Just curious about your NetRatings system, because it appears to me that it uses very simple statistical functionalities (descriptive statistics) for its analysis. Correct me if I am wrong here, that NetRatings only does simple analysis and not the advanced methods that can give your clients a deep understanding of their data. I can pick perhaps 3 obvious analytical techniques that are missing from NetRatings :

#1) It can’t do time-series sub-sequence matching of surfing behavior (matching one time-series to another one or to a group of other time-series to find out which is the closest match). Looking at those graphs above, there is no way to tell which time-series that have the same co-movement with which other series. This is important as the analyst can pick out important patterns.

#2) It can’t do sequential pattern extraction of users surfing sequence in a website. This is important that you can track the behavior of users. It gives the analyst the capability to see if there is a need for improvement of the website design, since there is emerging pattern from the log data that shows a high rate of a repeated sequence.

#3) I don’t know whether you system cluster users according to their browsing behavior. It is important to know this, since you can automatically predict what similar users going to be doing when they come to the website. For e-commerce , you can recommend products to such users who belong to same clusters (same browsing & buying behavior).

I understand that your system is developed in Australia, if you want more info on advanced data analysis, then I am happy to point you out to more info regarding those techniques.

SPSS has a web-analytic (web-mining) product, which they are moving in to those territory that NetRatings is occupying. All the functionalities I have listed above and other advanced analysis methods are already implemented by SPSS. SPSS is a data-mining vendor so, they will implement the latest state-of-the-art web analytic algorithms that appear in literatures.

You may be interested in our Monthly Online Retail Monitor. We survey around 750 Kiwis each month on actual spend and report via category, retailer, local vs. offshore, etc. A link some findings from the latest release can be found Here. There are a couple of large e-commerce sites in New Zealand that you probably wouldn’t think of but they are there.

Thanks,
Glen
Neielsen//NetRatings

Trade Me spends a huge amount of effort on making the process of finding and buying as smooth as possible. It’s hard (sadly) in NZ to find an easier buying process, and with over 800,000 items on sale it is a continuous challenge to keep ‘finding’ easy.

Unfortunately the Trade Me buying $ metrics are not published, but I for one would tend to judge shopping sites on a simple ‘how much do they sell?’ basis.

The function of a *shopping* site must be to present the products as logically as possible so I can find what I want, make the purchase decision – then go through the checkout procedure to get them on their way… isn’t it?

I’d be interested in seeing the metrics of time on site vs. unique sessions vs. completed transactions as I think this would be a more valuable comparison for displaying the effectiveness of a shopping site. In terms of auction sites then perhaps page impressions and time on site are good measures to compare them with each other. Is there a way to capture the number of sucessful auctions vs. failed auctions vs. buyouts vs. offers accepted?

Unfortunatly, I can’t think of any simple way of capturing enough those stats (at least not in a site agnostic, consistent manner), given that you’d need broad buy-in across the ‘big’ shopping sites to make it a worthwhile comparision. Can anyone else suggest a better way to measure the sucess of shopping/auction sites?