Methodology

How we decide who goes first

Published in full, because a ranking you can't interrogate is just an advertisement with extra steps. If you think we've got a business wrong, this page tells you exactly which lever to argue about.

What goes into the order

4 factors, weighted as below. The weights are the same in every category and every location.

Status: live. The score these weights produce is computed once per data refresh and is identical for every visitor. It is shown out of 100 on each listing. Within a page, businesses on close scores are then ordered by how well they match that particular trade and place — explained below.

  • Review history

    35%

    Volume first, rating second. The difference between five reviews and fifty is meaningful, so volume is scored on a log scale and weighted more heavily than the star average — a 4.9 from six reviews does not outrank a 4.6 from four hundred. The average itself is adjusted for how much of it there is: a perfect 5.0 from a single review is pulled most of the way back to the directory average, while a 4.8 from four hundred barely moves. Nearly 40% of the businesses here hold a nominal 5.0 and their median review count is 13, so an unadjusted star rating separates almost nobody. The median business in the directory has 40 reviews.

  • Site reachability

    25%

    Whether the business's website actually resolves, and how recently we last genuinely reached it. A dead site is excluded from the directory outright rather than ranked low. A site that answers but blocks automated readers scores below a live one and well above one we could not reach at all. Having no website is not the same as having a broken one, and is scored as the far milder thing it is.

  • Independent corroboration

    20%

    How many separate sources describe the same business, and how strongly they agree. Open mapping data maintained independently of Google is the primary cross-check, and it has been run for 45,403 of the 66,104 listings so far. This is a floor rather than a ladder — most listings sit at or near the same value, so it marks out the thinly-sourced bottom of the directory rather than ordering the top of it. Where the match has not been run, that half of the component is the directory average and the listing's coverage figure says so. It used to carry the largest weight on this page, which overstated what it actually does.

  • Contact integrity

    20%

    Whether the phone number on the listing matches the one the business publishes itself. A contradiction is the only thing that scores zero here; having nothing to compare is treated as missing evidence, not as a mark against the business.

Two different questions

The score above answers one question: is this business real, contactable and well regarded? That is a fact about the business. It is the same number on every page it appears on, and it is the number printed on the listing.

Ordering a page needs a second question: is this a good answer to this trade, in this place? That one depends on the page as much as the business, so it cannot be baked into a single global figure — and until August 2026 we tried to use one number for both. It did not work. Across the 176 town and city pages with twenty or more businesses, 18.6% of top-ten places were held by a business more than 25km away, and one in five had somebody at least 15km closer whose score was within three points. On one page a business 48km out sat above one 3km out on identical scores, because distance was not even a tie-break and review count settled it instead.

For a directory whose whole job is finding someone who will come to your house, that is the wrong answer. So proximity and trade focus now adjust the order — within a limit.

The limit, stated plainly

Fit can move a business by at most 6 points either way. It reorders businesses whose scores are already close together and can never lift a weak nearby business above a strong one. Fit breaks near-ties; the score decides real differences.

That is checkable, and we checked it: after the change, the highest-scoring business on a page is still in its top ten everywhere, having lost at most 0.1 of a point of ground.

  • How close it is. Measured against how spread out that page actually is, not a fixed radius. 25km from the middle of Melbourne is most of the city; 25km from Broome is next door. A business at the page's median distance sits in the middle, closer scores better, and the same rule works for Sydney and for Tennant Creek without a special case. Coordinates more than 500km out are treated as a mapping error rather than a long drive, and count as unknown.
  • How squarely it does this trade. Taken from the categories the business lists for itself, where it publishes them. A plumber listed as a plumber is a better answer to “find me a plumber” than a general contractor who lists plumbing fifth. Related categories do not count against anyone — a plumber who also lists gasfitting, drainage and hot water is being specific, not vague. This is a weak signal and is weighted as one: 88% of the businesses that publish any categories at all score full marks.

Neither of these is a fact about quality, which is why neither is in the score. A business that is far away or does many trades is not less trustworthy — it is just a worse fit for what you asked. And as with every other check here, a business is never penalised for something we do not know: where a location or a category list is missing, the average is used instead of a zero.

A national list has no place to be near, so on those pages the order is the score alone.

When a check hasn't run

Not every check completes for every business. A site can time out, a phone number can have nothing to compare against, and some of the directory is simply still in the queue. There are three different states here and they are not scored the same way:

  • Checked, and it passed. Scored on what we found.
  • Checked, and it failed. Scored low. This is real evidence and it counts against the listing.
  • Not checked, or nothing to check. The component contributes the directory average, not a zero. A business is not marked down because we haven't got to it.

