How tutor listings rank

In plain English

Tutors are sorted by their review score, smoothed against a platform-wide baseline so a single 5-star review can't leapfrog a tutor with fifty 4.6-stars.
Older reviews count less than recent ones. A tutor's score reflects recent performance, not all-time history.
Safeguarding-verified tutors (DBS, PVG, AccessNI) get a small lift on their score.
Scores are rounded to one decimal place to match the star rating you see displayed on tutor cards.
When two or more tutors come out with the same final score, their order rotates each day so every tutor in that tied group takes a turn near the top. The animation below shows it in action.
Three page-1 slots are held open for tutors who joined in the last 30 days and who wouldn't otherwise have made page 1, clearly labelled as new. If a new tutor has already earned their spot in the ranked order, they keep it, no label, no special treatment. If no recently-joined tutor needs a lift, the slots fill from the ranked order like every other position.

How page 1 is built

Below is a 30-day simulation of a single search showing both layers in one view. The two-column grid is the actual page-1 layout, 24 slots. Two tutors at the top are tied at five stars and swap each day; the tutor below them has a uniquely lower score so they stay put; further down, tied tutors take turns. Slots 10, 15, and 20 hold space for tutors who joined recently and would otherwise be off-page, they rotate among that newcomer pool. Watch the "views" counter on each card accumulate.

Example scenario · 30-day daily rotation on page 1

Day

of 30

Top-rated tutor

Score-tied tutor

Baseline tutor

New to Tutorperch (reserved slot)

Slots 10, 15, and 20 are reserved for tutors who joined recently (carrying the "New to Tutorperch" label). The other 21 page-1 slots fill from the ranked order, with tied tutors taking turns each day. Watch the "views" counter on each card: page-1 tutors climb steadily while off-page tutors barely move.

Why reserve slots for newcomers rather than boost them inside the ranking? Promoting newness inside the score would say a new tutor is somehow "better" than an established one with the same evidence, which they aren't. A separate, labelled surface gives new tutors a visibility floor without distorting the ranking of everyone else. The research note works through the trade-off in more detail.

The formula

For the curious: the smoothing is a Bayesian-weighted average. The formula sits below, with the variables explained underneath. The result is rounded to 1 decimal place; the safeguarding lift is added after that rounding, so it sits a tutor just above non-verified peers with the same rounded score rather than promoting them across a whole rating tier.

score = \frac{(n_{eff} \times r_{eff}) + (m \times C)}{n_{eff} + m} + β

β applies only to safeguarding-verified tutors.

The two ingredients carry the time decay. Each review is weighted by its age, so a fresh review counts for full weight and an old one counts for a fraction. A review that is age days old gets weight:

weight = 2^{- \frac{age}{547}}

547 days is an 18-month half-life: a review halves in weight every 18 months.

Summing those weights across a tutor's reviews gives the effective count n_eff; the weighted mean of the ratings gives the effective average r_eff. Those two feed the formula above in place of a plain count and average. In the live directory the curve is applied in five age bands rather than the exact exponential, a portability concession that tracks the smooth curve closely; the principle is identical.

n_eff: The tutor's effective review count: every review summed by its decay weight from above. Reviews from the last six months count for full weight; older ones count for progressively less, with reviews more than five years old contributing very little.
r_eff: Their effective average review score (1–5 stars), weighted by the same decay factor so recent reviews count for more than older ones.
m: = 10. Prior weight. The number of reviews we treat as "enough" to trust a tutor's average. Below this, the formula pulls toward C; above it, the tutor's own average wins.
C: = 4.3. The baseline rating an unrated tutor sits at. Set conservatively below the typical review average so a tutor with many average-mean reviews genuinely outranks a tutor with no reviews. We'll recalibrate as the directory grows.
β: = +0.05. Additive bump for tutors with a current DBS, PVG, or AccessNI badge. Modest by design: a strong review record beats it.

m is the calibration knob. It sets how many reviews a tutor needs before their own average matters more than the platform mean. We've set it to 10: a tutor with around 10 reviews splits the score 50/50 between their own rating and the prior, and the tutor's reviews dominate above that. We'll raise m as the directory grows and tutors accumulate more reviews on average. Any change is announced here first.

