Personalised Learning: From Big Idea to Classroom Practice

Personalised learning is a term I suspect may eventually fade. Not because the idea behind it is wrong, but because it’s come to mean too many different things at once. It sits comfortably in policy documents, accreditation frameworks, and school strategy statements, but that comfort comes at a cost. When a term can mean almost anything, it becomes hard to say what’s actually happening in classrooms, or whether it’s working.

Here’s the argument I want to make: personalised learning isn’t one thing. It’s at least three. And they’re not equally effective. Understanding the difference matters, because schools have limited time and limited resources, and those should go where the evidence is strongest.

A Term That Outgrew Its Meaning

Personalised learning emerged from a simple and widely shared concern: learners are different, and one-size-fits-all provision doesn’t serve everyone well. The original intention wasn’t to prescribe a method, but to signal a shift away from uniformity and towards greater responsiveness to individual learners. In that sense, it was always an aspirational idea rather than a tightly defined pedagogical model.

Over time, the label stretched. As schools, systems, and technologies evolved, personalised learning came to include:

• adaptive software and digital platforms
• flexible pacing and personalised pathways
• independent or self-directed study time
• choice-based tasks and projects
• goal setting, reflection, and learner dashboards
• targeted support and small-group teaching

This expansion wasn’t accidental. A flexible term is useful in policy and accreditation contexts, where it can signal a commitment to recognising learner difference without specifying how. But that same flexibility became a problem. When personalised learning covers this much ground, it’s hard to evaluate, hard to improve, and easy to mistake visibility for impact.

All of these things are, in some sense, genuine attempts to respond to learner difference. The question isn’t whether they count as personalisation. They do. The question is what each of them actually does for learning.

Three Levels, Not One

A more honest way to think about personalised learning is as three distinct levels, each operating at a different distance from learning itself, and each with a different evidence base.

Level one: personalisation of the learning experience

This is about the structure and organisation of schooling. Subject choices, course pathways, timetabling flexibility, learner dashboards, and choice boards all sit here. Students have different routes through a curriculum and more agency over what their educational journey looks like.

This matters. A student who has genuine choices about what to study, and real agency over how they progress, is in a different relationship with their education than one who has none. Pathway personalisation can increase engagement, motivation, and a sense of ownership. These aren’t trivial outcomes. Emotional engagement with learning matters, and in John Hattie’s synthesis of over 1,800 meta-analyses, student self-reported expectations and goal-setting carry meaningful effect sizes.

But this level operates at a distance from learning itself. It shapes the conditions of schooling, not the act of teaching. A student can have a highly personalised pathway through a school and still experience very little personalised teaching within it.

Level two: personalised delivery and content

This is closer to learning. Adaptive technology, pacing models, and content variation all sit here. Platforms adjust the difficulty or sequence of content based on learner responses. Some of this is genuinely useful. The Education Endowment Foundation rates digital technology at +4 months of additional progress, but with an important caveat: it’s only effective when it makes a specific learning mechanism more accessible. Platforms that adjust content based on click-through rates or confidence scores are personalising data, not learning.

The evidence on adaptive technology is promising but conditional. It tends to work well for procedural knowledge in clearly defined domains. It’s considerably weaker on conceptual understanding, transfer, and the kind of learning that requires someone to notice something unexpected and respond to it.

It’s also worth naming what doesn’t belong here, even though it’s often described as personalisation. Differentiated tasks, three versions of the worksheet, activities tiered by perceived ability, were once widely practised as a form of personalisation. The problem is that they cap expectations. When the task has already decided what a learner can do, the ceiling is built in. This is a good example of what Ed Haertel called a lethal mutation, a practice that began with sound intentions but was implemented in a way that quietly undermined what it was trying to achieve. In this case, the intention was to support all learners. The effect was often to sort them into different destinations.

Scaffolding is different, and the distinction matters. The goal is the same for everyone. The support varies, and is deliberately temporary. You’re not lowering the ceiling. You’re adjusting the ladder.

