Reasons for the fight between the two companies with browser shares

Currently, the Global browser market share is regularly published by StatCounter and NetApplications. Their statistics are based on the website server's browser access records. When a browser accesses a website, the header of the message sent by the browser to the server contains a statement about the browser version, which can also be forged). The server then identifies the browser that accesses the website. StatCounter and NetApplications obtain relevant statistics by deploying statistical tools to the server, and then process the data to reach a conclusion.

As we have repeatedly seen, the conclusions of the two market share statistics are always very different. According to StatCounter data, Chrome's market share in June was 32%, slightly higher than ie. According to NetApplications statistics, Chrome only has 19%, far less than 54% of ie.

Later, I am not going to discuss who is better about the two types of market share data. I just want to explain how the statistical differences are produced:

I. Sample differences

Currently, about 4 million of the websites where StatCounter deploys statistical tools and about 0.3 million of NetApplications are deployed. The statistical samples of StatCounter are much larger than those of NetApplications. Therefore, theoretically, this is more advantageous for StatCounter to obtain data closer to the overall situation. However, it is clear that 0.4 million of websites are already a large sample, which is enough to draw a relatively accurate conclusion. This is not an important reason for the huge gap between the two data.

Ii. Chrome pre-loading

The Chrome address bar uses similar technologies as Google instant search to record users' keys and pre-load the URLs that users may enter. This technology is used by Chrome 17. Obviously, if the traffic generated by the pre-loading occupies a large share of the total traffic, After Chrome 17 is launched, the Chrome browser's traffic share will be suddenly increased, but the fact is that there is no. In that month, Chrome's traffic share was only increased by 1%. Even if the 1% increase is all produced by the pre-loading, it will not have a great impact on the final conclusion. So this is not the reason for the huge difference between the two statistical calibers.

3. Traffic Adjustment

StatCounter statistics directly use the original traffic data monitored by the server. However, NetApplications re-adjusts traffic in countries and regions. NetApplications believes that due to the different preferences of Chinese people on websites and some regulatory factors, you know ), as a result, the traffic data monitored on these servers does not represent the actual situation. For example, the access records of a large number of Internet users on these websites are poor. How can this problem be solved? Weighted adjustment. For example, if the poor traffic monitored by a large Internet power on the server is increased by ten times and recorded to an ie or Chrome user, it is regarded as ten, so that the traffic it generates matches with the identity of a large country.

StatCounter and NetApplications deploy statistical tools on servers mainly in Europe and America. The user browser preferences recorded by servers are greatly affected by users in Europe and America, while Europe and America are also regions with high Chrome penetration rate. Therefore, weighted adjustment is beneficial to ie. After using the same method to adjust StatCounter data, ie's market share has increased by nearly 10 percentage points. It should be said that the statistical difference produced by this adjustment is very large and is one of the major sources of the huge data difference produced by the two statistical methods.

Iv. Market Share calculation criteria

Repeat the preceding statement. The StatCounter statistics directly use the original traffic data monitored by the server. StatCounter refers to the market share of traffic data generated by different browsers.

However, netapplication once again adopted a complicated adjustment method, which tends to use the number of users as the market share standard. One browser browses 100 Web pages within one day, and the other browser opens only one online banking page within one day. According to the idea of NetApplications, the market share of the two is the same user.

However, because the server cannot obtain a unique identification ID with high accuracy just like the browser version number. Therefore, the statistics of NetApplications still have complicated technical problems. How can we determine which access requests come from the same user, and which are not? In principle, there are only two feasible solutions. However, the two solutions are hard to say how accurate they are. One solution is to track cookies, and the website server writes traces to the browser cookies, so that when the browser accesses it again, it can know that it has been here based on cookies. But what if the user clears cookies midway through? Or, if the user has been enabling the private browsing mode to declare to the server that he does not accept cookies for writing, each access to the browser will generate a "new user ". By the way, Microsoft will enable the private browsing mode in ie10 by default. It is clear that this will create a large number of ie users in statistics based on cookies. Another solution other than cookies is Tracking ip addresses. The server records the ip source of each access page, and the same ip address serves as the same user. However, this solution also has obvious defects because it cannot identify the situation where multiple users share ip addresses, or when a user uses dynamic ip addresses. All in all, it is impossible to rely on website servers to strictly identify users on the statistical layer. Therefore, it is difficult to clarify the differences between the user volume data adjusted by NetApplications and the actual situation.

However, this data is different from the StatCounter data. After the difference between the weighted traffic adjustment based on country units is eliminated, the market share of ie has increased by 10% again. This at least indicates that the average webpage access volume of ie users is lower than that of Chrome. Or, more simply, users who often use browsers tend to prefer Chrome.