Students who started below 50% gained an average of 24.2 percentage points, and 92% of them improved. Drawn from 10,000+ marked past papers between 2018 and August 2026, compared like-for-like within the same paper type, the average went from 71.6% to 78.2% — a gain of 6.6 points, with 63% improving and 37% gaining 10 points or more. The lower a student starts, the more the tutoring moves them; students already above 80% hold station rather than climb, because there is little room left on the same paper.
Improvement by where the student started
This is the finding that matters, and the reason a single average is misleading. Each record is one student, in one subject, at one paper type, with at least 3 marked papers.
| Started at | Average change | Improved |
|---|---|---|
| Below 50% | +24.2 pts | 92% |
| 50–65% | +12.5 pts | 76% |
| 65–80% | +5.5 pts | 68% |
| 80% and above | -1.0 pts | 45% |
The bottom row is published for the same reason as the top one, and it is the row that changed most when we corrected our method. It used to read −4.9 points, because we were comparing a student’s first paper with their most recent one even when we had moved them onto a harder paper in between — foundation, then new-spec foundation, then old-spec higher, then new-spec higher. Escalating a student is what we do when they get good, and 63% of records in this band crossed a tier boundary, so the old figure was measuring a jump in difficulty and calling it a fall in the student. Comparing like-for-like, the band reads −1.0. A student marking 88% has little room left on the same paper type. If a tutoring company shows you only its best band, that is the row it left out.
By level
| Level | First paper | Latest paper | Average change |
|---|---|---|---|
| GCSE | 71.4% | 78% | +6.6 pts |
| A-Level | 69.2% | 76.3% | +7.1 pts |
| IGCSE | 76.5% | 80.9% | +4.3 pts |
| 11+ | 87.3% | 94% | +6.7 pts |
By subject
Every subject with enough records to report, ordered by average change. None is left off.
| Subject | Average change |
|---|---|
| GCSE English Language | +11.4 pts |
| A-Level Maths | +10.2 pts |
| GCSE Chemistry | +7.9 pts |
| GCSE Biology | +7.2 pts |
| A-Level Chemistry | +6.5 pts |
| GCSE Maths | +5.7 pts |
| GCSE Physics | +5.6 pts |
| IGCSE Maths | +5.5 pts |
| A-Level Biology | +5.3 pts |
GCSE Maths sits mid-table at +5.7 points, and it is the subject our method correction moved furthest: it used to read +1.2 points, because Maths is the most heavily tiered subject we teach and so the one where comparing across paper types did most damage. A-Level Biology is now our weakest published result at +5.3 points. It is on the table because leaving it off would make every other number less trustworthy.
How this was calculated
- Source: Ucademy's own marked past papers from 2018 to August 2026, drawn from 10,000+ marked past papers — the benchmarking records logged in our platform, together with the in-person paper records kept before the platform carried them.
- A record is one student, in one subject, at one paper type. A record is included only if that student marked at least three papers of that type, so a single paper cannot create an apparent trend.
- Comparisons are like-for-like. Students are escalated to harder papers as they improve — foundation, then new-spec foundation, then old-spec higher, then new-spec higher — so a record never spans a change of level, subject, tier or specification. Improvement is the difference between the first and most recent paper within one paper type, in percentage points.
- The median record covers 5 papers over 6 weeks.
- No record is excluded for performing badly. The 80%-and-above band averages −1.0 points and is published unchanged.
- Correction, 15 September 2026: the earlier version of this page compared each student's first paper with their most recent one regardless of paper type. Because 63% of the highest-starting records crossed a tier boundary, that method recorded a jump in paper difficulty as a fall in the student, and the 80%-and-above band read −4.9. Holding paper type constant corrects it to −1.0. Every other band improved.
What these figures cannot tell you
These are past papers, not final grades
This page measures movement on marked past papers, which is a different thing from a grade awarded by an exam board. Ucademy does hold official exam-results data, but it is a separate dataset and none of it is mixed into the figures here — it will be published on its own page rather than blended into these.
There is no control group
We compare students to their own earlier scores, not to similar students who did not use Ucademy. Some of this improvement would have happened anyway through ordinary school teaching and practice.
Students who start high hold station rather than climb
The 80%-and-above band averages −1.0 points. Most of the steeper fall this page reported before was an artefact of students being escalated to harder papers, and correcting for that lifted the band from −4.9 to −1.0. What remains is a ceiling: a student marking 88% has little room left on the same paper type, and an unusually strong first paper is usually followed by a more typical one. The same effect flatters the below-50% band. Both are shown.
Only committed students are counted
Requiring three papers excludes students who logged one and stopped. That selects for the more diligent, which is a genuine limitation of the figure and not a neutral filter.
Nobody independent has checked this
The figures are ours, calculated from our own records. They have not been audited by an external evaluator, and the underlying records are student data we cannot publish. The method above is published in full so the reasoning can be challenged even where the raw data cannot be handed over.
Questions about these results
How much do Ucademy students improve?
Drawn from 10,000+ marked past papers and compared like-for-like within the same paper type, the average student went from 71.6% to 78.2% — a gain of 6.6 percentage points. The gain depends heavily on where they started: students beginning below 50% gained 24.2 points on average, and 92% of them improved.
Are these real exam grades?
No. These are marked past papers, not grades awarded by an exam board, so no claim is made on this page about final grades. Ucademy does hold official exam-results data, but it is a separate dataset and none of it is mixed into these figures — it is published separately rather than blended in here.
How many papers are these figures based on?
They are drawn from 10,000+ marked past papers between 2018 and August 2026. A record is one student, in one subject, at one paper type, and it counts only once at least three papers of that type have been marked, so a single paper cannot create an apparent trend.
Do all Ucademy students improve?
No. 63% of records improved overall. Students who started at 80% or above average −1.0 points, with 45% improving — they hold station rather than climb, because there is little room left on the same paper type and an unusually strong first paper is usually followed by a more typical one. That band is published unchanged.
Has anyone independent verified these results?
No. The figures are calculated by Ucademy from its own platform data and have not been audited by an external evaluator. The full method is published so the calculation can be checked and challenged.