Research

What 943 one-to-one lessons show about expert tutoring.

Phoenix Tutors is the tutoring practice of Educave. Over two years, Educave Research recorded and analysed expert one-to-one lessons to describe what good one-to-one teaching is actually made of. This is a short summary for parents.

1,418
lessons recorded
943
analysed in detail
7,546
mistakes classified
74,258
questions classified

The study in brief

Human tutor intelligence, by Carl Morris, was published by Educave Research in August 2026. It draws on 1,418 one-to-one science and maths lessons recorded between August 2024 and June 2026, about 1,404 hours of teaching. Of those, 943 lessons with 48 regular students were analysed in detail: who spoke and for how long, what was asked, what went wrong, and what the tutor did next.

It is a description of expert practice. It is not a trial, and it does not claim that one way of teaching gets better grades than another.

Not all mistakes are the same

Every mistake corrected in these lessons was sorted into one of three kinds. Choose one to see what it is and what happened next.

  • Careless slip
  • Misconception
  • Missing prerequisite
Misconception: 30%

What it is. The student is working from a wrong idea of how something works.

What happened next. Guided or retaught 83% of the time. The answer was simply handed over 17% of the time.

So 45% of mistakes were not careless. A grade shows that a mark was lost. It does not show which kind of mistake lost it, and each needs a different response.

Share of 7,546 mistakes corrected across 943 lessons. "What happened next" comes from a smaller set of 121 corrections.

Would you give the answer?

Four moments from a lesson. Decide what you would do, then see what happened in the lessons studied.

The student is partly right
The student is wrong
The student is truly stuck
The student asks for the answer

When to hold back depends on where the student is. A tutor that always explains would have made the usual call 46% of the time.

From 308 moments set aside for testing.

The questions asked

Most were concrete. Open questions, and questions about the student’s own thinking, were the smallest share.

  • Closed or factual43.5%
  • Procedural25.5%
  • Socratic21.5%
  • Open and exploratory7%
  • About the student’s own thinking2.4%

Share of 74,258 questions.

How much students talk depends on the subject

The same teaching does not suit every subject. In maths the student did almost twice as much of the talking as in biology.

  • Science, general23%Youngest learners, most interactive
  • Maths21%Most student-led; mistakes are mostly slips
  • Chemistry15%Balanced across every measure
  • Physics15%A lot of guided, prompted support
  • Biology11%Most explained by the tutor; most misconceptions

Median share of the talking done by the student, from an earlier sample of 738 lessons.

How to read these numbers

The study calls its own figures exploratory. Two people independently coded a sample of ten lessons and agreed closely on the mix of question types and mistake types, so the proportions are reliable; the raw counts are less so. Coding was 81% complete when the study was published.

The findings describe expert teaching as it was recorded. They are not proof of what works for every child.

What it means for a Phoenix lesson

A grade records the outcome and leaves out the reason. That is why every Phoenix lesson is written up the same day: what was taught, what landed and what comes next. A person on our team reads each note before it reaches you.

See how a lesson is recorded

Want to know what is behind your child’s marks?

A private consultation with Carl, Saif or one of our education consultants. Twenty minutes, free.

QuickestWhatsApp+44 7440 496063Message us
Most detailA few linesThree questions. A person on our team reads it and replies.Write to us
In your own timeEmailhello@phoenix-tutors.comEmail us