Large, consistent gains
A meta-analysis of 96 randomized trials found that tutoring raised achievement by an average of 0.37 standard deviations, a large effect by the standards of education research.1
Evidence & approach
We follow the research, and we measure the results.
Our program combines two strong bodies of evidence: what makes tutoring effective, and how people learn and remember. This page sets out both, along with exactly what we measure and how we report it.
What the tutoring research says
Few interventions have stronger evidence, if the design is right.
Design determines impact
Effects are largest when tutoring is frequent (three or more sessions a week), delivered in small groups by a consistent, trained tutor, during the school day, and connected to classroom content and data.2 4
Proven at scale here
In the NYC Public Schools HIT initiative, middle school math students averaged 1.5 years of progress in one year on i-Ready. The share of K–2 students at or above literacy benchmark rose an average of 13 points a year on Acadience.3
Those results are the initiative’s, not ours. We are a new provider with no outcome data of our own yet. We cite them because they define the standard we intend to meet, and because our model follows the same design principles.
What the learning science says
Mental Arts teaches the methods with the strongest evidence.
Decades of cognitive science agree on which study methods work, and on which ones students actually use. These are often not the same. Mental Arts closes that gap.
Retrieval practice
Recall beats rereading
Practicing recall produces far better long-term retention than restudying the same material for the same time.5
Spacing
Spread it out
Review spaced over days and weeks is retained much better than the same amount of review crammed together.6
Metacognition
Familiar isn’t learned
Learners often mistake familiarity for mastery, and prefer methods that feel easy but work poorly. Checking yourself corrects this.8
A major review rated practice testing and distributed practice “high utility,” and rereading and highlighting “low utility.”7
Data use & progress monitoring
Data at every stage of the cycle.
We use data for three things: forming the right groups, planning each session, and showing your team whether tutoring is working. We agree on goals with your HIT team before the cycle starts.
Before
Your benchmark data and our short diagnostic set each student’s starting point and target skills, and determine the groups.
Every session
Attendance, minutes, and an exit ticket. The tutor plans the next session from that day’s results.
Every two weeks
A progress report to your HIT team. Groups and goals are adjusted if a student is moving faster or slower than expected.
End of cycle
A post-assessment and a summary comparing each student’s baseline and final results, with a recommendation for next steps.
| What we measure | How | When it’s reported |
|---|---|---|
| Dosage delivered | Minutes each student actually received, counted per session | Every session; reported every two weeks |
| Attendance | Taken live by the tutor; absences sent to the HIT coordinator | Same day |
| Skill mastery | Unaided skill checks; mastered at 90% or better | Every session; cumulative every two weeks |
| Growth | Baseline vs. end-of-cycle on our measures; school benchmark (e.g., i-Ready) where available | End of cycle |
| Fidelity | Observations against our session model and the HIT standards | First two weeks, then at least monthly |
| Experience | Short student and teacher surveys | Mid-cycle and end of cycle |
Our internal standard
Mastery, not just exposure.
Most tutoring measures what students were exposed to. We measure what they can now do on their own. A targeted skill counts as mastered only when a student scores 90% or better on an unaided skill check, and still can when it returns in spaced review.
This is the A-level standard we hold ourselves to. It tells our tutors when to move on and when to keep going, and it is how we judge our own work. It is never used to label, sort, or grade students. Their teachers and schools do that.
We will share results with each partner school in full, and publish aggregate, de-identified results on this site, including results that fall short.
Sources
References
- Nickow, A., Oreopoulos, P., & Quan, V. (2020). The impressive effects of tutoring on PreK–12 learning: A systematic review and meta-analysis of the experimental evidence. NBER Working Paper 27476.
- Robinson, C. D., Kraft, M. A., Loeb, S., & Schueler, B. E. (2021). Design principles for accelerating student learning with high-impact tutoring. EdResearch for Recovery, Annenberg Institute at Brown University.
- ExpandED Schools. (2026). High-impact tutoring in NYC: Lessons from a citywide initiative. expandedschools.org
- National Student Support Accelerator, Stanford University. Toolkit for Tutoring Programs and tutoring quality standards. nssa.stanford.edu
- Roediger, H. L., & Karpicke, J. D. (2006). Test-enhanced learning: Taking memory tests improves long-term retention. Psychological Science, 17(3), 249–255.
- Cepeda, N. J., Pashler, H., Vul, E., Wixted, J. T., & Rohrer, D. (2006). Distributed practice in verbal recall tasks: A review and quantitative synthesis. Psychological Bulletin, 132(3), 354–380.
- Dunlosky, J., Rawson, K. A., Marsh, E. J., Nathan, M. J., & Willingham, D. T. (2013). Improving students’ learning with effective learning techniques. Psychological Science in the Public Interest, 14(1), 4–58.
- Bjork, R. A., Dunlosky, J., & Kornell, N. (2013). Self-regulated learning: Beliefs, techniques, and illusions. Annual Review of Psychology, 64, 417–444.
- Ericsson, K. A., Krampe, R. T., & Tesch-Römer, C. (1993). The role of deliberate practice in the acquisition of expert performance. Psychological Review, 100(3), 363–406.
- Rohrer, D., & Taylor, K. (2007). The shuffling of mathematics problems improves learning. Instructional Science, 35, 481–498.