A worked example

Four example tutors plus a row for you. Dial in your numbers below and watch where you land in the table, the constants are the same ones the live directory uses.

Try your own

Name Reviews Avg (0–5) Safeguarding-verified

Tutor	Reviews	Avg	Safeguarding	Actual score	Rank score
#1 Ben	50	4.6	—	4.55	4.60
#2 Anya	1	5.0	—	4.36	4.40
#3 Chiara	0	—		4.35	4.35
#4 David	0	—	—	4.30	4.30
#5 You you	0	—	—	4.30	4.30

Actual score is the full-precision Bayesian average. Rank score is the value the sort uses: the Bayesian average rounded to 1 decimal place, with the safeguarding bump added afterwards. The rounding is why two tutors who look a hair apart in the actual column can land on the same rank score and rotate day-to-day.

Ben: Fifty reviews at 4.6. The real average dominates the prior.
Anya: One 5-star review. The prior pulls her score toward the platform average, she sits above the baseline but below tutors with more reviews. Around six perfect reviews would round-match Ben at 4.6 and bring her into the daily rotation alongside him.
Chiara: New tutor, no reviews yet. Sits at the prior plus the safeguarding bump.
David: New tutor, no reviews, no safeguarding badge. Sits exactly at the prior.

What we don't reward

Response time isn't in the formula. You might be teaching, away, or choosy about who you take on, and none of that says anything about how good a tutor you are. Hourly rate isn't in there either: the default sort doesn't push cheaper tutors above pricier ones (students who want that sort pick it explicitly). Recency of joining doesn't lift a score directly; instead, when two tutors come out with the same final score, their order rotates day-to-day so every tied tutor takes a turn near the top.

Tutorperch doesn't sell placement. No boosted tier, no featured slot. Tutors pay no fee for being listed and no fee to rank higher; the £9.99 finder's fee and the £3.00 identity-verification fee are the only payments on the platform, and neither moves a tutor up the page. Admins can hide profiles for safeguarding or compliance reasons but can't push one tutor above another. The ranking code is the same for everyone, including us. If we ever change the formula or the constants, you'll see it announced on this page first.

For a deeper read on how we designed the rotation tiebreaker, the alternatives we considered, and the simulations we ran, see our research note: Fairness in tutor rankings.

Reviews are gated. Only a student who's paid the finder's fee for a specific tutor can review that tutor, and the review window opens 30 days after the payment (so reviews reflect at least one real lesson). Coordinated review manipulation, sock-puppet reviews, and reviews of tutors you haven't actually engaged are all grounds for sanction under our Terms and Code of Conduct.

Tiers and city searches

The score above orders tutors within a single group. Two things put a tutor into a different group: whether they're currently accepting new students, and whether the search names a specific city.

Tutors not currently accepting students sit at the bottom of every list, regardless of rating, safeguarding badge, price, or location. Their profiles stay visible so past students can still find them to write reviews, but a parent shouldn't waste time messaging someone who can't take them on this week. This applies to every sort option, including "Price: low to high" and "Recently joined".

City-based searches like /tutors/maths/in/manchester tier the results by location: in-person tutors based in the named city first, then online tutors, then everyone else. The Bayesian formula orders within each tier. Geographic relevance beats a small ranking gap when parents are looking for nearby help.

The homepage avatar strip

The strip below the homepage hero shows 8 tutors, refreshed daily. Each tutor's score is their weight in a daily random draw. Stronger-scoring tutors come up more often, and the lineup changes every day.

A tutor shows up more often by earning reviews or by being safeguarding-verified. The same two signals search ranks on, nothing else. The draw has a fairness floor so the same handful of strong-rated tutors can't fill the strip every day; newer tutors with no reviews still have a real chance of appearing, and they'll show up more often once they've earned reviews. Tutors can't pay to be in the strip, same as the search rankings.

Robert S., Co-founder

How tutor listings rank.