How the index is measured
Everything needed to reproduce our counts. If you find a mistake in here, email info@list-it.nz and we will correct it on this page with a date.
Sources
Two, and only two. The New Zealand Business Number register, taken from the monthly bulk data release published by the Companies Office, which gives us the population of registered entities and what each record carries. And our own reads of business websites: we fetch the business ’s own site and test it for four facts. No third-party data, no panel, no survey, no estimate.
What we read
Four facts, each one a yes or no, each taken from the business’s own website and nowhere else. The exact tests:
- About text: 40 characters or more of text describing what the business does.
- Phone: a dialable number in the page text, 6 to 12 digits.
- Logo: a logo image on the page.
- Socials: a link to at least one social profile.
Nothing List-It has added to a listing counts towards that listing’s score. This matters more than it sounds: the very first run ranked the Waikato top because 37,203 descriptions in our database had been written by our own AI enrichment, mostly there. Reading only the business’s own site is what makes the number mean anything.
Hours and photos are not counted. Both can only be added by an owner who has claimed their page, and the index should not punish a business for a field it had no way to fill.
The pass mark
At least 3 of the 4 facts. A business that shows three of the four is counted as findable: an assistant has enough to identify it, describe it and put a customer in touch.
There were five facts until 13 September 2026, when services was dropped. It was read from a field we write during enrichment, and enrichment only ever ran for claimed listings, so for the roughly 97,000 unclaimed businesses in the index it was structurally always false and read 0% in every region and every trade. It also contradicted the rule above, that every fact comes from the business’s own site. The index had therefore always been a 3-of-4 measurement. Every figure it published was a correct 3-of-4 measurement; dropping the fifth changed no result, it made the code say what it was doing.
Reachability
A site that does not resolve when we visit scores zero of four and counts as not findable. An assistant cannot read a dead site any more than we can. That is a real findability failure and it stays in the headline number. Because it is also a different KIND of failure from a site that loads but is incomplete, we publish both figures every time: the share findable across all sites read, and the share findable among only those that loaded.
The denominator caveat
Register-wide percentages use every entity on the register as the denominator, including dormant companies, holding companies and sole traders who have never traded. That inflates the “missing” shares against the population of actively trading businesses. We do not filter it, because any filter would be a guess. Read these as figures for the register, not for the trading economy.
Order of reading
Websites are read in NZBN order, which is uncorrelated with region, trade and town. Coverage therefore grows evenly across the country rather than filling up one region first, so a partial read is still a fair comparison between regions.
Minimum sample sizes
A bucket is not published until it holds enough businesses to mean something: a region 100, a trade 200, a town 150, an AI engine bucket 50. Below those, the row simply does not appear. Every table carries its n, and a row without n is not a row.
What we never do
- We never name a business at the low end. A region or a trade can rank last. A business never does.
- We never compare this index to an overseas one built on a different measure.
- We never publish an engine bucket under 50 checks, or without its sample note.
- We never run a headline this page cannot defend in one paragraph.
Our own scores
We publish our own two listings against the same test, whatever it says. Both currently sit at 75 and neither is recommended by name yet, which is the honest state of it.
| Business | About text | A dialable phone number | A logo image | A link to at least one social profile | Findable | AI score | Recommended |
|---|---|---|---|---|---|---|---|
| Bay Web Co baywebco.nz | Yes | Yes | Yes | Yes | Yes (4 of 4) | 75 | No |
| List-It list-it.nz | Yes | Yes | Yes | No | Yes (3 of 4) | 75 | No |
Four facts read 13 September 2026. AI check last run 17 September 2026.
Corrections and change log
Anything that changes how a number is computed is dated and listed here, permanently.
- 13 September 2026: Services dropped from the test. It was read from a field List-It writes during enrichment, which only ever ran for claimed listings, so it was structurally false for the ~97k unclaimed businesses in the index and read 0% in every region. The index had always been a 3-of-4 measurement; the code now says so.
- 15 September 2026: The web and SEO category headline was withdrawn. The bucket was too loose to carry a headline, and the true figure for named marketers was 76.2%.