Spaced Learning Practice
Evidence supports pilots, not universal promises
A 2025 classroom meta-analysis found a moderate advantage for distributed over massed practice, d = 0.54, with substantial heterogeneity remaining. Sell evidence-informed design and experimentation — not a universal memory formula.
Cases we cite
| Evidence case | Design and finding | What it reveals | Business response |
|---|---|---|---|
| Classroom meta-analysis | Pooled advantage d = 0.54 (22 reports, 31 effect sizes, >3,000 learners) | Average benefit does not determine a particular client’s outcome | Justify a pilot, not a guaranteed uplift |
| EEF SMART Spaces | 125-school evaluation; chemistry effect −0.01 — no additional months of progress | A research principle can fail when packaged as a particular programme | Test the whole intervention, not merely the theory |
| Primary-school classroom study | Retrieval beat rereading; short vs long spacing intervals did not differ significantly (mean accuracy 44.6% vs 45.9%) | Retrieval quality may matter more than a small cadence change | Invest in authentic questions and feedback before algorithmic optimisation |
| Microsurgery RCT | 42 medical students; one 8-hour session vs four weekly 2-hour sessions. At one month: spaced 31.60 vs massed 27.63, p = 0.04 | Equal nominal hours can produce different delayed performance | Hold content and time broadly constant when testing cadence |
SMART Spaces trained teachers to deliver six structured GCSE Chemistry revision lessons. High teacher engagement; 70% said they would deliver it again — yet no evidence of improved GCSE Chemistry outcomes. Acceptability is not attainment. Separate four questions: Did people participate? Did they retain? Did they apply? Did the business outcome change?
- Appropriately spaced opportunities to recall and apply can improve later performance in many settings — on average, with heterogeneity
- Delayed follow-up is a stronger transfer signal than post-course satisfaction
- A paid pilot with a measurement agreement is the defensible commercial entry
- No universal forgetting curve or fixed day sequence for every skill
- No guaranteed ROI or guaranteed uplift for a named client
- No claim that quiz accuracy alone proves compliant workplace behaviour
- No treatment of vendor marketing figures (for example “up to 170% retention”) as peer-reviewed constants