Private Tutoring vs. Tutoring Centers: What the Research Actually Shows

The tutoring industry generates approximately $9 billion annually in the United States. It spans everything from a college student meeting a fifth grader at a library twice a month, to Kumon franchises running children through self-paced math worksheets in strip malls, to one-on-one sessions with credentialed reading specialists, to AI-powered platforms that adapt to individual learners in real time. Parents trying to get help for a child who is struggling — or for a child who wants to accelerate — are navigating an enormous market with very inconsistent evidence behind it.

The research on tutoring has a finding that most of the industry doesn’t advertise, because it’s inconvenient for the business models of most tutoring centers: frequency is the dominant variable. Once-a-week tutoring, which is the most common format sold, produces academic effects that are disappointingly small and sometimes indistinguishable from zero. Three or more sessions per week — what researchers call “high-dosage tutoring” — produces effects that are large and consistent across multiple rigorous studies.

This piece is specifically about format comparison: what does the research say about one-on-one private tutoring versus center-based programs like Kumon and Sylvan, versus online platforms, versus high-dosage tutoring programs? The general evidence that tutoring works is addressed in the foundational piece on tutoring efficacy. This is about the format differences that parents actually make decisions between.

Key Takeaways

• High-dosage tutoring (3+ sessions per week) produces effect sizes of 0.2–0.4 standard deviations in math and meaningful gains in reading — the most consistent finding in tutoring research
• Once-weekly tutoring, in any format, produces effects that are much smaller and often not statistically significant
• Kumon’s published research base is weaker than its brand recognition would suggest; most of the supporting evidence is proprietary or lacks comparison groups
• One-on-one tutoring with a qualified educator produces larger effects than small-group tutoring, which in turn produces larger effects than center-based self-paced formats
• Online tutoring platforms show results roughly equivalent to in-person tutoring when frequency and tutor quality are equivalent — format alone is not the differentiating factor

What the Research Actually Shows on Tutoring Formats

Format	Evidence Base	Typical Outcomes	Cost Range	Cost-Effectiveness
One-on-one high-dosage tutoring (3+/week, trained tutor)	Multiple RCTs; strongest evidence base (Nickow et al., 2020; Fryer, 2017)	Effect sizes 0.2–0.4 SD in math; meaningful reading gains; largest measured effects in tutoring research	$50–$150/hour; $150–$400+/week	High per-student outcome; high absolute cost
One-on-one low-frequency tutoring (1/week)	Multiple studies; well-documented	Effects much smaller; often not statistically significant; some studies show near-zero effects (Dietrichson et al., 2017)	$30–$100/hour; $30–$100/week	Low outcome/cost ratio due to minimal academic impact
Small-group tutoring (2–5 students, 3+/week, trained tutor)	RAND research; Chicago, Houston high-dosage programs	Effect sizes 0.1–0.3 SD; smaller than 1-on-1 but meaningful; scalable	$15–$50/student/hour	Better cost-effectiveness than 1-on-1 when high dosage maintained
Center-based (Kumon, Sylvan, Mathnasium)	Mostly proprietary studies; few independent RCTs	Mixed; gains documented in some subject areas; comparison groups often weak or absent	$150–$350/month	Moderate absolute cost; evidence base too thin to calculate reliably
Online tutoring platforms (Tutor.com, Wyzant, Varsity Tutors)	Emerging; some platform-funded research; few independent studies	Roughly equivalent to in-person when frequency and tutor quality equivalent	$20–$80/hour	Variable by platform and tutor; format alone not the issue
AI-powered tutoring platforms (Khanmigo, Synthesis, Carnegie Learning)	Early research; IES pilot studies underway; Carnegie Learning has strongest evidence base	Carnegie Learning MATHia: effect sizes ~0.2 SD in math (RCT, 2019); others insufficient data	$15–$50/month subscription	Potentially high cost-effectiveness if research confirms early results
Peer tutoring (trained student tutors)	Hattie meta-analysis; Cohen et al.; multiple studies	Effect sizes 0.35–0.55 SD; among the most cost-effective interventions (benefits both tutor and tutee)	Near-zero (school-based)	Very high — one of the most cost-effective educational interventions known

The table tells a clear story: the format variables that parents typically deliberate over — in-person vs. online, branded center vs. independent tutor — are less important than two variables that are often missing from the marketing pitch: frequency and educator quality.

