Welcome to WebmasterWorld Guest from 184.108.40.206
Forum Moderators: mademetop
I know cultural differences do weight but I would like to have global stats.
The thing with search engine market share is very very tricky and it is easy to misunderstand them. An engine can for instance have a top global market share, without being really important anywhere - simply by being number 5 or 6 or 7 or even lower almost everywhere in the world.
The absolutely best source is MMXI Media Metrix [mmxi.com], but they only cover 16 countries. Their full country reports are extremely expensive ($3500/country) but the top 10 sites are free and published on their site. We usually quote them in the European forum [webmasterworld.com] here at WebmasterWorld, where you can find the figures in each of the country discussions.
Additional material is added by members from local sources. This morning for example, Marcin published a lot of very fine statistical information on Poland.
sorry on the link stuart. there has to be atleast one blank line at the top of the post - its a legacy issue from the old board. but you can "linkize" any http with a back slash in front of it like this message. You can also use the url= style syntax (help file [webmasterworld.com])
Yes, that Mycomputer one is probably the best one going for stats. Everything else is really suspect. From search engine partnerships, to bad methodology, to cheating, there are some that are just wacko.
Sorry Rencke, I'm not a media matrix believer at all. They are the same ones who have done TV ratings in the US for decades. Some estimates put their tv ratings margin of error at 25%.
Like you will hear from time to time.
Company A: We received 2 million users this month
Media Metrix: Well our stats show that Company A only received 500,000 users for this month.
Has anyone thrown verified data into their faces showing how many users they(A) have had compared to Media's "research"?
1. Websnapshot is based on measuring software installed on webservers in English speaking countries. They do not measure traffic to other servers. This introduces a huge bias, as demonstrated by their domain figures: .com 50.8%, .net 40.9%, .edu 5.7% and the entire rest of the world with the majority of the online population - a mere 2.7%.
2. The selection method for these servers is not a random sample. Instead it is based on webmasters voluntarily joining the program, downloading and installing Websnapshot software. This method introduces a bias beyond all control. All we can be sure of is that the webmasters who elected to partcipate differ in some way from the webmasters that elected not to participate or who never found out about Websnapshot.
3. We know that only 32% of all surfers access the net in English. The rest use other languages. American sites, which consitute the overwhelming bulk of all sites measured - are only sparingly translated into foreign languages. Since they are measuring English sites only, it follows that visitors from non-English speaking countries are hugely underrepresented, as witnessed by the language settings in browsers as measured by Websnapshot:
[websnapshot.mycomputer.com...] These figures, have little to do with the composition of the world's online population, where North Americans are now in minority.
In short, Websnapshot uses an inexpensive method, which may or may not be reliable even for American sites, but lacks all relevance for the rest of the world and which therefore shows nothing about the situation globally.
By contrast, MMXI uses an expensive method, that conforms to good scientific practise.
A random sample of consumers is pulled from population registry in the 16 countries concerned, representing about 0.05% of the total population. Cash incentives are provided to induce these people to install measuring software on their computers and to answer questions about themselves, age, sex, education, income etc. The software measures the behaviour of these people and transmits it to the databases of MMXI or their subcontractors. (I happened to be pulled out of the Swedish population registry for their new Swedish panel and have gone through all the motions described.)
MMXI have demographics under complete control, and the bias introduced is the same as for any consumer report - people who refuse to respond and who might differ from the average in some way. In addition, people who fail to install the self installing software, due to inexperience with computers would be underrepresented. This cumbersome method is the reason why MMXI reports are so expensive.
Nielsen/Netratings use the same method as far as I can see from their site.