A results figure you can actually check
Improvement is the whole point of exam preparation. A student wants to walk into the hall knowing more than they did in September, and a parent wants to know the hours and the subscription bought something real. So improvement is the thing we measure most carefully, and this page shows you exactly how we do it.
We think the method matters as much as the number. A figure you can interrogate is worth more than a figure you have to take on faith, and by the end of this page you should be able to tell us where our approach is strong and where it is limited. We have set out both.
Everything below is the actual procedure: the two comparisons we run, the sittings we set aside before counting, and the standard a figure has to clear before we will show it.
The problem with measuring improvement at all
Suppose a student sits a practice paper in January and scores 45%, then sits a different practice paper in April and scores 60%. Did they improve by fifteen points?
Maybe. Or maybe the April paper was easier. Papers vary. A student moving from foundational material into harder territory can work extremely hard, learn a great deal, and watch their raw score fall, because the questions got tougher faster than they did.
This single problem invalidates most casual measurements of progress. If you compare a student's early scores with their later scores across different papers, you are measuring the papers as much as the student.
There is an obvious fix: compare a student against themselves on the same paper. That holds difficulty constant. But it introduces a different problem, which is that the second time you sit a paper you have seen it before, and part of your improved score is simply recognition.
Neither approach is clean on its own. So we use both, and we publish both.
The two things we measure
| Measure | What it compares | Its strength | Its weakness |
|---|---|---|---|
| Same paper | A student's first sitting of a paper against their second sitting of that same paper | Difficulty is identical, so any change is about the student | The second sitting carries some familiarity with that particular paper |
| New papers | A student's earlier first attempts against their later first attempts, across papers they had never seen | No familiarity is possible, so this is the cleanest evidence of transferable learning | Different papers differ in difficulty |
Each measure covers the other's blind spot. Where they agree, we have something worth saying. Where they disagree, we have a reason to be careful rather than a reason to pick the flattering one.
The rule we hold ourselves to: both measures are published together, and we fix that pair in advance. Choosing what to report before seeing the result is what keeps the reporting steady from one month to the next.
What we throw away before counting
Most of the integrity of a statistic lives in what you exclude. These are our exclusions, and the reasoning behind each.
We only count objectively marked questions
Multiple choice questions have a right answer. They are marked by the same rule every time, for every student, on every sitting. Nothing about the marking can vary, which makes them the soundest available basis for a claim about attainment.
So these figures rest on objectively marked questions alone. Written work is marked and returned as part of a student's preparation and it is valuable to them, but a public claim about outcomes deserves the one form of evidence where marking involves no judgement at all.
It is the firmest ground available, so it is the ground we build on.
We only count papers the student actually attempted
Students open papers and abandon them. They start something, get interrupted, and never come back. Those sittings sit in the record with very low scores attached.
Counting them would be badly misleading in an interesting way: an abandoned first sitting followed by a real second sitting produces an enormous apparent improvement that is entirely fictional.
We exclude a sitting when the student did not work through most of the paper. Importantly, we decide this by looking at how many questions they actually engaged with, not by looking at the score. A student who attempted every question and scored badly is a real result and we count it. A student who answered five questions out of fifty and closed the tab is not, however high or low the score.
We only count students who genuinely practise
A visitor who tries two papers in one afternoon is not evidence of anything. Improvement takes time to happen and time to detect.
A student enters our figures only after sustained use: several sittings, spread over a period of time rather than crammed into one session. We also ignore a repeat that happens minutes after the first attempt, because reopening a paper immediately is one study session, not practice followed by a retry.
Every student counts once
Some students are enormously more active than others. Left unchecked, our heaviest users would dominate the figures, and the published number would describe a handful of people rather than our students in general.
So we average each student's results into a single figure first, and only then combine across students. A student who re-sits forty papers has exactly the same weight in the result as a student who re-sits two.
The standard a figure has to clear
This is the part that most distinguishes a real measurement from a marketing claim.
Any group of students contains a range of results, so an average by itself is not enough to justify a public claim. A small group with a wide spread can produce an encouraging average by chance alone.
Every figure we publish therefore has to clear two tests. The first is statistical confidence: we check that the improvement still holds under an unlucky draw of who happened to be in the group, which in technical terms means requiring the lower end of the confidence interval to stay above zero. The second is that the typical student agrees with the average, because an average can be lifted by a few exceptional results while most students experience something different.
A figure that clears both is one we are willing to stand behind, and it travels with the number of students behind it so you can weigh it for yourself.
How often the figures are refreshed
Weekly. A number that moved every time someone finished a paper would invite you to read meaning into noise, and a weekly figure is settled enough to be worth looking at.
The same method across every board we cover
We apply this method identically whether a student is preparing for Cambridge IGCSE, Pearson Edexcel International GCSE, OxfordAQA International GCSE, or the West African and Nigerian examinations, WAEC, NECO and JAMB.
That consistency is what makes the figures comparable, both between boards and across time. One method everywhere means a change in the number reflects a change in how students are doing, rather than a change in how we chose to count.
There is one practical consequence worth naming. Boards differ in how much of their assessment is multiple choice, and since we only count objectively marked questions, some subjects contribute more data than others. A subject assessed largely through extended writing will be thinly represented in our figures. We name it because a figure is only as useful as your ability to place it.
Commitments we hold ourselves to
- Survey results stay labelled as surveys. Asking students whether they feel they improved measures confidence, which is worth knowing but is not attainment. If we run one, we will say so plainly.
- Every qualifying sitting counts. We do not keep a student's best paper and quietly drop their worst.
- A repeat score is never presented as a predicted grade. It measures something useful, but not that, for the familiarity reason above.
- Every figure travels with its method. If we show you a number, this page is how it was produced.
What this means for you as a student
The same logic that governs our published figures should govern how you read your own practice scores, whether you are preparing for Cambridge, Pearson Edexcel or OxfordAQA IGCSE papers, or for WAEC, NECO and JAMB.
- Treat your score on a paper you have never seen as your honest prediction.
- Treat your score on a paper you have redone as a measure of how well you absorbed that paper, which is useful but different.
- Keep a few papers untouched until close to your exam, so you always have a clean measurement available.
- If your score on a repeated paper stops climbing, stop repeating it. That ceiling is showing you a topic you have not learned yet.
Checking our work
We publish the sample size alongside any improvement figure we show, so you can see exactly how much data sits behind it and judge its weight yourself.
If you are a school considering Green Bridge for your IGCSE cohort and you want the methodology in more detail than this page provides, ask us. We will walk you through it.
Our aim is a figure that still looks right in a year, that a teacher could pick apart without finding a hole, and that a student could reproduce on their own practice scores. A number worth trusting is worth the work it takes to earn.
A plain account of how Green Bridge measures whether students improve: the two comparisons we run, the sittings we set aside before counting, and the standard a figure has to clear before we publish it.
Maoni