Three things,
and expertise is the easy one.
A tutor in a small centre does three jobs at once. They teach, which requires knowing the subject and — much more discriminating — being able to explain it more than one way. They manage a relationship, because the family is buying attention as much as instruction, and because the tutor is the person whose name gets mentioned when a friend asks for a recommendation. And they are an operational unit: they arrive on time, they remember what happened last week, and they tell you when something is wrong before it becomes your problem.
Hiring processes overwhelmingly test the first third of that, and the least distinguishing part of it at that. A first-class degree tells you almost nothing about whether someone can sit with a fourteen-year-old who is embarrassed about not understanding fractions, and the person who can do that will keep families for years while the more qualified candidate loses them politely. This is not an argument against expertise — a tutor who does not know the material is useless — but expertise is a threshold to clear, not a scale to rank on.
Six signals,
ranked by how much they predict.
The first row is the one worth designing your whole process around. Ask a candidate to explain something — long division, the present perfect, whatever fits — and then ask them to explain the same thing to a student for whom that explanation did not work. Strong tutors have a second and third route ready and will ask you what the student found confusing. Weaker ones repeat themselves more slowly and with more vocabulary, which is exactly what they will do with a struggling child.
The third row catches something that matters more in tutoring than in classroom teaching. One-to-one students test boundaries within a few sessions, and one of the tests is asking something the tutor does not know. A tutor who can say "I'm not sure — let me check and I'll tell you next week" keeps the relationship. One who bluffs loses it, usually permanently, because a fourteen-year-old can tell.
Pay for it,
and use a real student.
Interviews select for people who are good at interviews. A trial session with an actual student selects for people who are good at the job, and it is the single most informative hour in the process. Pay for it — asking someone to teach for free is both a poor signal about the centre and, in some places, a problem in itself — and be present, or join part of it if you are online.
Choose the student deliberately. The instinct is to give a candidate an easy, cooperative learner, which tells you nothing you did not already know. Give them a student who is a little reluctant, or who is behind, or who says very little, because that is the population a tutoring centre actually serves. What you are watching for is not polish. It is whether they find out where the student actually is before they start teaching, whether they notice when they have lost them, and whether the student says more at the end of the hour than at the start.
Afterwards, ask for three sentences about that student — where they are, what they need next, what you should know. This doubles as the writing test in row five and tells you whether the tutor was paying attention to a person or delivering a prepared lesson at them.
Four sources,
and the best one is already yours.
Your own former students are the strongest source and the most overlooked. Someone you taught to a good result, who is now a few years older, already knows your standards, your subject approach and your families' expectations — and they are usually delighted to be asked. The second is teachers wanting additional work: experienced, reliable, and constrained to exactly the evening and weekend hours you need filled, which is a rare alignment. The third is postgraduate and final-year students, who bring enthusiasm and recent experience of the exams but need the most support on the relationship half of the job.
The fourth is referrals from tutors you already have, which becomes your best source once you have two or three good ones, because people who are good at this know others who are. Ask explicitly rather than hoping — a standing offer that you will always look at someone they recommend costs nothing and produces better candidates than any advertisement.
What rarely works well at this size is a general job board, which produces volume and very little signal, and consumes the screening time of the person who was already too busy. Use it if the other four are exhausted, and expect to filter hard.
Checks, references
and status.
Three things need settling before a new tutor sits with a student, and all three are local questions this article cannot answer for you. Whether a background check is required for someone working with children, and which one, is determined by your jurisdiction. Whether your insurance covers a person who is not you is determined by your policy, and is worth a phone call rather than an assumption. And whether the person is an employee or self-employed is determined by the facts of the relationship rather than by what the contract calls it, in a number of countries — directing someone's hours, materials and method while treating them as a contractor is a common and expensive mistake.
References are worth taking and are usually taken badly. The useful question is not whether the candidate was good but a specific one: "Would you give them your most difficult student, and why?" It produces an answer with content, where the general question produces a polite non-statement. Get local advice on the first three before the second tutor rather than after the fifth.