Quillattice · Practical guide · UK education
Prepare education data for a useful analysis.
A focused question and a clear account of the data make it easier to assess what an analysis can show. Use these prompts for assessment, learner or programme data before analysis begins.
Mathematical data analysis serviceOpen the project brief template
01 · Frame the question
Connect the analysis to a decision or comparison.
A useful question sets boundaries before anyone chooses a calculation or chart. It says what you want to explore, who needs the answer and what form of output would help them use it.
Name the decision or comparison
What do you need to decide, or which groups, measures or periods do you want to compare? Keep the question within the data you expect to have.
Identify the audience
Who will use the analysis: a subject lead, programme team, school leader or another group? What context should the reader already know?
Choose a useful output
Would a short written summary, a table, a chart or a presentation make the answer easiest to discuss? Say what the output needs to help the audience see.
A question to adapt
For [cohort or programme], how does [measure] compare across [periods or groups], and what view would help [audience] consider [decision or next discussion]?
The bracketed parts are prompts to fill in from your own context, not claims about a particular school or dataset.
02 · Describe the dataset
Give the numbers their context.
A short description helps someone assess whether the available records fit the question. Include what you know; mark anything that is uncertain or not available.
Source and purpose
Where did the data come from, who collected or prepared it, and what was it originally intended to record? Note whether you are working from a system export, an assessment file or a prepared summary.
Measures and definitions
Name each measure and explain its units, scoring rules or categories. If a score, threshold or label has a local meaning, define it rather than assuming every reader knows it.
Time period and comparison
Give the dates or terms covered. If the question compares periods, say how they were chosen and whether the same measure and assessment conditions were used in each.
Cohort and coverage
Describe who or what is represented: for example, a year group, classes, learners taking part in a programme, or all records available. Mention exclusions, changes in membership or uneven coverage.
Data dictionary, if available
Include a data dictionary, codebook or short field list when you have one. Define abbreviations, missing-value codes, category labels and any derived fields that matter to the question.
Known limitations or constraints
Flag missing or duplicated records, small groups, changes in how data was recorded, measures that are not directly comparable, and any limits on what can be shared or interpreted.
03 · Interpret and report carefully
Keep the finding close to what was measured.
A report can make a pattern easier to inspect without turning it into a causal explanation or a claim about an outcome the data did not measure.
Describe what is visible
State the group, measure and period. Give the count as well as a percentage where it helps the reader understand the denominator, and name the comparison being made.
Check the comparison
Look at how the groups, measures and coverage were defined. A difference in a summary can be worth exploring, but it does not by itself explain why the difference appears.
Keep the limit beside the finding
Say which records the result describes, what is missing or uncertain, and what the data cannot establish. Separate the observation from possible explanations and from outcomes not measured.
A careful reporting frame
Separate the observation from what it does not establish.
“In the supplied records, [count or pattern] was visible for [defined group and period]. This describes the records available for this comparison. It does not by itself explain why the pattern appears or show an outcome that was not measured.”
Quillattice · Mathematical data analysis
Exploration is supported by judgement and review.
AI supports exploration and drafting. Mathematical judgement, education context and human review shape how visible patterns are interpreted and what the supplied data can support.