How to use assessment data to drive instruction

A score is the beginning, not the answer

A class average or a student's percentage tells you that something needs attention, not what. Data drives instruction only when it changes what happens in the next lesson: who you work with, on which skill, and how. If nothing in Tuesday's lesson changes because of Monday's check, the check was record-keeping.

The cycle, week by week

Keep the checks short and low-stakes

You do not need more tests. A short check every two or three lessons is enough, and it helps learning on its own: in a well-known experiment, students who practised recalling a text remembered it better a week later than students who reread it (Roediger & Karpicke). Rosenshine's principles of instruction put the same idea first: begin lessons with a short review of previous learning.

Keep the checks ungraded or lightly graded so students treat them as practice. A check that feels like a test produces careful guessing, not honest answers.

Read the results by skill

Tag each question with the one skill it tests and look at results per skill. Start with the questions the class lost most and the wrong answers most students chose. Then look per student: which skills each one keeps getting wrong on fresh questions. Two readings, ten minutes: one for the class, one for individuals.

Decide: whole class, small group or one student

If most of the class shares one wrong answer, reteach that idea to everyone, using a different representation from the first time. If a skill is wrong for a handful of students, pull them into a short group while the rest practise. If one student is wrong across several skills, look for the earliest skill they miss; that student needs a conversation, not another worksheet.

A worked week

Monday: a Year 9 science class takes a ten-question check on forces, one skill per question. The average is 64%, which on its own says little.

Monday after school, ten minutes: two questions lost most of the marks. On one, 14 of 26 students said a ball thrown upwards has an upward force on it at the top of its path. That is one shared misconception, so it is a whole-class reteach. On the other, wrong answers were scattered; five students missed it and also missed a similar question, so they are a small group.

Tuesday: the lesson opens with ten minutes on forces at the top of a throw, taught with a free-body diagram instead of the description used the first time. The five students work with the teacher on reading the graph while the others practise.

The following Monday: two new questions on each skill go into the next check. 22 of 26 now get the first one right; the small group gets the second one right. Both skills close. The numbers in this example are illustrative.

What a head of department should look at

Department leaders rarely need individual answers. What helps is the same view for every class: which skills each class has and has not mastered, out of how many students, and whether the gaps closed after reteaching. When the same skill is weak in every class, it is a curriculum or sequencing question, not a teacher question.

Common mistakes

• Collecting data more often than anyone reads it.
• Reading only averages, so every class looks roughly the same.
• Reteaching with the same explanation that did not work the first time.
• Never checking whether the reteach worked.