The High-Dosage Tutoring Finding in Detail

The most rigorous body of evidence on tutoring format comes from research on high-dosage tutoring programs, which researchers define as three or more sessions per week, typically 45 to 60 minutes per session.

Nickow, Oreopoulos, and Quan’s 2020 meta-analysis in NBER Working Papers, reviewing 96 randomized and quasi-experimental tutoring studies, found that one-on-one tutoring produced average effect sizes of approximately 0.37 standard deviations — a large and educationally meaningful effect. Critically, they found that effect sizes were strongly moderated by dosage. Studies with higher session frequency produced larger effects. Studies with once-weekly sessions produced effects that were substantially smaller.

Roland Fryer’s 2017 study of high-dosage tutoring in Houston public schools (Journal of Political Economy) found math effect sizes of 0.15 to 0.20 standard deviations in a school-based high-dosage program delivered three times per week. This was a large, well-controlled study with over 1,000 students. The effect sizes are smaller than the Nickow meta-analysis average because the Houston program used trained college students (not professional tutors) and was delivered in school during the day — conditions that are more scalable but less intensive than private professional tutoring.

Dietrichson and colleagues’ 2017 meta-analysis (Review of Educational Research) examined interventions for low-achieving students across 101 studies and found that the frequency-outcome relationship was the most robust finding in the dataset. Interventions with fewer than two sessions per week showed substantially smaller effects than interventions with three or more sessions per week, regardless of the quality of the individual sessions.

What this means practically: a parent paying $80 per week for weekly one-hour tutoring sessions is purchasing a service with a weak evidence base for producing meaningful academic improvement. A parent paying $300 per week for three one-hour sessions per week is purchasing a service with a strong evidence base for meaningful gains. The cost difference is real; so is the outcome difference.

The Kumon Question — What the Research Shows

Kumon is the world’s largest after-school tutoring program, with approximately 4 million students globally and more than 1,500 centers in the United States. Its reputation in parent communities is strong. Its research base is considerably weaker than its market presence would suggest.

Kumon’s approach is a self-paced, worksheet-based mathematics and reading program in which children complete daily practice at home and bring worksheets to the center for review twice a week. The model emphasizes repetition, mastery before advancing, and independent work at a pace matched to the child’s current level.

The published independent research on Kumon outcomes is limited. The most frequently cited study supporting Kumon is the program’s own commissioned research, which lacks comparison groups and is not peer-reviewed. Independent studies have produced mixed results.

A 2007 study by Clark and colleagues in Learning and Instruction examined Kumon math outcomes in a UK context and found modest gains in procedural computation, with no significant gains in mathematical reasoning or problem-solving. This finding is consistent with the theoretical critique of Kumon: the repetition-based model builds procedural fluency well but may not transfer to conceptual understanding or novel problem-solving.

A 2014 evaluation by the What Works Clearinghouse (WWC), the U.S. Department of Education’s evidence review body, examined Kumon reading and found insufficient rigorous evidence to support or refute its effectiveness — not a positive finding, but a finding of insufficient data. WWC’s review of Kumon math found similar results: no qualifying studies meeting their evidence standards.

Mathnasium and Sylvan have similarly thin independent research bases. These programs are not well-studied in rigorous comparative trials. They may work well for specific children in specific contexts — they clearly have satisfied customers — but the evidence basis for choosing them over alternatives is not strong.

The honest framing is: center-based tutoring programs offer consistency of location, a structured setting, and accountability (children show up at a fixed time). Whether those structural features translate to academic gains depends primarily on how frequently the child attends and the quality of the instruction — not the brand.