Level three: personalised learning through instructional responsiveness

This is where the strongest evidence lives, and it’s the level most directly proximal to learning itself.

Personalised learning in this sense happens when teaching responds to what learners actually understand, not to a profile, a predicted level, or a prior assessment. It depends on teachers eliciting evidence of understanding in the moment and adjusting instruction in response. The goals remain high and shared. What changes is how each learner is helped to get there.

Formative assessment and feedback produce some of the largest learning gains of any educational intervention. Black and Wiliam’s landmark 1998 meta-analysis found effect sizes of between 0.4 and 0.7 for classrooms where formative assessment was consistently used, larger, they noted, than most other interventions studied. The Education Endowment Foundation rates feedback at +6 months of additional progress with a very high evidence base. But this isn’t feedback as a grade or a general comment. It’s feedback that’s timely, specific, and focused on what the learner should do next, feedback that requires action, not just reading.

Metacognition and self-regulation sit at the top of the EEF Teaching and Learning Toolkit, rated at +7 months of additional progress. Teaching students to plan, monitor, and evaluate their own learning, to know where they are, where they’re going, and what they need to do next, has consistent effects across ages and subjects. This isn’t the same as assuming students can self-direct. It requires explicit, deliberate teaching.

Responsive instruction has equally strong support. Rosenshine’s Principles of Instruction, which synthesise findings from cognitive science and decades of classroom research, point consistently to the same habits: reviewing prior learning, presenting new material in manageable steps, checking for understanding frequently, and providing guided practice before releasing students to work independently.

Adaptive scaffolding, the deliberate provision and gradual fading of support as understanding develops, keeps expectations high for all learners whilst ensuring that gaps in prior knowledge don’t become permanent barriers.

What all of these have in common is that they require teaching to change in response to evidence of what learners actually understand. That’s the mechanism through which learning moves forward.

Assessment for Learning and Adaptive Teaching: the Engine

The practices above aren’t separate strategies. They’re expressions of the same underlying process: Assessment for Learning followed by adaptive instruction.

Assessment for Learning makes learner understanding visible. Adaptive teaching is the process through which that understanding shapes instruction. When these work together, teaching becomes genuinely responsive to what learners know, misunderstand, and are ready to learn next.

This cycle, elicit, interpret, adjust, verify, is well established in the research. It’s worth noting that it’s been repackaged in various forms over the years. One recent and useful framing comes from René Kneyber and Valentina Devid at Toets Revolutie, a Dutch organisation whose work on formative action places the emphasis firmly on what teachers and students do with evidence of learning, not just the gathering of it. Their five-step model (orientate, generate, evaluate, act, verify) is described in their 2023 book Formative Action: From Instrument to Design and has been written about extensively by Tom Sherrington at WalkThrus. The underlying idea, that assessment is only formative when it leads to instructional action, sits squarely in the tradition that Black and Wiliam established in 1998.

Personalised learning, at its most effective, isn’t a strategy layered on top of teaching. It emerges from the quality of the teaching itself.

What This Looks Like in Practice

The strongest examples of personalised learning tend not to look flashy. They look like disciplined, responsive teaching within a thoughtfully designed learning environment.

In practice, this means classrooms where:

• learning goals are clear, shared, and genuinely ambitious for all learners
• teachers use evidence of understanding to make decisions about what to do next, mid-lesson, between lessons, and across a unit
• scaffolds are provided deliberately and faded as understanding grows, rather than being fixed features of particular tasks or particular students
• learner agency is taught and guided, not assumed, students learn how to set goals, interpret feedback, and make informed decisions about their learning
• flexible grouping is used responsively, based on learning need, not fixed by perceived ability

The evidence does not support setting or tracking as a general strategy to improve outcomes. Where grouping is beneficial, it is typically short-term, responsive to specific learning needs, and accompanied by changes in instruction, particularly for learners who require additional support.