This changed in August 2026, and the previous behaviour was a real fault rather than a refinement. Until then an unchecked signal scored identically to a failed one, and the effect was large: a business where all our checks had completed averaged more than double the score of one where none had, while actually holding fewer reviews. The ranking was substantially a readout of our own crawl coverage presented as a judgement about businesses.

Because a score can now rest partly on an average rather than a finding, every listing publishes the difference. Alongside the score you will see what proportion of it comes from checks that actually finished. A 68 we have fully verified and a 68 we have partly assumed are different claims, and the page says which one it is making.

How far each check has got, in numbers. Of 66,104 businesses we have fetched 37,038 websites, compared 31,654 phone numbers against the business's own site, and run 45,403 cross-source matches. Those gaps are a backlog, not a verdict, and they are why a great many listings show a coverage figure well under 100% — comparing two businesses means reading their scores and their coverage. The cross-source match in particular has only been run over the businesses this directory started with, so a newer listing shows no corroboration finding either way.

The imputation is per component, and it can be partial. Corroboration is two things — how many separate sources describe a business, and how confidently we matched it to one of them — and the second is unavailable for any business we have not run the matcher over. That half is imputed and charged to the coverage figure; the half we do know is scored as found. Treating the missing half as a zero would have cost 10.7 points out of 100 for a lookup nobody had done, which is exactly the fault described above in a smaller place.

There is no way for a business to obtain the average by avoiding a check. The checks are ours to run, and a website that answers gets read.

What we check

  • Corroborated by an independent source

    We match every listing against open mapping data maintained separately from Google. One source is an entry; two independent sources agreeing is evidence the business exists as described.

  • Review history, not just the average

    4.7 from twelve reviews is a different fact from 4.7 from three hundred. We hold the count alongside the rating, weight the count more heavily, and show both.

  • A website that actually loads

    Checked on a schedule, and we record the date we last genuinely reached it rather than the date we last tried. Businesses whose site is dead are excluded from the directory entirely.

  • Contact details that agree with each other

    The phone number on the listing checked against the number on the business's own website. A mismatch usually means a stale listing or a lead reseller sitting between you and the person doing the work.

  • ABN confirmed against the national register

    Where a business publishes an ABN we confirm it and show the registered entity alongside the trading name. Most small businesses don't publish one, so this is a bonus signal rather than a baseline.

  • Published opening hours

    Read from the listing rather than the business's marketing copy. Where hours could not be read we say so rather than guessing, because publishing an open business as shut is worse than saying nothing.

What is deliberately ignored

  • Whether a business advertises with us — there is no advertising product
  • Whether a business has claimed its listing
  • How much a business spends on anything, anywhere
  • Whether a business has complained about its position
  • Any relationship, commercial or otherwise, with any other company

Where this method is weak

It cannot see when reviews arrived. Forty reviews earned steadily over six years and forty earned in a fortnight score identically here, and the second is far more likely to have been bought. We hold review counts but not the dates behind them, so until that changes this page will keep saying so rather than implying a check we don't run.

A confirmed ABN is not part of the score. It looks like it should be, and it used to be. But the lookup has run for about a tenth of the directory and has never returned a negative result, which makes it a record of what we have checked rather than a fact about the business. A check that cannot fail cannot rank anything, so scoring it meant nine businesses in ten lost points for our backlog. It is still shown on the listing and you can still filter for it.

It still favours businesses with a web presence, though less sharply than it did. A superb sole trader working entirely on word of mouth gives us less to verify, and an evidence score will reflect that. What changed is the size of the penalty: having no website is now scored as the mild thing it is rather than as a failure, because businesses without one average 4.59 stars against 4.70 for those with a live site — a gap far too small to justify what we were doing to them. It remains a genuine limitation rather than a claim about anyone's work.

It is slow to reflect change. A business that has recently declined can carry a good position for a while on the strength of its history, and a new business with excellent work has to accumulate evidence before it climbs.

It cannot see the job you are actually hiring for. Our checks measure whether a business is who it says it is. They cannot tell you whether it is good at bathroom waterproofing specifically. Read the reviews for that, and get more than one quote.

Think we've got something wrong?

Corrections are free and we act on evidence, not volume of complaint. 66,104 businesses are listed here and some proportion of what we hold about them is wrong.

Report a problem