📋Rubrics Teachers Can Trust

Ask most teachers how they feel about being observed and evaluated, and you'll get a careful answer. Underneath it, often, is a mix of anxiety and mild cynicism. Someone walks in, sits at the back, writes things down, and some days later a score appears. The teacher rarely knows exactly what was being looked for, or what to do differently next time.

I understand that reaction. And having spent a lot of time thinking about how we evaluate educators across 120+ schools, I've come to believe that most of the problem lies in the design. If evaluation feels like surveillance, it's usually because it was built like surveillance.

The alternative is evaluation that feels like coaching. That's harder to build, but not mysterious. It comes down to a handful of principles about the rubric itself and how its results are used.

Criteria you can see

The first test of a rubric is whether a teacher could read it and know what good looks like.

Criteria must be observable. "Creates an engaging learning environment" is a feeling. "Checks for understanding at least once before moving to independent work" is something an observer can actually see happen, or not happen. The more a criterion depends on the observer's mood, the less a teacher can trust the score.

And the rubric must be published. Teachers should have it in their hands long before anyone walks into their classroom. There is no good reason for evaluation criteria to be secret. If we want teachers to get better at something, the least we can do is tell them what it is.

Weight what matters

Not every criterion is equally important. A rubric that treats "board work is neat and legible" the same as "students leave the lesson able to do something they couldn't do before" is quietly telling teachers that both matter equally. They don't.

So the criteria should be weighted, so that what matters most counts most. This sounds obvious, but in my experience it's often skipped, because weighting forces uncomfortable conversations about priorities. Those conversations are exactly the point. A weighted rubric is a statement of what the organisation actually values in teaching.

Experts own it, and it evolves

A rubric for teaching mathematics should be shaped by people who deeply understand how mathematics is learned. The same goes for languages, sciences and everything else.

I've become firm that subject-matter experts should own the rubric and refine it over time. Not administrators, not a central office working from a generic template. And because teaching evolves, and because we learn from using any tool, the rubric should be versioned. When it changes, everyone should know what changed and why, and old evaluations should still be read against the version that applied at the time.

A rubric that never changes is one nobody is learning from.

A shared core, plus depth

Across a network of schools, there's a real tension here. You want to compare. You want to know whether teaching quality is improving across the network, which schools need support, where a particular strength is concentrated. But you also know that a good English lesson and a good physics lesson look quite different.

The answer we've found useful is a common core of criteria that applies across every subject, things like lesson planning, checking for understanding, classroom culture, and use of time. That core gives you comparability. Then, on top of it, subject-specific criteria that capture what's distinctive about teaching that discipline well. That gives you depth.

Without the core, you can't see the network. Without the subject layer, you can't see the teaching.

Every score comes with evidence

This is the principle I'd defend hardest.

A score on its own is close to useless to a teacher. A "3" on questioning tells them very little. What helps is the evidence behind it: what was actually said, what was observed, what the students did. "You asked eleven questions in the first twenty minutes; nine were answered by the same three students" is something a teacher can work with.

And every evaluation should end with two to four concrete action points. Not twelve. Not a vague "continue to develop questioning strategies". Two to four specific things to try before the next observation. That's what turns a judgement into a coaching conversation. The teacher walks away knowing what to do on Monday.

Name the bands for growth

Language matters more than we sometimes admit. If the result of an evaluation is "pass" or "fail", or "meets" and "does not meet", the teacher hears a verdict.

We use bands named for growth: Strong, Good, Developing, Needs Support. Each of those says something about where a teacher is and, implicitly, where they're heading. "Developing" tells a teacher they're on the way. "Needs Support" puts some of the responsibility on the organisation, which is where it belongs. It's a small choice of words, but I've seen how differently people respond to it.

The teacher sees it first

The person being evaluated should see their own insights before anyone else does. Before the principal reviews a summary, before anything rolls up into a network report, the teacher should have the chance to read their evaluation, sit with it, and understand it.

This does something important to trust. It signals that the evaluation exists primarily for the teacher's growth, and only secondarily for anyone's dashboard. When teachers learn about their own results from a meeting where everyone else already knows, the whole exercise starts to feel like it was about them rather than for them.

Internal, always

Teacher evaluation data is internal. It is never shown to parents.

I hold this line firmly. Teachers need to be able to be "Developing" on something without that becoming a reputation. If evaluation data could end up in front of parents, every teacher would rationally optimise for looking good rather than getting better, and the honesty that makes coaching possible would disappear. Parents deserve good teachers. The way to get them is to protect the space in which teachers can improve.

Why structure beats free text

There's a natural instinct, when you want evaluation to feel humane, to let observers write freely. Just describe what you saw. Give real feedback.

Free-text feedback has its place, and good observers write useful notes. But it can't be aggregated. You can't add up a hundred differently-phrased paragraphs and learn where questioning is weak across a region, or whether a training programme moved anything. Across a network, structure is what makes feedback useful beyond the individual conversation.

So the design I believe in combines both. Structured, weighted, observable criteria that can be compared and tracked. Evidence and action points that keep it human and specific. A teacher should feel the evaluation was about their actual classroom. The network should be able to learn from thousands of them together.

What trust looks like

I don't think teachers will ever look forward to being observed, exactly. But I do think they can come to trust it. That trust comes from knowing the criteria in advance, knowing that experts in their subject shaped them, getting evidence instead of impressions, leaving with a few clear things to try, seeing their results first, and knowing those results won't be used against them.

When those things are in place, an evaluation stops being something done to a teacher. It becomes one of the more useful conversations they have all term.