Hi Ahmad,

1.) Statistics provided me additional data beyond what was published in the ISP report
2.) I did not use Google IPv6 information, I used APNIC data, and there were more than 60k samples.
3.) Vector Fibre has been a retailer for some years now
4.) It doesn’t matter if addresses are NATTed, the amount of traffic from an AS represents a user count, not a count of held or advertised IP blocks.

LikeLike

Maybe I’m missing something, can you explain further, as far as Statistics NZ does not provide any break up by ISP except a total numbers, and Google IPV6 information only indicates availability of IPV6 and it is small sample of ~4K for NZ . A network having an AS number or allocated IP address block by APNIC does not means they are definitely using it for their subscribers. Also Vector for example is not a retailer and 2deg are only mobile, even if tethered laptops, the will be likely NATed

Regards

LikeLike

ah, so this is based on fixed browser usage (from the gogle adwords numbers?) Would the share estimates therefore be skewed towards ISPs that have users that simply use the internet more?
i.e. spark probably has a load of old biddy customers who use the internet very rarely, vs, say, Snap who’s customers will be much higher net users. Would this skew spark’s share downwards vs Snap’s?

LikeLike

Great question. I think this is users who are tethering their laptops to their phones or wireless hotspots. My guess as to why the number is so small is that almost no one tethers full-time.

LikeLike

LikeLike

Hi Bill. There are a few things going on. The ComCom’s April 2014 report covers 2013. It also cites 1.32 million connections, which is 600k less than Stats New Zealand reports (see their ISP survey). Additionally the numbers I’ve used break out some significant users like universities and very large businesses. I have excluded those users from my data, though next time I have a go at it I will use ASN path to assign these large systems to upstream carriers. In many cases it’ll be Spark who is the upstream carrier. It’s likely Spark’s share is somewhere splitting the difference between the two figures.

LikeLike

There’s a huge difference between Spark (Telecom) on 49 percent market share at the Commerce Commission and 41 percent here. What do you put that down to?

LikeLike

Hi Jesse, I manually merged data to take into account mergers & acquisitions and will continue to, as we stop seeing traffic on some of the old ASNs. The data set goes back four years, so samples are biased towards operators who have been around the whole time. I will adjust my next estimate to take into account new entrants. I have a feeling MyRepublic and a few of the rural WISPs will get a bump.

LikeLike

Same goes for Vodafone and the old TelstraClear.

Does anyone know how old those APNIC samples are or is it a rolling 30 day data set (data collected 30+ days ago being discarded) ?

LikeLike

Hi John, thanks for the feedback. I will consider splitting out sub-brands in the future.

LikeLike