Useful assessment analysis begins with a teaching question, not a chart. Identify what students understood, where a misconception affected the class and which next action will make a difference. Question-level marks can guide that enquiry, but the student responses explain it.
The aim is not to produce a larger report. It is to leave the analysis with a small number of defensible decisions.
Check the assessment record first
Before interpreting results, confirm that the data represents the marking accurately.
- Every expected student is included or marked absent.
- Every scored question has a result or a clear unattempted state.
- Question maximums and the assessment total agree.
- Manual rubric overrides have been reviewed.
- No script has missing or duplicated pages.
- Grade boundaries are the intended version.
- Moderation changes are reflected in the final scores.
An incomplete column can look like a learning gap. A data-quality check prevents the analysis from giving a confident answer to the wrong question.
Begin with the class overview
Use the total score distribution to understand the shape of the assessment, not to diagnose individual topics.
Look at:
- median and range;
- clusters or large gaps;
- unexpected outliers;
- completion rate;
- the relationship between total marks and grade boundaries;
- whether the result differs sharply from recent comparable assessments.
Avoid reading too much into the class average alone. Two classes can share the same mean while one has a tight middle and the other has a split between very high and very low scores.
Move quickly to question-level results
For each question, calculate or view:
- average marks as a proportion of the maximum;
- number of full-mark responses;
- number of zero or unattempted responses;
- spread of scores;
- performance by subquestion or rubric criterion where available.
Normalising by maximum mark helps compare a two-mark item with a twelve-mark item. The purpose is to locate questions worth inspecting, not to rank every item from best to worst.
Question-by-question marking makes this data a natural by-product of the marking process rather than a separate spreadsheet task.
Inspect the work behind the number
A low-scoring question can have several explanations:
- the concept was not secure;
- students misunderstood the command word;
- a prerequisite skill failed;
- time pressure led to incomplete answers;
- the question or diagram was unclear;
- the mark scheme omitted a credible approach;
- a scanning or page-order problem hid evidence;
- marking was inconsistent.
Read a purposeful sample: one strong response, several common middle responses, a zero and any unattempted scripts. Look for a shared pattern in the actual work.
Do the same for unexpectedly high performance. A good result might reflect secure understanding, but it could also mean the item did not discriminate or closely repeated a rehearsed example.
Distinguish class needs from individual needs
Use three levels of response.
Whole-class action
Choose this when a large proportion of students share the same misconception or method gap. Reteach the concept with a different representation, model a stronger response or run a focused correction task.
Small-group action
Choose this when a cluster of students has a specific prerequisite gap. A short targeted session is more efficient than repeating the full lesson for everyone.
Individual action
Use this for an unusual misconception, missing work, an access need or a pattern that appears across several assessments for one student.
The score helps identify the group. The response determines the action.
Use feedback patterns as qualitative data
If the marking system records reusable feedback or tags, count the recurring themes carefully. A frequently used comment about unsupported evidence may be more instructive than the average score for the essay.
Do not treat every comment as a perfectly standardised category. Teachers edit feedback for context, and several phrases may describe the same underlying need. Combine related themes, then verify them against sample scripts.
The feedback comment bank guide explains how to organise reusable phrases around questions and criteria so these patterns remain meaningful.
Write a one-page assessment response
Keep the final analysis short enough to use. A practical record might contain:
- What was secure: one or two strengths supported by question data and scripts.
- Priority misconception: the most important shared gap.
- Students needing targeted support: a named secure list stored in the appropriate school system.
- Next teaching action: what will happen, when and with whom.
- Assessment issue: any item, mark scheme or administration change for next time.
- Follow-up evidence: the short task that will show whether learning improved.
This is more valuable than a decorative dashboard with no decision attached.
Compare over time with care
Trend analysis is useful only when assessments are reasonably comparable. A lower average may reflect a harder paper, different content coverage or a changed cohort rather than weaker teaching or learning.
Where possible, compare:
- the same skill across several assessments;
- performance as a percentage of available marks;
- question design and demand;
- the conditions under which the assessment was completed;
- students present in both datasets.
Use trends to form a question for further investigation, not as a standalone judgement about a teacher, class or student.
Export only what another system needs
If totals must move to a management information system or department spreadsheet, export them once from the final marking record. Preserve stable student identifiers, understandable column headings and the assessment version.
Avoid maintaining a parallel spreadsheet throughout marking unless it has a clear purpose that the marking workspace cannot meet. Duplicate records drift and create another checking task.
Assessment insight in Nib
Nib records results as teachers mark, connecting each score and feedback decision to the student, question and original script. That makes question-level analysis a by-product rather than a second data-entry exercise. The product is designed to reveal useful patterns while keeping the evidence one tap away.
For the workflow that creates this record, begin with the digital marking guide.