James Gardner, Head of Innovation and Research at Lloyds TSB was recently wondering:
it is users themselves, everywhere else, that are deciding not only how things are personalised, but also what content should be shown in the personalised context. We can see where social finance is going when examining Mint, Wesabe and others of their ilk.
So,
sitting around the table listening about "personalisation", I couldn't
help but wonder if we're somehow stuck in a financial services time
warp. Or at least, why the institutional part of the market is.
.. to which I replied:
I think it's a mix of the two -- that personalisation done right draws from both
- software-initiated (implicit personalisation), and
- user-initiated (explicit personalisation).
Amazon is the poster child of software-initiated customisation, aka
'collaborative filtering' aka 'recommendation-engine driven
personalisation'. All the features and items are derived entirely
implicitly from the user's actions. Last.fm is another great example --
working secretly in the background as you listen to music to figure out
what you might like. They're both screaming successes so I'd not want
to deemphasise this aspect.
Netvibes could, correspondingly be seen as the poster child of
user-initiated customisation. Netvibes has spawned well, an entire
class of web sites now called 'start pages'. This concept is Also a
great success, particularly in the intranet space.
Both work well, for solving certain things, and I think they're mutually supporting.
What's particularly fascinating in the social computing space is how
it spans both too -- recommendations can be based on your own past
behaviour ('the page you made' in Amazon) while also based similar
patterns with other users. User-initiated customisation at the social
level is a bit more tenuous but is starting to take shape shape in the
form of a kind of 'group evolution' -- Yahoo Pipes! for instance allows
cloning of pipes so you can make a better pipe. You've customised your
experience, but it's based on what someone else did.
So why isn't implicit personalisation more popular then? When the
rubber hits the road, it involves some pretty tricky problems that
require some great business intelligence, which is often hard in an
enterprise. First base being single view of customer (SVC) of course,
or least what I coin 'incrementally better view of customer' ('IBVC').
As an aside into what's involved for recommendation engines, check
out a precis of Google New's recommendation engine at on my personal
blog.