Thriving by Dis-Automation - ' The Difference Between Worker '

Knowing the Difference Between Worker and Ware

Savage Beast's deliverable is specific to one field, but the basic function is essential to a global economy that requires consumer choice without limits.

The deliverable is music discovery—the ability to find new music consumers will enjoy (and buy) based on what they've liked in the past.

The means of doing this are not the myriad automated-pattern recognition processes other music services use, though some are surprisingly good.

It's not collaborative filtering through affinity networks mapped by computers (If Fred Flintstone loves A and B but hates C, and you love A but hate C, you get a recommendation for B).

According to founder Tim Westergren, Savage Beast uses a team of trained music analysts to describe each song using a set of attributes that make up what Savage Beast calls the Music Genome Project.

Just as the human genome project maps genes and gene combinations against the characteristics that appear when they're expressed, this music genome project attempts to map core musical forms and identify them in specific songs.

Customers define what they like, and software scans the entire genome, identifying songs to suggest by matching the pattern of a customer's likes with the pattern in a set of more than 400 attributes assigned by a human analyst at Savage Beast.

The resulting matches range across styles and genres, turning "categorization" on its head because instead of refining choice down in ever-narrowing hyper-optimizations, it offers a listener an opportunity to grow into other genres with a few songs they have a high probability of liking but would not encounter if they stuck to familiar genres.

There are three things that make Westergren's analysts functional, only the last of which might someday be sensibly automated: formal music theory background ("at least a four-year degree"), corporate training ("so that everyone identifies the same way"), and what Westergren calls a good ear.

Once analysts start tagging each cut, they have to labor in physical proximity with other analysts to build uniformity of decision-making and attribute assignment.

Once identified, the knowledge gets committed to a database, and that's where the technology comes in.

Savage Beast customers can plumb a deep database and get quick, informed suggestions that match their musical proclivities.

How did Savage Beast get to this very uncommon, anthrocentric, approach? Westergren says the company's founding CTO Will Glaser (who is no longer with the firm) made it clear that company founders needed to define what they wanted as an end product first, and then allow the team to build towards it. That's an old model, but a highly effective one.

The Genome Project became the bridge between what exists (hundreds of thousands of musical cuts) and what was sought (a way to use past preferences to explore for currently attractive choices).

The Westergren/Glaser model for high-impact projects demands building to a defined goal, and applying humans to do what technology can't do well while applying technology for what it does well.

How does this affect you? It could give you a much more effective way not only to organize your projects, but to put the technology you already use to a much more effective use than you did in the past.

In the next column, I'll suggest ways you can think about the model in your own line of work.

Jeff Angus is a management consultant and has been working with IT since 1974. He has held IT management positions in user interface design, marketing, operations and testing/analysis. Look for his book, "Management by Baseball: A Pocket Reader." Jeff's columns have appeared in The New York Times, The Washington Post, the St. Louis Post-Dispatch and the Baltimore Sun.

Check out eWEEK.com's for the latest news, reviews and analysis on IT management from CIO Insight.