Percentage of traffic from Google

**lifeplayer** · 6 June, 2010, 15:47 PM

[quote name='David' date='06 June 2010 - 10:36 AM' timestamp='1275820569' post='16965']

10 samples is nothing. Say if you were finding out the ethnicity of the population in USA and you get 10 random people. 8/10 you choose are Asian. Does that mean 80% of the population are Asian? I wouldn't think so. They are same in figures, but have different meanings when it comes to the fairness of statistical variances.

[/quote]

If you think 10 can not present the actual, then how many are enough to represent the actual? 10000? 20000? 30000?

You can never get the actual, if you dont believed, mark down your current status

You are getting 20000 out of 40000, wait until you get 80000 traffic and see it will be exact 40000 from google search

**David_** · 7 June, 2010, 07:23 AM

[quote name='lifeplayer' date='07 June 2010 - 02:47 AM' timestamp='1275839269' post='16978']

If you think 10 can not present the actual, then how many are enough to represent the actual? 10000? 20000? 30000?

You can never get the actual, if you dont believed, mark down your current status

You are getting 20000 out of 40000, wait until you get 80000 traffic and see it will be exact 40000 from google search

[/quote]

You are not speaking reality. 10 samples will not help anyone. Its a fact. How can you judge a site's power if you have 10 hits, as opposed to Microsoft's billion hits? If Microsoft has 30,000,000,000/70,000,000,000 hits (43%) are from Google and Admin Community has 5,000/10,000 hits (50%) from Google, does that mean AC is better than Microsoft in terms of Google traffic success? The answer, in anyone's right mind, is no.

A fair trial is if everyone compares hits at the same time, with the same software. It can never be fair, so a fair amount would be at least 100,000 hits/pageviews (of course, you should consider using Calculus to determine which site has the higher success.)

**lifeplayer** · 7 June, 2010, 09:01 AM

[quote name='David' date='07 June 2010 - 07:23 AM' timestamp='1275895427' post='17011']

You are not speaking reality. 10 samples will not help anyone. Its a fact. How can you judge a site's power if you have 10 hits, as opposed to Microsoft's billion hits? The answer, in anyone's right mind, is no.

A fair trial is if everyone compares hits at the same time, with the same software. It can never be fair, so a fair amount would be at least 100,000 hits/pageviews (of course, you should consider using Calculus to determine which site has the higher success.)

[/quote]

Let answer your question

Question: If Microsoft has 30,000,000,000/70,000,000,000 hits (43%) are from Google and Admin Community has 5,000/10,000 hits (50%) from Google, does that mean AC is better than Microsoft in terms of Google traffic success?

Answer: The percentage is not for comparing, it show that which source are provide more traffic to your own site. Assume that

Microsoft has 30,000,000,000/70,000,000,000 google hits (43%), 30,000,000,000/70,000,000,000 Bing hits (43%) and 10,000,000,000/70,000,000,000 direct hits (14%). And ABC.com have 5/10 google hits, 3/10 bing hits and 2/10 direct traffic. Then the conclusion is ABC.com is doing better on getting google traffic "INTERNAL" compare to Microsoft. That's nothing related to the "success" or "more" since the "success" is directly refer to the figure, not the percentage.

By the way, not going to talk about the sample size is whether enough to provide the "good result" since it involve in too many statistical calculation, but i can tell you for the DOE, 10 samples can be enough

**David_** · 7 June, 2010, 11:45 AM

[quote name='lifeplayer' date='07 June 2010 - 08:01 PM' timestamp='1275901305' post='17012']

Let answer your question

Question: If Microsoft has 30,000,000,000/70,000,000,000 hits (43%) are from Google and Admin Community has 5,000/10,000 hits (50%) from Google, does that mean AC is better than Microsoft in terms of Google traffic success?

