Andrew Gray:

There's a couple of estimates on: http://en.wikipedia.org/wiki/Global_Internet_usage though they look a little dated.

Alternatively, users by country is reasonably well estimated, I think, and you could try estimating based on languages from that.

I do agree that numbers of editors per million internet users (not internet connections, see below) would be an important metric. It would also be more within our sphere of influence than the current participation metric (but not closer to the ultimate goal that we strive for, as advocated by our mission statement). However it seems to me that today lots of practicalities stand in the way to make this a dependable metric.

Can I first try to explain my point by analogy? On road signs and maps the most used metric by far is distance in miles or kilometers. For planning a car ride it would be more useful to know average travel times, taking into account number of traffic lights (for city trips) and current traffic congestion (and if that is not known: time of the day, day of the week, and average traffic densities on this route) etc. In fact electronic travel planners can do this, using super precise electronic maps and even a real time data feeds about changing conditions. Several years ago when this was not possible pure distances by main routes were an incomplete, sometimes hard to interpret, but despite shortcomings unambiguous and trustable piece of information.

IMHO the same applies to participation stats: given the current level of available information we better use a metric that is not ideal, not extremely sophisticated, but reasonably solid, even when there is serious interpretation needed to put it to effect.

http://en.wikipedia.org/wiki/Global_Internet_usage has two tables: The first is outdated, 5 years count for a lot in web stats, especially in parts of world where internet connectivity is sharply on the rise. (BTW it mentions for Chinese speakers 874 million instead of 1.3 billion). The second only mentions 10 languages where we need information for 280 languages.

I am not saying proper and sufficient data are not available, but frankly despite serious attempts I have not found it online and publicly available. Most connectivity stats are per country, not per language. One would also need good stats on distribution of languages per country, assume equal distribution of connectivity over language groups (in many countries, language spoken and social-economic status are not independent). For parts of the world where census data are less dependable or outdated this would put estimation error on estimation error.

Another complication is that e.g. in China and large parts of Africa (and India?) most web activities happen in internet cafes and on other shared computers. In those countries the number of accounts, and the number of web connected people can differ dramatically (even in western countries there is no simple relation, many people have access to several computers, at home and at work). Also should we factor in broadband vs modem connections? How much can a person contribute on 10-20 seconds response times?

So, in short, yes I see the limitations of the new metric, but it does draw attention to a significant part of the story of where we are in relation to our mission statement. And hopefully we can build on this and refine our metrics when global, dependable, up to date and non skewed data for internet usage per language become available.

Erik Zachte