JohnButt: A few points to include in the debate:
We are trying to get the same number of probes per ISP, having said that it is easier to get Telecom probes so there is more, but not proportional to their market share (which by the way is around 50% from the last stats I saw)
Yes this traffic is only National and that is a smaller proportion of NZ traffic, we are testing more but have not the time yet to publish - I hope to correct that in the near future.
The browsing test is however some representation of a lot of usage issues, international browsing for example is unlikely to be much better than National, so it does give some indication, poor DNS and poor latency are also not going to help browsing.
Telecom has not always been "the best", maybe they put the greatest effort into getting to be the best
And lastly, the gap is small, but the dummy page is also small at 300k, a larger page would drive download times up, not necessarily proportional.
Yes Telecom has put a lot of work into their recursive DNS clusters [server clustering, running a fast DNS engine and doing things like re-writing TTLs and caching them for longer than normal....] to ensure their users get the fastest DNS response times,
My comments around the international traffic was not to show better results, I'm expecting worse across the board, however that's the point as that's more representative of user experience as ISPs peering, , prioritization, caching, over-subscription etc will all come into play.
Is it possible to have the test page be more loaded in terms of content as is say the homepage to www.nzherald.co.nz or www.cnn.com where longer loading times will reduce the margin of error?
I do look forward to the expanded testing results and hope your newly won contract will help you get there.