Answer: The percentage is not for comparing, it show that which source are provide more traffic to your own site. Assume that

Microsoft has 30,000,000,000/70,000,000,000 google hits (43%), 30,000,000,000/70,000,000,000 Bing hits (43%) and 10,000,000,000/70,000,000,000 direct hits (14%). And ABC.com have 5/10 google hits, 3/10 bing hits and 2/10 direct traffic. Then the conclusion is ABC.com is doing better on getting google traffic "INTERNAL" compare to Microsoft. That's nothing related to the "success" or "more" since the "success" is directly refer to the figure, not the percentage.

By the way, not going to talk about the sample size is whether enough to provide the "good result" since it involve in too many statistical calculation, but i can tell you for the DOE, 10 samples can be enough

[/quote]

I used the percentages to make things look easier

Microsoft will always do better than 99.99% of all other sites (top 100). Traffic wise, Microsoft will always be more successful no matter what. Its not the percentage that matters, its the amount of hits. If you have 5 hits and I have 10, then I've done better than you. If you put it this way, you would prefer 8/10 total hits from Google rather than 1000/5000 from Google. I'd prefer the 1000/5000 hits as I have more hits than you.

As for your statistics comment, 10 is nothing. Please confirm this: I select 10 random people to represent USA. 8/10 I find are of an Asian background. 1/10 is from a caucasian (white) background and 1/10 is from an African background. Does this sample reflect reality? The answer is NO, unless you can confirm this. I'd say about 7/10 are white, 2/10 are African American and 1/10 are other. But with your 10 samples, you find something different, so that clearly means, in your mind, that USA is 80% Asian.

**lifeplayer** · 8 June, 2010, 02:12 AM

[quote name='David' date='07 June 2010 - 11:45 AM' timestamp='1275911158' post='17017']

I used the percentages to make things look easier

Microsoft will always do better than 99.99% of all other sites (top 100). Traffic wise, Microsoft will always be more successful no matter what. Its not the percentage that matters, its the amount of hits. If you have 5 hits and I have 10, then I've done better than you. If you put it this way, you would prefer 8/10 total hits from Google rather than 1000/5000 from Google. I'd prefer the 1000/5000 hits as I have more hits than you.

As for your statistics comment, 10 is nothing. Please confirm this: I select 10 random people to represent USA. 8/10 I find are of an Asian background. 1/10 is from a caucasian (white) background and 1/10 is from an African background. Does this sample reflect reality? The answer is NO, unless you can confirm this. I'd say about 7/10 are white, 2/10 are African American and 1/10 are other. But with your 10 samples, you find something different, so that clearly means, in your mind, that USA is 80% Asian.

[/quote]

Seems it is hard for me to explain to you according to sample size but it doesnt matter since many people dont actually understand about this. By the way, you might visit http://www.surveysystem.com/sscalc.htm, you can find out how much population that 10 sample can represent. I dont mean 10 sample can represent all , it depends to the population. And you dont really need to go or all the US member to know the precentage of the races

**David_** · 8 June, 2010, 07:30 AM

[quote name='lifeplayer' date='08 June 2010 - 01:12 PM' timestamp='1275963127' post='17051']

Seems it is hard for me to explain to you according to sample size but it doesnt matter since many people dont actually understand about this. By the way, you might visit http://www.surveysystem.com/sscalc.htm, you can find out how much population that 10 sample can represent. I dont mean 10 sample can represent all , it depends to the population. And you dont really need to go or all the US member to know the precentage of the races

[/quote]

The site barely has credentials, nor is it a educational website, or established organization. Thus it has very little reliability and validity.

As for the bold, you just proved yourself wrong. Just a few posts up, you said that 10 was enough. Now you say that 10 isn't always right, and that you can have a certain amount to determine the ethnicity of a country. Also, I never said we needed the whole population to determine validity, I clearly said 10 is nothing. (BTW, I do the highest level of mathematics in Australia. Statistics is my domain.)

