Olex delivers consistent, examiner-level marking so teachers can spend less time on assessment admin and more time on what matters: teaching. Independently tested and verified against working examiners.
Book a Demo ›Teachers stay in control. Olex gives them consistent, reliable marks so they can focus on feedback, planning, and their students.
“Put 30 essays through Olex and every single one would be within a few marks of what I would give it as a senior examiner. This is the most accurate marker I have ever come across.”
“As an AQA examiner, I’d say Olex marks are ‘fair and reasonable’ 99.9% of the time and likely more accurate than the markers in your department. We’ve been double-marking 10% of scripts by hand throughout the year, and the Olex marks have held up almost every time.”
Accuracy is not only about the mark. The written formative feedback Olex produces is accurate, detailed and personalised, drawing on references from each pupil's own writing. Teachers tell us it gives pupils specific, usable next steps, not generic comments.
The study began with 98 scripts marked independently by Olex and multiple experienced examiners under blind conditions. This established the human comparison used throughout the results.
The evidence base was then expanded with a further 59 scripts carrying standardised seed marks supplied by a relevant awarding organisation. Olex has now been evaluated across 157 scripts in total.
RMSE (root mean squared error) is the average distance between the Olex mark and the official examiner mark. Lower is better.
| Metric | Experienced examiners | Olex ★ |
|---|---|---|
| RMSE (lower = better) | 4.98 | 2.1357% lower |
| Scripts within examiner tolerance (10%) | 51% | 88%+37pp |
| Scripts within examiner tolerance (12.5%) | 57% | 91%+34pp |
| Scripts within examiner tolerance (15%) | 70% | 99%+29pp |
| Worst single difference | 11 marks | 5 marks55% smaller |
| Metric | Experienced examiners | Olex ★ |
|---|---|---|
| RMSE (lower = better) | 4.32 | 2.0752% lower |
| Scripts within examiner tolerance (10%) | 62% | 78%+16pp |
| Scripts within examiner tolerance (12.5%) | 68% | 93%+25pp |
| Scripts within examiner tolerance (15%) | 77% | 97%+20pp |
| Worst single difference | 9 marks | 5.33 marks41% smaller |
The experienced-examiner figures come from the original 98-script blind-marking phase. The Olex tolerance results reflect the expanded 157-script analysis. The human figures remain fixed because the participating examiners did not mark the additional 59 scripts.
The RMSE figures come from the initial phase, where Olex and the experienced examiners marked the same scripts.
Olex is evaluated across three tolerance bands, providing a transparent view of its performance against both the strictest benchmark and the wider variation encountered in live marking. This gives teachers confidence in every mark.
The 10% tolerance is treated as the gold-standard benchmark. The 12.5% tolerance is used for the headline result, while the 15% band provides additional context.
Tolerance measures how close a mark is to the established reference mark, relative to the marks available for the task. For example, on a task worth 40 marks: 10% tolerance represents four marks in either direction; 12.5% tolerance represents five marks in either direction; 15% tolerance represents six marks in either direction.
Results based on blind-marking 157 GCSE scripts across English Literature and Language, covering multiple attainment levels and a range of texts and papers. All marks, from Olex and from experienced examiners, were compared independently against an established seed mark. Tolerance bands of 10%, 12.5% and 15% are drawn from published inter-examiner research; 12.5% is the most commonly cited live-marking threshold.
Full methodology
The study evaluated 157 anonymised GCSE English scripts in two phases.
The combined evidence base includes 124 English Literature scripts and 33 English Language scripts across different attainment levels, task types and mark totals. The study includes:
English Literature and Language are reported separately because their task types, mark ranges and assessment demands differ.
Phase one: blind marking by experienced examiners
The initial 98 scripts were marked independently by Olex and multiple experienced examiners. The examiners did not see one another's marks or the marks generated by Olex. Olex did not have access to the examiner marks or established reference marks during marking. This phase established the comparative human results and produced the direct RMSE comparison between Olex and the experienced examiners.
Phase two: standardised seed-mark validation
The study was expanded with a further 59 scripts carrying standardised seed marks supplied by a relevant awarding organisation. The seed marks were established before Olex assessed the scripts and were not visible to Olex during marking. No additional human marking was conducted during this phase. Its purpose was to test Olex against a larger set of authoritative, pre-established reference marks. The participating examiners' figures therefore remain unchanged, while the Olex tolerance results reflect the expanded 157-script evidence base.
Measuring accuracy
Every Olex mark was compared with the appropriate established reference mark for that phase of the study. Accuracy was evaluated using:
The 10% tolerance is treated as the gold-standard benchmark. The 12.5% and 15% bands are also reported to show how performance changes across wider tolerance ranges.
Confidentiality and study integrity
The identities, schools and affiliations of the examiners who participated in the blind-marking study are kept confidential. We also do not identify the awarding organisation that supplied the standardised seed marks used during the expanded phase.
The underlying source materials, examiner-level results and identifiable student scripts are not published where doing so could compromise examiner confidentiality, student privacy, confidentiality requirements attached to the source data, or the integrity of future validation work.
These safeguards do not prevent transparency about how the study was conducted. We publish the total number of scripts tested, the Literature and Language breakdown, the size and purpose of each study phase, which figures relate to the human comparison, which figures relate to the expanded Olex analysis, the tolerance definitions, the accuracy measures, and the aggregate results for each subject.
Get a 30 minute demo from our team, to see how Olex applies assessment criteria, generates marks and keeps teachers in control. Simply pick a time below, or get in contact.