Skip navigation

Building Signal vs. Filtering Noise

Marshall Kirkpatrick over at Read Write Web is on to something again.

“In a world more swamped with content options every day, recommendation technology is poised to make a huge difference in our experience online.”

The part which caught my attention is that recommending is very different than the filtering. And I happen to think it’s much better.

Turn it up, it’s loud in here

Lots of folks people think that there are too many silos each containing little bits of information.  The first impulse is to build an aggregator to collect and then spew out all those updates from one central domain.  (We tried too, with our erstwhile feed reader.)

Whoops. Suddenly, there are too many updates, there is too much noise and people start clamoring for a filter. What happened?  Clay has the snappy title of “It’s not Information overload, it’s Filter Failure” which seems to prove itself, with 40,000 references already. 

More data doesn’t always win

I think of recommending as subtly different than filtering.

Recommendation done well makes me think about selecting and building up salient data components to arrive at a useful conclusion.  Filtering makes me think of starting with too much crap and trying to figure out what to throw away.  

I’m going to posit that attempting to filter feeds or streams will have limited widespread success, because in the act of aggregation, the filter loses the original context which contained so much signal. Parked domains can look like perfect matches to search engines, too.

On the other hand, technology that does a good job recommending will feel like it’s building signal instead of taking away noise. It will retain the context.

What might this magic recommendation technology look like? 

Consider Amiad Solomon’s keynote suggesting web 3.0 will offer “detailed data exchange to every point on the internet, a ‘machine in the middle’ with

1) Smart internetworking:  to analyze collective online behavior

2) Communication via api to a single point of reference

3) Distributed databases to search millions of nodes and scan billions of records at once.

Marshall simply says: “Bring on a smart future augmented by powerful recommendation technology!”

I couldn’t agree more.

In fact, I think these gentlemen just may be mind readers. We’ll see.


Post a Comment

You must be logged in to post a comment.
%d bloggers like this: