Data provided by StatCounter can't be directly compared to data from Net Applications
Chrome is either just about to overtake IE to become the most popular browser in the world or has been losing users every month for the past three months. Both points of view are backed up by numbers, but they can't both be true. The optimistic view comes from StatCounter, the dire one from Net Applications.The two analytics companies have never reported identical numbers, but at least they agreed on the general trends.
That is no longer the case. The difference is notable but is easily explained by the different methodologies employed by the two firms.
The biggest factor is that Net Applications weights the data it gets from the sites it monitors, some 40,000 of them. For example, if the tracked sites aren't particularly popular in China, visits from the country would only make up a small portion of the total visits it counts.
This means that, even though it's the largest internet market in the world, China would not influence the monthly usage stats much.
So, even though IE is still very popular in China, due to a number of things, the biggest being the fact that most people are first time computer owners and many people rarely go online even if they do have a computer, this would not affect the global stats too much.
Weighting means accounting for the discrepancy by using market size numbers - how many people use the internet in any given country - provided by the CIA. This way, even if you only see a small number of visits from China, they'll weigh more than visits from the UK, for example.
The problem with this is that the market size numbers for different countries may be from different years and may be outdated which is a problem especially in fast growing markets like China. Weighing adds almost as many problems as it fixes.
StatCounter does not use any weighting and also looks at visits, not unique daily visitors, like Net Applications. This produces pure usage numbers, but they're only representative to the sites StatCounter monitors, more than three million. Turkey for example is a big influence on StatCounter numbers, but a much smaller one on Net Applications'.
Both approaches have advantages and disadvantages. Microsoft prefers Net Applications, for obvious reasons, but you're free to make your own informed decision. Both tools provide useful information, in any case.