Randomised controlled trials (RCTs) remain the most reliable way of identifying whether an educational approach causes improved outcomes. But they are necessarily slow.
Designing and refining an intervention, recruiting schools, delivering training, and measuring attainment before and after the intervention. It all typically takes a few years.
For school leaders and teachers, however, decisions cannot wait that long. Teachers must constantly decide between plausible options: various feedback routines, different homework structures, how to design a retrieval practice strategy, or ways of explaining new vocabulary effectively.
How can we bridge this gap? A micro‑randomised controlled trial (micro‑RCT) is a small, rapid, and cheap classroom‑level trial designed to test these sorts of everyday choices.
How micro-RCTs work
Randomisation can occur at two levels.
Entire classes may be assigned to an intervention and compared with similar classes continuing with business-as-usual practice. Or, pupils within the same class may be randomly allocated to small groups that either receive the intervention or continue with typical teaching.
Outcomes are measured over a short period, usually several weeks or half a term. The result is a quick evaluation of that practice in that context.
Micro-RCTs are particularly suited to high‑frequency decisions (e.g. feedback routines, retrieval activities, homework design, fluency practice or scaffolding strategies) where teachers repeatedly choose between alternative approaches.
If a micro-RCT is then replicated ten or more times they can also provide an idea of the spread of effects across contexts, as well as an overall average with a high degree of certainty about the result.
Replication also adds almost nothing to the cost. Once a trial is set up on the platform, it can be used by as many teachers as can be recruited.
What challenges might this overcome?
Robust educational evaluation faces two central challenges.
The first is interpreting the results from an inconclusive trial. Findings from a large-scale trial tell us only whether the intervention worked under the specific conditions in which it was tested. It cannot tell you why it did (or did not) work. Micro-RCTs offer a way of testing the components of an intervention, or optimising aspects of it (such as the frequency or intensity of use or evaluating the separate components of a more complex intervention).
Second, there is a translation gap, even when something is shown to be successful. In theory, large trials provide generalisable evidence. But too often they feel distant from everyday classroom decisions. Therefore, schools rely heavily on professional judgement and local experience to decide whether the findings are applicable (or not) in their context. Micro‑RCTs offer a bridge between these worlds by allowing teachers to test ideas in their own context while preserving some causal inference.
So micro-RCTs might help overcome:
- inconclusive findings from expensive large scale RCTs
- long delays between innovation and feedback
- uncertainty about implementation in specific contexts
- reliance on anecdotal impressions of impact
A method of enquiry
However, important caveats remain. Small samples produce imprecise estimates. Individual results are uncertain and easily over‑interpreted. Classroom interaction can contaminate groups. Short-term, proximal assessments may have limited reliability.
Micro‑RCTs therefore function best not as definitive evaluations but as a method of structured enquiry. As a method, they are most useful when multiple replications are aggregated or combined to provide an idea of the spread of results.
Feasibility pilot
To explore whether micro‑RCTs can operate in real school conditions, we developed a feasibility pilot using the What Worked Teachers platform. The intention was not to evaluate a specific teaching intervention, but to evaluate the evaluation process itself.
In effect, this is a proof‑of‑concept: can teachers realistically run fair tests within ordinary teaching?
This mirrors early stages of large education trials, where feasibility studies precede efficacy trials. Here, however, the innovation is methodological. The intervention is not a pedagogical approach, but a research infrastructure – a system enabling teachers to generate causal evidence locally.
Initial findings from our pilot
It’s too early to say anything definitive. But we can make some observations from our early experience.
One is that the idea is intuitive to teachers: many already experiment informally. The idea of adhering to randomisation to maintain a ‘fair test’ is widely understood.
However, we’ve also seen some barriers emerge:
- Time. Teachers must plan lessons, manage behaviour, and cover curriculum content; any additional task competes with these priorities. Simplicity therefore becomes critical. The process must feel lighter than a typical research initiative.
- Confidence with data. Teachers are comfortable interpreting pupil assessments, but may be less confident interpreting variation and uncertainty. Automated reporting helps, but interpretation guidance is likely to matter as much as the analysis itself.
- Culture. School improvement often relies on collective adoption of practices. Micro‑RCTs introduce a temporary divergence with different pupils experiencing different approaches. While this is educationally reasonable, it requires careful explanation to staff and leaders to avoid misunderstanding.
These reflections suggest that feasibility depends less on statistical complexity and more on implementation clarity. The limiting factor may not be randomisation but practicality.
What micro-RCTs could be used for
The promise of micro-RCTs is modest but important: it is not faster answers to big questions, but rather quicker and cheaper learning about small ones.
These are exactly the questions teachers that face everyday and the challenge that academics have about turning theory into effective practice.
Possible applications include:
- testing alternative feedback strategies
- evaluating retrieval practice routines
- comparing homework formats
- refining scaffolding and modelling approaches
- piloting new materials before wider adoption
- testing different components when refining an intervention
Perhaps most importantly, they may shift how evidence is used. Instead of evidence being delivered to schools, schools participate in producing it. Teachers move from consumers of education research to contributors to it. That’s something worth aiming for.