Exam Marking & Grade Boundaries: Mark Schemes, Double Marking, Moderation Adjustments, Boundary Setting from Distributions & Component Weighting

What you'll build

What you will be able to do

Each lab ends with a concrete result that you build yourself and the checker confirms, cell by cell.

• Total the question marks of ten fictional exam scripts, express each total as a percentage of the maximum, count marks that exceed a question's maximum and highlight scripts with missing marks before any result is released.
• Compare two independent marks for each script, apply a fictional tolerance rule to find the agreed mark, measure how often the markers disagree and whether one of them is systematically harsher.
• Compare three fictional markers: the mean and standard deviation of the marks each one gave, how far each mean lies from the mean of all scripts, and how big that gap is measured in team standard deviations.
• Compare a moderator's re-marking of eight sample scripts with the original marks, fit a straight line with SLOPE and FORECAST.LINEAR, and use it to adjust the marker's other scripts within the allowed range.
• Build the cumulative distribution of 24 fictional final marks, set a grade boundary from a target share with PERCENTILE.INC and check what share of candidates actually reaches it – ties and whole marks make the two differ.
• Grade 24 fictional candidates with an approximate-match XLOOKUP, then move the B boundary by a couple of marks and count how many candidates change grade and how the share at B or above responds.
• Scale three components with different maximum marks to their fictional weights, add them to an overall mark out of 100, grade it from a boundary table and see why simply adding raw marks gives a different answer.
• Apply re-mark results to a cohort of 20 fictional candidates, measure how often a re-mark changed the mark, count the final grades per band and chart the grade distribution for the results report.