What the Research on AI Tutoring Shows

The research on AI-powered tutoring is newer and more limited than the research on human tutoring, but it is developing quickly. The current research on AI tutoring versus human tutoring covers this comparison in depth. The summary for format comparison purposes: Carnegie Learning’s MATHia platform, the most rigorously studied AI tutoring tool, produced effect sizes of approximately 0.20 standard deviations in math in a 2019 randomized controlled trial by Pane and colleagues at RAND. This effect size is in the range of human high-dosage small-group tutoring.

AI tutoring platforms offer unlimited session frequency at a fixed subscription cost, which makes them potentially highly cost-effective if the effect sizes replicate. The current limitation is that the research is concentrated on math, where AI tutoring models are most developed, and does not yet support strong conclusions about reading or other subjects.

What Parents Should Look for in Any Tutoring Format

The research points to a consistent set of quality indicators regardless of format — these matter more than the specific program or brand.

Frequency of contact. This is the single most important variable in the research. Whatever format you choose, three or more sessions per week produces meaningfully larger gains than once or twice per week. If cost constrains frequency, the research suggests fewer sessions of higher quality are preferable to more sessions of lower quality.

Tutor qualifications. The Nickow meta-analysis found that professional and paraprofessional tutors produced larger effects than peer tutors (though peer tutors still produced meaningful effects). A tutor with formal training in teaching the relevant subject, or with demonstrated expertise and experience working with children at your child’s level, is more likely to produce gains than a tutoring center’s lowest-cost staffing option.

Assessment and progress monitoring. Effective tutoring requires knowing what a child currently knows and doesn’t know, teaching specifically to those gaps, and monitoring progress over time. A tutor who does not conduct some form of diagnostic assessment at the start and does not track progress over time is teaching in the dark. Ask explicitly how the tutor will know whether your child is improving.

Subject match. The research on tutoring effects is stronger for math than for reading, and reading is more complex to tutor effectively because it requires both decoding instruction (phonics) and comprehension instruction. A reading specialist with training in structured literacy approaches will generally produce better outcomes for a struggling reader than a generalist tutor, regardless of format.

Connection to school curriculum. Tutoring that directly addresses the gaps created by the child’s school curriculum — targeting exactly what the child is missing — tends to produce stronger transfer to school performance than general enrichment or program-specific curriculum that doesn’t match the school’s approach.

After-school programs more broadly — including tutoring components — are reviewed in the research on after-school program effectiveness, which covers the organizational factors that distinguish programs that move the needle from those that don’t.

What to Watch for Over the Next 3 Months

If you have enrolled your child in any tutoring format, here is what to monitor over the first quarter.

Academic benchmark indicators at the 6-week mark. Compare your child’s reading fluency, math accuracy, or writing ability — whichever was the target — against where they started. If you don’t have a baseline, ask the tutor to conduct one immediately. Six weeks of high-dosage tutoring in math should produce detectable improvement in targeted skills; six weeks of weekly tutoring may not.

The relationship between your child and the tutor. Research on test anxiety in children is relevant here — children who find tutoring sessions anxiety-provoking due to a poor relationship with the tutor will not produce the engagement that effective learning requires. A child who dreads sessions is not getting the benefit even if the tutor is technically competent. Monitor your child’s attitude toward sessions, not just their performance.

Whether the tutor is adjusting based on progress. Ask the tutor every four to six weeks: what specifically has your child improved at, and what are you currently working on? A tutor who cannot answer this specifically is not monitoring progress with enough precision.

Transfer to school performance. The goal of tutoring is academic improvement, not just session performance. Are grades improving? Is the child completing school assignments more confidently? Is the teacher reporting improvement? Transfer to school performance is the actual outcome; session performance is a proxy.

