In the course of a week-long benchmark study of portals around the world we ran recently for one of our clients, I stumbled upon the home page for the Yahoo! portal in spain.
Notice anything unusual? What fascinates me is how the top third of the page mentioned locations in no less than three countries. How? Why?
Near the top-left of the page, I can see an icon depicting a Canadian flag, along with a link to the Canada-specific Yahoo! portal. What this means is that the Yahoo! system knows that I am browsing from Canada (I have been, therefore, geolocated). If you view the site from another country, chances are you will be getting a different flag and a different link. And if you view it from Spain? I do not know. When I do follow the hyperlink to Yahoo! Canada, I get a similar link to Yahoo! Quebec, and when I go to Yahoo! Quebec, I get a link to Yahoo! Canada.
Interestingly enough, one of the most common elements of personalised user experience is missing here: language selection. There is no way I can read the news from Spain in English – or indeed, in Català or Gallego. Neither do I get a « Français » link on Yahoo! Canada, or « English » on Yahoo! Quebec.
Knowing where I am, the system could assume I necessarily want to view local content in my local language and therefore redirect me to the Canadian/English portal. Fortunately, it does not, and lets me browse news from Spain as much as I like, and limits the effect of geolocation to a subtle but clear hint that local content is also available.
How do they do it? Most likely, the IP address of the machine I used to access the site was compared against a GeoIP database such as this one or that one. Note that, unlike in many mobile applications, the IP-to-location translation is done automatically, without ever prompting the user for her agreement.
Next in our exploration of the personalisation elements is the page-wide bar of text links, calls for action and teaser questions serving a single purpose: convert the visitor into a user, either by registering, or by logging into an existing account.
Notice how the bar employs about about every trick know to web designers to entice us to log in: give regular users a quick link to their well-known personal space; showcase some of the added value gained through registration (preview of my inbox, directly in my news portal); comfort the potentially puzzled new visitor (New here?) and tease the social fiber (What are you doing? – obviously a go at copying the now-abandoned twitter tagline?).
The most surprising element comes last. How come, given that I am not logged into Yahoo!’s system, I am shown a weather widget for the city of Tokyo, Japan? I could have understood if it were Madrid (local to the portal, with an educated guess about which city I may be interested in) or Montréal (through geolocation) – but Why Tokyo? And how?
How do they do it? I quickly validated my initial hunch: Yahoo achieved this through the use of cookies – although other client storage options such as the localstorage in html5 would have been just as valid. My problem here, however, is that:
- The cookies are opaque – even with a cookie browser (most browsers with developer extension offers one) I would hardly be able to know which cookie was telling Yahoo! thay my preferred location was Tokyo. I had to delete cookies, one by one, until the weather widget changed to a geolocated default of Montreal.
- I cannot tell how and when this cookie was set in my system. Sure, I could set my browser to have me validate any cookie sites want to store, but given how cookies are entirely opaque, how would I know the meaning and consequences of allowing Yahoo (of flickr, or any site in the Yahoo ecosystem) to store a cookie that sets key
Tto…CAuNRMqKiCOsvekk….
And there’s more… All these modes of personalisation have been enabled without me even logging into the system. Imagine the many personalised features that can be implemented for recurrent, identified users: social filtering (content that my friends like, for instance), content adapted to my choices or patterns of use (recommendations for you), and much more. Has any content-heavy site ever tried to recommend their users items that they are unlikely to « like », or at least that they are likely to be surprised or challenged by?


I’m wondering since the participation to the Privacy Workshop how the transparency on how are used the cookies could be made. Sometimes a simple sentence could help users to decide the nature of the cookies. Not exactly how they are used, but what are their purpose.
For example, the one you have been described could be something along
description: « Geolocation information set by Acme Inc visiting this page http://example.com/I/see/you«
[...] This post was mentioned on Twitter by karl dubost, olivier Thereaux. olivier Thereaux said: How content sites personalize User Experience – the Yahoo case: http://lab.pheromone.ca/2010/07/19/yahoo-personalised-ux/ [...]
Karl, I thought this was impossible due to the fact that cookies are a mere key/value pair. A quick look at RFC2109 proved me wrong:
Can we see an example of good practice regarding cookies and geolocating using the services you mentioned?
I think people just don’t know how to do it right…