Using Assessment Data Without Being Driven by It

Data-driven teaching is, in principle, a good idea. Decisions based on evidence about what children know and don't know are better than decisions based on gut feeling. The problem is not data; it is how data is used.

The pathologies of data use in primary schools are well-documented: teachers who spend more time recording children's progress than addressing it; data meetings that focus on accountability rather than action; assessment systems that measure what is easy to measure rather than what matters.

What makes assessment data useful

Data is useful when it tells you something specific that you can act on. 'Class 4's reading comprehension is below target' is not actionable — it is a general statement about a general problem. 'Seven children in Class 4 cannot reliably infer character motivation from text — they are treating the text too literally' is actionable. That tells you what to teach next.

The distance between data collection and data action matters. Assessment data collected in October that is reviewed in February cannot change October's teaching. The formative cycle needs to be short: assess on Monday, adjust on Tuesday.

Red herrings in data reading

Average class scores obscure the distribution. A class with 28 children at expected and 2 far below expected has the same average as a class where most children are slightly below. They need completely different responses.

Single data points are rarely reliable. A child who scores unexpectedly low on one test may have been tired, unwell, or had a bad morning. A pattern across multiple data points is meaningful; a single point rarely is.

What to do in data meetings

The purpose of a data meeting is to decide what to teach, not to explain why things are the way they are. Any meeting that spends more than 20% of its time on explanation and less than 80% on action planning is not working.

A useful structure: What does the data tell us? (5 minutes) → Which children need what? (10 minutes) → What specifically will we do, when, and who will check? (10 minutes). The third question is the only one that matters.