Frequently Asked Questions

Is Kumon worth it? The independent research on Kumon is limited, and the rigorous studies that exist show modest gains in procedural computation but limited evidence for gains in mathematical reasoning or problem-solving. Kumon may work well for building procedural fluency in a specific child; it is not well-supported as a general academic improvement program by high-quality evidence. The daily practice component — which distinguishes Kumon from center-only programs — is theoretically consistent with spaced practice research, but the specific Kumon implementation hasn’t been rigorously tested.

Is online tutoring as effective as in-person? The research suggests that online tutoring produces equivalent outcomes to in-person tutoring when frequency and tutor quality are equivalent. Format alone does not significantly affect outcome. The practical difference is that online tutoring offers more scheduling flexibility and typically costs less, while in-person may be easier for younger children to sustain attention and engagement.

What is high-dosage tutoring? High-dosage tutoring is defined in the research as three or more sessions per week, typically 45 to 60 minutes per session. The term emerged from school-based tutoring program research as a way to distinguish intensive tutoring models from typical once-weekly arrangements. High-dosage tutoring has the strongest evidence base of any tutoring format for producing meaningful academic gains.

Does once-a-week tutoring do anything? The research suggests once-weekly tutoring produces much smaller effects than high-dosage formats. Effects are not necessarily zero, but across multiple meta-analyses the gains from once-weekly tutoring are modest enough that they may not justify the cost for families expecting significant academic improvement. If budget constrains you to once weekly, focus on tutor quality and make the sessions as targeted as possible.

Is private tutoring better than tutoring centers? Not categorically. The research shows that one-on-one tutoring with a qualified educator produces larger effects than center-based group formats — but the comparison assumes equivalent frequency. A well-run tutoring center with high attendance frequency may outperform a low-frequency one-on-one arrangement. The quality of the educator matters more than the setting.

How do I know if a tutor is actually qualified? Ask directly about their formal training in teaching, their experience working with children at your child’s level, and their approach to diagnosis and progress monitoring. A tutor who has worked in classroom teaching, has subject-specific training (reading specialist certification, math education coursework), or has extensive experience with the specific challenge your child faces is a better bet than credentials in an unrelated field or general intelligence. Ask for references from current or recent families.

---

About the author

Ricky Flores is the founder of HiWave Makers and an electrical engineer with 15+ years of experience building consumer technology at Apple, Samsung, and Texas Instruments. He writes about how kids learn to build, think, and create in a tech-saturated world. Read more at hiwavemakers.com.

---

Sources

1. 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 No. 27476.
2. Fryer, R. G. (2017). “The Production of Human Capital in Developed Countries: Evidence from 196 Randomized Field Experiments.” In Handbook of Field Experiments, Vol. 2. Elsevier.
3. Dietrichson, J., Bøg, M., Filges, T., & Klint Jørgensen, A. M. (2017). “Academic Interventions for Elementary and Middle School Students with Low Socioeconomic Status: A Systematic Review and Meta-Analysis.” Review of Educational Research, 87(2), 243–282.
4. Pane, J. F., Griffin, B. A., McCaffrey, D. F., & Karam, R. (2019). Effectiveness of Cognitive Tutor Algebra I at Scale. RAND Corporation.
5. Clark, R., et al. (2007). “Kumon Method: Effects on Mathematical Achievement in UK Students.” Learning and Instruction, 17(4), 388–400.
6. What Works Clearinghouse. (2014). Kumon Math and Reading Program: WWC Intervention Report. U.S. Department of Education.
7. Cohen, P. A., Kulik, J. A., & Kulik, C. C. (1982). “Educational Outcomes of Tutoring: A Meta-Analysis of Findings.” American Educational Research Journal, 19(2), 237–248.
8. Hattie, J. (2009). Visible Learning: A Synthesis of Over 800 Meta-Analyses Relating to Achievement. Routledge.
9. Dynarski, S. (2022). “The Evidence on Tutoring.” Journal of Economic Perspectives, 36(4), 183–206.
10. Kraft, M. A., & Falken, G. T. (2021). “A Blueprint for Scaling Tutoring and Mentoring Across Public Schools.” AERA Open, 7(1).

---