Cross posting to wikitech-l
2009/5/8 Ian Woollard ian.woollard@gmail.com:
Well, the wikipedia is big... google is big also.
One of the tricks google use is they try stuff out on victims... I mean users. They pick a small percentage of the internet and do something slightly different for them, and see if it works or not. The advantage of only doing it for a small number is that it means you can write the test with prototyping tools, rather than having to make it run fast, and you'll have a lot less complaints if it doesn't work very well. It also means you can do back-back comparisons stuff like: 'we tried this, and it showed a 23% improvement in referencing'.
I think if at all practical, the wikipedia needs to start doing stuff like that.
That sounds like an excellent idea to me. Is it achievable?
El 5/8/09 7:22 PM, Thomas Dalton escribió:
Cross posting to wikitech-l
2009/5/8 Ian Woollardian.woollard@gmail.com:
Well, the wikipedia is big... google is big also.
One of the tricks google use is they try stuff out on victims... I mean users. They pick a small percentage of the internet and do something slightly different for them, and see if it works or not. The advantage of only doing it for a small number is that it means you can write the test with prototyping tools, rather than having to make it run fast, and you'll have a lot less complaints if it doesn't work very well. It also means you can do back-back comparisons stuff like: 'we tried this, and it showed a 23% improvement in referencing'.
I think if at all practical, the wikipedia needs to start doing stuff like that.
That sounds like an excellent idea to me. Is it achievable?
Such A/B testing is planned for various things where feasible. Some things will be easier to do this way than others. :)
-- brion
Brion Vibber brion@wikimedia.org writes:
One of the tricks google use is they try stuff out on victims... I I think if at all practical, the wikipedia needs to start doing stuff like that.
That sounds like an excellent idea to me. Is it achievable?
Such A/B testing is planned for various things where feasible.
Speaking of Google tricks:
A friend of mine in his 80's knows not to click on the sponsored links, the right column of Google search results.
But when Google occasionally takes one of those sponsored links from its right column, and slips it in to its left column legitimate search results, with only background tint and fine print to distinguish it, well, the rest is http://groups.google.com/groups/search?as_umsgid=87eivk5lum.fsf%40jidanni.or...
Thomas Dalton wrote:
That sounds like an excellent idea to me. Is it achievable?
It would need to be properly advertised to the relevant guys at the wiki, so when the different configuration breaks for an user, people at the community don't spend hours wiht an irreproducible behavior.
wikitech-l@lists.wikimedia.org