Summary

  • Read on 20–21 October 2026, ix.nz described the operator as not-for-profit and carrier-neutral, named three cities and named nine facilities with street addresses; a paginated participant directory showed one entry and no total.
  • The same reading produced no exchanged-volume figure carrying a unit, an averaging interval and a site or aggregate scope. A metrics path existed; a pricing path and an alternate metric host returned nothing because the retrieval request failed, which leaves them unread rather than empty.

The record the exchange publishes

A national exchange's public surface starts with a claim about its own character. The landing page at ix.nz describes the operator as not-for-profit and carrier-neutral, and names the three cities in which it operates: Auckland, Wellington and Christchurch.

Physical presence is then set out street by street. The retained excerpt names nine facilities: Datacentre220 at 220 Queen St, Auckland; 2Degrees DC at 7a Parkhead Pl, Albany, Auckland; Spark MDR at 31 Airedale Street, Auckland; DataVault at 162 Grafton Road, Auckland; DataCom – Orbit Datacentre at 6 Orbit Drive, Rosedale, Auckland; Xtreme DC at 191 Thorndon Quay, Wellington; Spark Wellington Central Exchange at 70 Featherston Street, Wellington; Christchurch Central Exchange at 109 Hereford St, Christchurch; and Computer Concepts Ltd at 21 Durham St South, Sydenham, Christchurch.

The excerpt carried elision markers, so nine is the number the retained text names, not a verified total. Facilities sit in commercial data centres bearing other operators' brands, which is what a carrier-neutral fabric looks like on paper — the exchange is present inside buildings it does not own.

Two registers complement that map. The participants path publishes a directory arranged across multiple pages; the retained text showed one entry, Verge Infrastructure Limited, and supplied neither a total nor a counting rule. At APNIC, the public member directory row read, verbatim:

| New Zealand Internet Exchange Incorporated | SMALL | NZ |

That row is administrative. It records a membership and a country code in the registry's own table. The reviewed material supplies no definition linking SMALL to traffic, utilisation, capacity, reach or commercial scale. BTW's directory entry for the entity, imported from that same member directory on 18 June 2026, carries one resource of type rir_membership classified as virtual, lists no ASN or IP resources, records a confidence score of 0.82, and states three data gaps — including verifying the legal registration record and mapping ASN/IP resources before any routing-footprint claim. The card also records a legal type of private_company inside a name that uses Incorporated; no primary legal-form document was retained, so the legal form stays unresolved here rather than settled by a word in the title.

So the record answers who is listed, where the fabric reaches, and what the operator says about itself. Those are real answers. They are not the same answer as how much traffic crosses.

Four elements, read together

An exchanged-volume claim can be tested against four elements that must appear together. A numeric figure. Its unit. Its averaging or sampling interval. And its scope — a named facility or the exchange in aggregate.

Each one carries weight. Without a unit the figure cannot be converted or compared. Without an interval, a peak over an unstated window, a mean and a monthly byte total are three different measurements that cannot be ranked against one another. Without scope, a facility-level reading cannot be aggregated into a national fabric or set against a single-site exchange. A timestamp or coverage period and a stated method complete the picture, because without them a reader cannot tell whether two published values describe the same thing.

Set against that test, the result of this reading is narrow and repeatable: no figure carrying all four elements appeared in the source material retrieved and retained.

What the metrics path held

The navigation carried a link labelled Metrics. Following it produced a page whose retained text did not supply a traffic figure, a unit, an interval or a scope. A navigation label establishes that the operator points readers toward a metrics surface. It does not establish what that surface contains, how often it refreshes, whether it is aggregated across sites, or whether any value on it is measured at the exchange or estimated elsewhere.

This distinction is not pedantry. Port capacity, participant counts and traffic volume move independently. An exchange can add a site without adding peers, add peers without adding volume, and add volume without any of it being comparable to another fabric's number. A reader who substitutes one for another gets an answer that no source supports.

The paths that did not open

Two further paths were part of this reading and returned nothing: a pricing path carrying port and connection options, and an alternate metric host reached from the operator's navigation. Both requests failed at the provider with a 503 and an unknown outcome, so no material was returned or retained.

That is a retrieval result, not a content finding. An unread page is not an empty page, and this article treats those two conditions differently throughout. If either path later becomes readable and carries qualifying data, the finding above should be updated rather than defended.

What this establishes

On 20–21 October 2026, BTW's review of the public page material it could retrieve and retain found no exchanged-volume figure accompanied by a unit, an averaging interval and a site scope. To state the boundary explicitly: this is a result about the reviewed public record. It does not establish that the exchange carries no traffic, that it keeps no internal measurements, or that any particular number is being withheld. Nor does it support a routing-footprint claim, because no ASN or IP resources were mapped, and it does not turn a facility list, a participant entry, a port option or an RIR membership tier into traffic evidence.

Why a register and a measurement answer different questions

A register answers who, where and listed-as. A measurement answers how much, over what interval and across what scope. The first is a claim about structure and presence; the second is a claim about load and behaviour. For an exchange that has built a multi-site national fabric, the presence record is genuinely informative: it tells a prospective peer which buildings to reach, which cities are covered, and that the operator presents itself as neutral between carriers. It tells almost nothing about whether joining will shorten a domestic path by a meaningful margin, and nothing at all about how much traffic already moves there.

That gap has practical consequences for anyone weighing interconnection. A capacity decision made on a register is a decision made on adjacency rather than on load. The four-element test is cheap to apply and it separates the two.

What would change the finding

A reachable page, feed or archive carrying a traffic number with a unit, an interval, a period and a scope would allow that measurement to be reported with attribution. A page showing only the word Metrics, an unlabelled graph, or a bare number without interval or scope would leave the measurement-basis gap exactly where it is. A complete participant export or an explicit total with a snapshot date would let a dated participant count be published; pagination alone does not. And a primary legal-form document would settle legal status rather than leave it to inference from a name.

None of those require the operator to say anything about strategy. Each simply converts a listing into a reading that a stranger can check.