The problem, as I've described on my blog regarding the Nielson/Netratings proposed metric, is that it can be artificially inflated very easily with no clearly measurable margins of error (http://agentultra.com/?p=70).
The great thing about page views is that we can parse out (with a level of certainty) which views were accidental browser refreshes and so forth.
There's also potential for gaming such a system that Nielson/Netratings proposes. One just has to look at Second Life and the problem it has with the "time spent" model. It spawned the "camping chair" where land owners would pay out a minimal fee to people who would hang out on their land.
The trick is to use multiple sources and to be good at statistical mathematics. :)
I'm just a bit queasy about time spent model. With modern browsers offering advanced features like tab based browsing et al, a user like me just prefers to open a new tab instead of navigating away from the page. Which does not capture the time spent correctly.
Also, lot of content consumption in blogs etc is through RSS, and which again means that amount of time spent is equivalent to the time taken to leech the feed xml, which again is false variable as i spend reading them in thunderbird for long.
I agree to one point that different websites serve different purpose and cannot be classified with a single performance model. I am working on to get some modeling done for my favorite blog reader for filtering feeds.
Ooo. Nice one. Something for my mind to feed on and try and analyze my traffic. Quite a few things have happened to my art blog in terms of traffic, and most of the times, I can attribute a change to something. But sometimes, it just leaves me scratching my head.
As you explain here, I realized that these visitors are sentient beings and go about a myriad ways. I need to model with more details (than just correlations) to understand them. (Special thanks for pointing to Sankey diagram :) )
The time model can be flawed, as mentioned by ashish, but I don't believe in just measuring pageviews. Unfortunately, I have to use multiple analytic services to get all the stats I want. This just bloats up my weblog with (seemingly unnecessary) javascript.
Beautiful diagrams by the way, done in illustrator i assume?
Liked your post and the idea of using Sankey diagrams to display website visits.
Although originally only for showing energy efficiency, Sankey diagrams today are used to display any kind of "flows": materials, energy, cash, cars, people, ...
BTW: I use a special Sankey diagram tool, so when changing the flow quantities (here: website visitor numbers) the magnitude of the flow will adapt automatically.
< Back to 'The Varying Virtues of Site Performance Metrics'
10 comments:
J Kenneth King wrote:
The problem, as I've described on my blog regarding the Nielson/Netratings proposed metric, is that it can be artificially inflated very easily with no clearly measurable margins of error (http://agentultra.com/?p=70).
The great thing about page views is that we can parse out (with a level of certainty) which views were accidental browser refreshes and so forth.
There's also potential for gaming such a system that Nielson/Netratings proposes. One just has to look at Second Life and the problem it has with the "time spent" model. It spawned the "camping chair" where land owners would pay out a minimal fee to people who would hang out on their land.
The trick is to use multiple sources and to be good at statistical mathematics. :)
Pete wrote:
Good article. I try to explain this to all my clients and a number of my colleagues on a regular basis. I think you just made my job easier :)
On a side note, you spelt 'acheived' incorrectly in your Sankey diagram.
:P
Stuart Brown wrote:
Bother. I think you're right. Serves me right for composing diagrams so late!
Stuart Brown wrote:
Fixed. And just in time for the August colours!
Ashish wrote:
I'm just a bit queasy about time spent model. With modern browsers offering advanced features like tab based browsing et al, a user like me just prefers to open a new tab instead of navigating away from the page. Which does not capture the time spent correctly.
Also, lot of content consumption in blogs etc is through RSS, and which again means that amount of time spent is equivalent to the time taken to leech the feed xml, which again is false variable as i spend reading them in thunderbird for long.
I agree to one point that different websites serve different purpose and cannot be classified with a single performance model. I am working on to get some modeling done for my favorite blog reader for filtering feeds.
Loved your earlier posts as well, keep it up!
Todd wrote:
what with broadband speeds on the increase, does the speed of your site even matter as much anymore (apart from the extremes obviously)?
Sometimes, I feel people stall for too long on how to increase the performance on their site.
Vyoma wrote:
Ooo. Nice one. Something for my mind to feed on and try and analyze my traffic. Quite a few things have happened to my art blog in terms of traffic, and most of the times, I can attribute a change to something. But sometimes, it just leaves me scratching my head.
As you explain here, I realized that these visitors are sentient beings and go about a myriad ways. I need to model with more details (than just correlations) to understand them. (Special thanks for pointing to Sankey diagram :) )
Daniel Swiecki wrote:
Another good post.
The time model can be flawed, as mentioned by ashish, but I don't believe in just measuring pageviews. Unfortunately, I have to use multiple analytic services to get all the stats I want. This just bloats up my weblog with (seemingly unnecessary) javascript.
Beautiful diagrams by the way, done in illustrator i assume?
Stuart Brown wrote:
Yep, Illustrator's my tool of choice these days.
phineas wrote:
Liked your post and the idea of using Sankey diagrams to display website visits.
Although originally only for showing energy efficiency, Sankey diagrams today are used to display any kind of "flows": materials, energy, cash, cars, people, ...
I made two variations of your diagram on my blog: http://www.sankey-diagrams.com/using-sankey-diagrams-for-visualizing-web-site-performance/
BTW: I use a special Sankey diagram tool, so when changing the flow quantities (here: website visitor numbers) the magnitude of the flow will adapt automatically.
Comments are closed.
< Back to 'The Varying Virtues of Site Performance Metrics'