Could Google’s browser statistics replace Net Application and StatCounter?

Mozilla's Asa Dotzler wants Google to start providing browser usage statistics again.

I completely agree.

Asa also thinks that Google's statistics will be more globally representative than what we have to day.

That, I don't agree with. …

Now, Asa has recently come to the same conclusion I came to some time ago: Current browser statistics are completely unreliable. But just because what we have today is flawed doesn't mean that we should replace it with something which could be equally flawed.

Maybe browser statistics from Google would be more representative than what we have today, but the problem is that there's simply no way to know. As a matter of fact, I can think of at least a couple of huge sources of error of the top of my head.

For example, Google has been pushing browsers like Firefox and Chrome heavily through their sites, which means that users of Google services are much more likely to use those browsers.

Google has also been showing "your browser is not supported" messages at times, which obviously means that people will either leave the service, or switch to a supported browser. In other words, Google can, to a certain degree, decide what its own browser statistic should look like.

Another problem is that Google really isn't that huge in all markets. Baidu seems to be dominant in China, Yandex is not far behind Google in Russia, and so on. This means that markets where the browser landscape might look completely different from Google's main markets will be seriously undercounted.

These issues will be in addition to the general problems with all browser statistics, such as different caching in different browsers, "User-Agent" string masking, and so on.

So while I would definitely welcome browser statistics from Google, if only out of curiosity, I fear that these statistics wouldbe misinterpreted the way current sources of browser stats like StatCounter and Net Applications are, as actually having a verified representative sample of the global internet population.

One step forward, and one step back again. That won't get people anywhere 🙂

Advertisements

12 thoughts on “Could Google’s browser statistics replace Net Application and StatCounter?

  1. Originally posted by haavard:

    One step forward, and one step back again. That won't get people anywhere.

    +1 🙂

  2. sometimes i think if google is a monopolist as a search engine in most markets, doesn't that oblige them legally to limit their advertisements to their products to certain limits ? like MS

  3. Google also has google analytics, which doesn't have the shortcomings of the google services. When the stats are only generated by google analytics, I think you get pretty clear stats. Because you don't have the "try youtube in Chrome" etc. stuff. Sure, still some sites promote a browser, but this isn't different for the stats generated by statscounter or netapplicatioins. And because of the scale of google analytics I think the user base would be much bigger then with the "traditional" stats generators.

  4. Google Analytics will suffer from the same problems as StatCounter and Net Applications: The statistics will be based on sites that choose to use the service. A non-representative sample (as Asa showed in his blog post).

  5. Forgot to note, of course we'll never have anything perfect. I'm looking for something _better_ and that's all.

  6. Haavard, what did you think about my Wikipedia suggestion. They've got billions of visits and are translated into lots of languages. They are starting to publish stats and don't have the downside of promoting a specific browser. I can't think of a bigger site other than search services.Note that I did suggest Google + Baidu + Yandex or similar to help cover markets where Google wasn't dominant.

  7. Originally posted by AsaDotzler:

    Note that I did suggest Google + Baidu + Yandex

    Hi Asa,This would still need engines such as Bing, Yahoo, and other engines like Ask on board. However one may not be able to trust the stats from certain search engines because they may endorse their own product. Also they may not want to collect this information on their own because it would be time consuming.Also another variable is some people do not always use a search engine. I use the address bar and speed dial therefore I do not search as much as an average person. So the usage of my browser may not be accounted for as much as the next person. We would have to also rely on normal sites to collect this information also for stats to be accurate.It seems your proposition about search engines independently collecting statistics then sharing it with others to create one stat would be difficult and it wouldn't give us an accurate view because people do not spend most of their time on searching but rather on other sites e.g. Facebook.Originally posted by Asa from his blog:

    … Or three, we could find one source that had solid global representation.

    +1…Originally posted by AsaDotzler:

    I'm looking for something _better_ and that's all.

    We could develop a new stat counter together. For this to be possible this would require a new organization formation with a representative from each party that would develop it, this way people would not think the results are skewed. Of course this would record the browser statistics (along with other statistics) on sites (not just search engines) such as Facebook, Google, Yandex, etc. This could be widely implemented across the web just like bugzilla is for bug reporting. This would begin to gain implementation across sites via endorsement of corporations and organizations that created this stat counter. (Opera Software, Mozilla, Google, etc.) Also this could be non-profit so users would get advanced usage information of their site for free instead of paying for it (this would gain the counter more popularity). Or we could stick with results from Net Applications or Stat Counter. 😦

  8. Stats for Wikipedia are obviously very interesting too, but once again you get a selected audience and not necessarily a representative sample.We need something better than today, but is there anything better? Replacing poor statistics with other poor statistics doesn't really get us anywhere.

  9. I think between Asa and Daniel there is the kernel of a great idea. Here's my take.1. Develop an open, standard means of reporting browser statistics, whether it be an XML format or a JSON format or something else. All participating statistics gatherers should use that format. It should include as much demographic information as possible and still remain anonymous. The demographic info should include visitors' information as well as the sites' information. So a Russian site getting lots of North American visitors can be determined.2. Provide a central reporting site that all reporters push their collected statistics to. A reporter should push relatively infrequently (no more often than daily) to help keep the central site from becoming overwhelmed.3. The central site then collates all the information and publishes it regularly via RSS/Atom and some open API.The central site, protcols, and APIs should be as open as possible and controlled by a non-profit in order to prevent monopolization of control by any one browser maker.All browser makers and statistics reporting services should be invited to participate (sit on the board, whatever; I'm not great at organizing 🙂 ). This should promote maximum participation and keep things as fair as possible because everyone has a say.Once the information is gathered, processing it would be left to third parties (e.g. the current statistics reporters) to present it in a usable form with pretty graphs and such. Also, open source software could be created that polls the service and creates reports for individuals. For instance, I could see a plugin for WordPress comparing your stats to the central service's global stats. And interested parties could find out how many North American visitors to Russian sites use Opera or Firefox or whatever.I think part of the key will be making the raw information available. Also important is to gather as much raw information as possible.I certainly can't work on this…but I wish I could. It would be a great and fun project, I think. The toughest part, I think, is getting the various parties to participate.

  10. Originally posted by haavard:

    Yandex is not far behind Google in Russia

    I would say it is even vice-versa. :)Google is (pretty far) behind Yandex in Russia.Originally posted by haavard:

    We need something better than today, but is there anything better? Replacing poor statistics with other poor statistics doesn't really get us anywhere.

    The more we have independent sources for statistics, the more chances that we can see whole picture.There probably should be multiple source statistics aggregators which can show composite picture of global or regional statistics.

  11. pitredbeard interesting idea as most sites already do a browser sniffer of some sort, why not send that data to a point (that is a neutral point) to can then see what the browser sniffer have picked up on different parts of the web in different areasIn Russia it shows that the same site is not the most popular in every country so it is not a easy way to get good results of what browser people use as they may not all go to the most popular site in a country

Comments are closed.