In this form, personalisation isn’t about doing different things for the sake of difference. It’s about ensuring that different learners get what they need to make meaningful progress, without sacrificing coherence or challenge.

Why the Distinction Matters

None of this means that levels one and two are worthless. Pathway personalisation can create conditions for engagement and ownership that matter for learners’ long-term relationship with education. Well-designed adaptive technology can provide targeted practice in ways that free up teacher time for more complex instructional work. These things have a place.

But the problem is that schools conflate all three levels, invest in the visible ones, and then wonder why learning isn’t moving. Or worse, they assume it is, because the structures look like personalisation. A school can have elaborate learner dashboards, extensive course choice, and sophisticated adaptive platforms, and still not be personalising learning in the sense that consistently moves learners forward.

The question worth asking, in any school claiming to personalise learning, is a straightforward one:

Where, specifically, is teaching changing in response to evidence of what learners understand?

If that question is easy to answer, if there are clear examples grounded in formative practice, then personalised learning is probably happening at the level that matters most. If the answer points mainly to structures, platforms, pathways, or student choice, then what’s in place may be valuable, but it’s operating at a different level.

Grouping, pathways, or pacing alone don’t constitute personalised learning unless they lead to different teaching. These are the personalisation of the learning experience. Real and worth having. But not the same thing.

Clarity here isn’t about winning arguments. It’s about aligning where schools put their energy with what the evidence suggests actually helps learners.

Ask yourself: if you removed the platforms, the dashboards, and the pathways, would learners still experience teaching that responds to what they understand? If yes, you have personalised learning. If no, you have the conditions for it, and that’s a different thing.

Personalisation that’s visible in a strategy document but invisible in a lesson isn’t personalised learning. It’s personalised administration.

A Note on Techniques

The following instructional techniques consistently appear in high-quality research on what moves learning forward. They’re the practical tools of level three personalisation, as distinct from the structural tools of levels one and two.

Diagnostic formative assessment. 

Using questioning, hinge questions, work scrutiny, and discussion to identify misconceptions and partial understanding, not just overall performance. Black and Wiliam are unambiguous: when teachers use assessment to support learning rather than simply measure it, learning gains are substantial and consistent.

Targeted, actionable feedback. 

Timely, specific, and focused on what the learner should do next. The EEF rates this at +6 months of additional progress. Generic comments and scores don’t carry this effect. What matters is that feedback requires the learner to act: to redraft, revisit, retrieve, or reconsider.

Explicitly taught metacognition. 

Teaching students how to plan, monitor, and evaluate their own learning. Assessment-capable learners, those who can say where they are, where they’re going, and what they need to do next, show some of the highest effect sizes in Hattie’s research. But this capability is taught, not assumed.

Adaptive scaffolding. 

Adjusting models, prompts, representations, and supports in response to evidence of understanding, and deliberately fading them as competence grows. High expectations remain; the route to meeting them is adjusted.

Flexible grouping. 

Regrouping students temporarily for targeted instruction, re-teaching, or extension, based on current learning need rather than fixed ability. The evidence does not support setting or tracking as a general strategy to improve outcomes. Where grouping is beneficial, it is typically short-term, responsive to specific learning needs, and accompanied by changes in instruction, particularly for learners who require additional support.

Multiple representations. Presenting ideas in different ways to support access and deepen understanding. This isn’t about matching instruction to supposed learning styles, for which there’s no good evidence. It’s about using varied representations, visual, verbal, concrete, to make ideas more accessible and connections more visible.

Strategic use of data and technology. 

Using platforms, analytics, and adaptive tools to inform instructional decisions, not replace them. Dashboards and self-reported confidence scores can support awareness. On their own, they personalise data rather than learning. The EEF rates digital technology at +4 months of additional progress, but only when it makes a specific learning mechanism more accessible, not simply when it makes learner data more visible.