**lifeplayer** · 8 June, 2010, 11:00 AM

[quote name='David' date='08 June 2010 - 07:30 AM' timestamp='1275982246' post='17055']

The site barely has credentials, nor is it a educational website, or established organization. Thus it has very little reliability and validity.

As for the bold, you just proved yourself wrong. Just a few posts up, you said that 10 was enough. Now you say that 10 isn't always right, and that you can have a certain amount to determine the ethnicity of a country. Also, I never said we needed the whole population to determine validity, I clearly said 10 is nothing. (BTW, I do the highest level of mathematics in Australia. Statistics is my domain.)

[/quote]

Then you shall understand all about this, this is very common used, 10 sample is enough for you to judge for million. And you shall very clear about this since the statistic can be understand by fresh college level. And great to meet another statistic guy here and i was major in statistical too, already complete my master many years ago.

**David_** · 8 June, 2010, 13:54 PM

[quote name='lifeplayer' date='08 June 2010 - 10:00 PM' timestamp='1275994817' post='17058']

Then you shall understand all about this, this is very common used, 10 sample is enough for you to judge for million. And you shall very clear about this since the statistic can be understand by fresh college level. And great to meet another statistic guy here and i was major in statistical too, already complete my master many years ago.

[/quote]

Even if 10 is enough, 100 would show a fairer amount, 1000 would show an even more fairer amount. The closer you can get to the target, the more fairer. In Australia, its about 10 per 50,000 people, ie. they ask around 5000 people to do surveys etc. If it was 10 per million people, only around 200 people would be selected. 200 people to represent the whole Australia? Sounds absurd. You may want to keep up with today's new methods. Maybe 10 years ago it was 10 samples, now its a whole lot more.

**lifeplayer** · 10 June, 2010, 07:05 AM

[quote name='David' date='08 June 2010 - 01:54 PM' timestamp='1276005251' post='17060']

Even if 10 is enough, 100 would show a fairer amount, 1000 would show an even more fairer amount. The closer you can get to the target, the more fairer. In Australia, its about 10 per 50,000 people, ie. they ask around 5000 people to do surveys etc. If it was 10 per million people, only around 200 people would be selected. 200 people to represent the whole Australia? Sounds absurd. You may want to keep up with today's new methods. Maybe 10 years ago it was 10 samples, now its a whole lot more.

[/quote]

Nothing is 100%, and if 10 millions people in Aus, the only thing is collect all the 10millions sample to give you the high confident level. However, during the data collecting, there have some changing (new born or die) which cant give you 100% accurate result . And therefore, statistical was developed and being common used especially for the variable.

For example, there are millions finished good to check and it is hard to spend lots of the, it might need to invest high cost on full checking. Therefore, the organization will use "sampling" which same as the survey company. All is about the confident level, and this is what google using on checking the adsense click too, that's why they use "suspect" in the email when ban the member which mean the clicks happen overall is lower than the confident level.

And yes, some of the adsense banned member can be innocent, but the probability is low, which is depend on the confident level.

**David_** · 10 June, 2010, 07:14 AM

[quote name='lifeplayer' date='10 June 2010 - 06:05 PM' timestamp='1276153514' post='17160']

Nothing is 100%, and if 10 millions people in Aus, the only thing is collect all the 10millions sample to give you the high confident level. However, during the data collecting, there have some changing (new born or die) which cant give you 100% accurate result . And therefore, statistical was developed and being common used especially for the variable.

For example, there are millions finished good to check and it is hard to spend lots of the, it might need to invest high cost on full checking. Therefore, the organization will use "sampling" which same as the survey company. All is about the confident level, and this is what google using on checking the adsense click too, that's why they use "suspect" in the email when ban the member which mean the clicks happen overall is lower than the confident level.

And yes, some of the adsense banned member can be innocent, but the probability is low, which is depend on the confident level.

[/quote]

I know, I gave a realistic sampling example in my previous post, which has logics.