.
Both started with diagnostic scores in the lower portion of their cohort's range. Both had a Quant trajectory that improved roughly on the engine's predicted curve and a Verbal trajectory that did not — Section Analytics had flagged both candidates' Verbal pacing as drifting in the back third of the section in months three through five. Both candidates' Mock Exam scores in the last six weeks of the Ultra window predicted a final test score in the 690–705 band, with a confidence interval the engine reported honestly as widening rather than narrowing into the test date.
Both renewals were processed under the contract as written. There was no dispute. The candidates were notified of the renewal pathway when their final Mock sat outside the floor's predicted range and chose to sit the live test anyway — which is the right of every Ultra candidate, and the situation the renewal remedy was written for.
What we would recommend a candidate try
Three changes the data above suggests, if you are currently inside an Ultra window or planning one.
Treat the 60-hour eligibility minimum as a floor on practice, not a target. Plan for 75 to 90 hours of adaptive practice across the window — roughly five hours a week, distributed. The floor-hitters in this cohort were not heroes; they were candidates who treated the 60-hour line as the bottom of the band rather than the top.
Open Section Analytics regularly. The surface is not a vanity dashboard. It is the engine's read on what your current vector says about your next mock. Two opens a month is enough — once when a new Mock score lands, once between Mocks to check what the per-section percentile is doing. The candidates who skipped it were not punished by us; they were punished by not having the read.
Space the three required Mocks evenly. Roughly forty days between consecutive Mocks across a 180-day window. The eligibility floor allows for tighter spacing; the data says the tighter spacing does not pay off. We will look at adjusting the contract's minimum spacing in the next revision.
What we want to see in 90 days
The April note will read the spring cohort. Two specific figures we want to test against this cohort's:
First, whether the 78-hour-versus-61-hour practice gap separating floor-hitters from misses holds at a larger sample. A 142-candidate cohort with two misses is a small sample. The spring cohort will be roughly three times that size, and a stable practice-density threshold across the two cohorts is the figure we would actually trust as a recommendation.
Second, whether the Section Analytics opening pattern holds. The Ultra-alpha cohort that received the surface from launch had the regular-opener pattern almost universally; the spring cohort, which will come in with the surface already on the dashboard rather than as a new feature, may have a different opening baseline. The number worth tracking is whether the engagement pattern still predicts the outcome at the same rate when the surface is no longer novel.
We will write the April note when the data is one quarter older. The consolidated guarantee document the year-end Letter forecast ships ahead of that — same terms as the November contract and the May Ultra addendum, written as one canonical page.
— Brightroom Research