Bangkok, ThailandMediHub101MediHub101 home

How the ranking works

MediHub101 publishes the 101 top-rated providers in each category in Bangkok, ranked by their Google rating and the volume of reviews behind it. The order is decided by one formula applied to every clinic in the same way, and nothing else. No clinic can pay for a place, improve its position, or have itself removed.

The problem with star ratings

A clinic showing 5.0★ from eight reviews is not better than one showing 4.7★ from three thousand. It is simply less measured. Sorting by rating alone puts the least-known clinics at the top of every list, which is the opposite of useful when you are choosing somewhere to have a medical procedure in a city you may not live in.

The score

Each clinic is scored with a Bayesian shrinkage estimator. It starts every clinic at the category average and lets it move away from that average only as far as its review volume justifies:

score = (v / (v + m)) × R + (m / (v + m)) × C

R
— the clinic’s own Google rating.
v
— how many Google reviews it has.
C
— the average rating across every clinic in the category, computed fresh on each run.
m
— the confidence weight, currently 150. At exactly 150 reviews, a clinic’s own rating and the category average carry equal weight. Below that, the average dominates; above it, the clinic’s own record does.

The floors

Before any scoring, a provider must have at least 50 reviews and at least 4.0★. A provider that fails either is not ranked low — it is absent. There is not enough public evidence about it to place it at all.

The ceiling at 4.9★

This is the least obvious rule and the most important one. In Bangkok’s aesthetic market, a perfect score is not rare: in the first full sweep, 266 clinics showed exactly 5.0★ with more than 50 reviews — more than a list of 101 can hold. Ranking on the raw number filled the entire list with 5.0s and pushed out the three most-reviewed clinics in the city, each sitting at 4.9★ with thousands of reviews.

Two things follow. Google publishes ratings rounded to one decimal, so 4.9 and 5.0 are a single rounding step apart — treating that gap as decisive reads more precision out of the number than it contains. And a true 5.0 sustained across thousands of patients would mean nobody was ever dissatisfied, which for elective medical work is not a quality signal so much as a sign that the review set is not a random sample.

So above 4.9★ we stop distinguishing. Everything at 4.9 and above is scored as 4.9, and the weight of evidence decides the order among them. Nobody is penalised and nobody is accused of anything — a perfect score simply stops buying rank the evidence cannot support. The rating you see on a provider’s page is always the real Google number, untouched. The ceiling applies to every category, on the same evidence: a rounding step is a rounding step wherever it appears.

One consequence is worth stating plainly. Near the top of a list, most providers are at or above the ceiling, so their ratings are effectively tied — and what separates one from the next there is the volume of reviews behind it. A provider is placed above another because more patients have rated it just as highly, not because we have judged its reviews to be better ones. That is why the lists are described as top-rated and ranked by review volume, rather than as a verdict on any provider’s quality of care.

At most 2 branches per brand

Chains run many branches, each with its own Google listing and its own reviews. Left alone, one chain took seventeen of 101 places. A list like that is a ranking of chains, not a way to discover a provider — so no brand may hold more than 2 places. When a third branch would qualify, its slot passes to the next provider down, and ranks stay contiguous.

What counts as belonging on a list

Google’s own category labels are unreliable in both directions — hair salons are filed as beauty clinics and genuine clinics are filed as doctors. Each candidate is therefore checked against the category it is being ranked in before it is scored, and anything in a different business entirely is dropped rather than ranked badly.

Where the data comes from

Ratings, review counts, addresses and contact details come from Google, via the Places API, and are shown exactly as Google reports them. MediHub101 is not affiliated with or endorsed by Google.

We deliberately do not store or republish the text of anyone’s review — the rating and the count are the signal we rank on, and the reviews themselves belong on Google, where every provider page links to them.

Every provider’s rating and review count is read again from Google about once a month, and every list is re-ranked from the new numbers. Places move up and down between runs, places that have opened since the last read are added, and the date of the run that produced the current list is printed at the top of every ranked page. Between those reads the numbers you see are the ones from the last run — so the date at the top of the page is how current the ranking is, not how recently the page was loaded.

What the treatment filters mean

Some lists can be filtered by treatment, and some provider pages show the treatments their reviewers named. Those come from a much smaller body of evidence than the ranking does, and it is worth being exact about how small. Google’s API returns at most 5 reviews per request, and Google chooses which 5. The one thing that changes its choice is the language the question is asked in, so we ask once in each of English and Thai and merge the answers, counting a review that comes back in both only once. That is at most 10 reviews per provider, and usually fewer. We count how many of them name each treatment, keep the counts, and discard the text.

A treatment is listed against a provider when at least 2 of the reviews we read name it, and a provider is listed with at most 3 treatments — the ones its own reviewers named most. A treatment earns a filter chip only once at least 3 providers carry it, because a filter that leads to a single result is worse than no filter. Every count shown is written as what it is: mentions out of reviews read, both numbers, always.

So the filter answers one question — whose reviewers mentioned this — and deliberately not the question you might expect. It is never a ranking. There is no order within a treatment, no best, and no claim that a provider is good at anything it is listed under. A Google star rates a visit as a whole and never one treatment, so a rating is never printed beside a treatment name. And a provider absent from a filter is not evidence that it does not offer that treatment; it is evidence about 10 reviews.

These counts are re-read on the same monthly cycle as the ratings, and each read replaces the previous counts rather than adding to them. What you see is what this month’s sample said, not a total accumulated over time.

Coverage, honestly

A ranking is only as good as the set it ranked. Our sweep works through a fixed list of search terms across a fixed set of Bangkok districts, different for each category, and it publishes what it has ranked so far rather than waiting for the sweep to finish — an empty page while discovery runs would be less useful and no more honest. So a list can be incomplete, and when it holds fewer than 101 the page says how many it holds instead of padding it. Coverage widens each night until the sweep is done, after which the whole list is re-read about once a month.

Each category is swept separately, with its own search terms and its own districts, so a place missing from one list is not evidence about any other. A business that genuinely operates in two of them — a hospital with an aesthetic department, a practice that is both dental and cosmetic — can appear in both, because it really does belong in both.

The picture at the top of a provider page

Where a provider page opens with a photograph, that photograph is the provider’s own, served from the provider’s own website, and credited to it underneath. We store its address, never a copy of the image, and we re-read it on the same monthly cycle as everything else — so when a clinic changes the picture on its site, the picture here changes with it. No image is bought from Google, and none is taken from a third party. Where a provider has no website we can read, the page opens on a plain colour instead, which is a statement about our reach and not about the provider.

Finding it means one request to the provider’s home page — the same request any chat app makes when somebody pastes a link — from a browser identified as MediHub101PhotoBot. We prefer the image a site nominates for link previews in its own markup; where a site nominates none, we take the picture its home page is already showing every visitor. Any operator who would rather we did not can say so in their robots.txt, addressed to that name, and we will stop.

Ranked, not verified

This is a measurement of public review evidence. It is not an inspection, an accreditation, or medical advice. MediHub101 does not visit clinics or hospitals, check licences, or vouch for anyone’s clinical outcomes, and a high rank is not a recommendation to have a procedure. Check a provider’s credentials, and your doctor’s, before you book anything.

What is ranked