Managing Last.FM's "Mountain of Data"

Rob Spengler writes "Last.FM co-founder Richard Jones says the biggest asset the company owns is 'hundreds of terabytes of user data.' Jones adds, '... playing with that data is one of the most fun things about working at the company.' Last.FM, for those who have been living on Mars for the last two years, is the largest online radio outlet, with millions of listeners per day. The company surpassed Pandora and others largely due to its unique datamining features: 'Audioscrobbler,' the company's song/artist naming algorithm, can correctly determine a track even with tens of thousands of false entries. Jones says sitting on that much data has even helped police: 'thieves listening to music on an Audioscrobbler-powered media player have helped police in the US, UK, and other countries track down users' stolen laptops.' Does sitting on a mountain of data make Last.FM powerful enough to start making a stand against the record industry? CBS certainly thinks so — they bought the company for £140 (~$200) million last year."

Comments

Re:Now What...

[Mad Merlin]

I have a similar site that I wrote (pre-audioscrobbler). Granted it's crap, but I have mountains of data also. Closer to 1 tb than hundreds of tb. The question is, how do you monetize the data?

If you could (accurately) answer that question, then you'd act upon the answer...

Why do you think Google ads are Google's bread and butter as far as cashflow goes? The reason is that Google has a treasure trove of user data, probably more than anyone else, so they can really make contextual ads work. Anyone can write an ad engine, but not everyone has access to mountains and mountains of user data.

You might be surprised at how important context is when you're trying to promote something. Say you're trying to promote an online RPG like Game!, if you took a random collection of people, probably less than 5% of them would be interested in playing, but if you can target gamers specifically, that number might jump to 50%. If you're paying for every impression, that makes a world of difference.

So not only do you need to understand your audience, you also need to effectively target them. Now, how do you do that? Data mining of course, and the more data the better.

Pretty much all data has value, figuring out how to turn that data into money is extremely subjective and might involve some black magic, and definitely requires luck too.

Surpassing Pandora

[Paaskonijn]

The company surpassed Pandora and others largely due to its unique datamining features

I would think that being available outside of the USA may have helped quite a bit as well.