As someone involved in creating recommendation systems for web sites based on various user behavior signals, I can see adding a "what type of browser do they use" and "what type of OS do they use" intelligence in the future to further segment users.
Mac ownership must be a proxy for "household income". Brilliant.
Would be easy enough to compare prices on windows/linux in a vm on my mac though. And take note of the companies that do it. Something else to add to comparison services as well, aka does site xyz.com discriminate based on os/browser?
Important note: the prices are not different. Just that more 5-star hotels are shown in the search results to Mac users. If they were charging more for the same hotel, it would be evil not brilliant. :)
We tried both adjusting individual prices for Mac users (and didn't see any statistical difference from the price sensitivity for non-mac users, same as we didn't see statistically significant differences between older and newer Windows version users), and skewing the search result price range for Mac users compared to non-Mac users (and we certainly saw statistically significant improvements there in terms of average dollars per room-night). In the long run though, although we proved to ourselves it worked, the Mac userbase back then was so small that the overall bottom line effect wasn't big enough compared to the engineering and marketing effort to keep it working - I think it all vanished in the big 2004 rewrite of the site's back end.
It's much harder, but possible. You could look at browser features in JS and compare them to what's known for browsers to do and guess the browser. Some browsers are highly likely to be certain ODs.
It's highly highly likely that they are just looking at user agent strings.
Well, the code to look at JS features must be written in JS itself (adding features to an array, sending them over JSON, redirecting to a new page, etc.). I'll pull out my Dev Console and fuck with it (set a breakpoint and change that part of the code), and make them think I'm using lynx instead of Safari.
So, I don't think you can ever really know what browser your users use.
Sure. And even if you could, you couldn't really know whether they like cheap hotels based on it. The question is whether you can get a signal that's statistically well correlated enough to be useful.
Actually in this case you could infer they like cheap hotels (or don't like expensive ones). If someone is willing to go to a lot of effort to hide who they are, to hunt for bargains, or check they are getting a good price, they are probably very price sensitive, and would like the cheapest hotel you have, they've shown they are willing to waste their own time to get cheaper prices.
As someone involved in creating recommendation systems for web sites based on various user behavior signals, I can see adding a "what type of browser do they use" and "what type of OS do they use" intelligence in the future to further segment users.
Mac ownership must be a proxy for "household income". Brilliant.