Two stories attracted my attention yesterday. The first was an excellent ArsTechnica overview of the Obama tech team that provided the backbone for their world-beating ground game. The second was a DealBook (a New York Times blog) story about Hewlett Packard’s disastrous acquisition of Autonomy, a British company in the “big data” business.
In the first, we learn that the Obama for America (OFA) team hired a raft of coders, used mostly open-source software, and hosted almost every aspect of the campaign on Amazon EC2 hosts. Romney’s campaign hired only a handful of tech people, made none of them integral to the campaign strategy, outsourced almost everything to friends and contributors, and spent about $15 million more. The OFA organization tool to communicate with volunteers was called Narwhal, they built it five years ago and have been improving ever since. A narwhal is a smallish arctic whale with only one known predator, the orca, so the Romney team called their system Orca. Having no IT staff, they worked with Microsoft and an unnamed consultant to build it, then didn’t even turn the thing on until 06:00 on election day, at which point they discovered that it didn’t really work.
Hewlett-Packard, a once-distinguished company built by engineers with a knack for creative development, is struggling. Their hardware business doesn’t have the sexy margins of other tech titans so they’re on the lookout for ways to build their software and service businesses. Analysis of high-volume non-structured data is currently hot: Witness IBM’s Watson and everything that Google does. They spent over $11 billion to buy the company last year and are just now figuring out that accounting irregularities (fraud?) had significantly jacked up the price. Autonomy doesn’t seem to have a either good products or loyal customers, I’d guess that they’re going to mark down this investment again, in addition to the recent $8.8 billion charge.
Buying Autonomy was a stupid thing for HP to do. If you want to build a crack big data business you don’t want to buy one somebody else built and wants to sell, you find an old barn and hire a handful of people that want to tackle the problem. You look for kids, maybe a lot of them that haven’t got their degree yet. (Stanford, alma mater to Bill Hewlett and Dave Packard, is right around the corner, and they’re doing some good work in this field.) Maybe you snap up a couple of seasoned veterans with experience in the field, maybe not. Recruit somebody from Dan Wager’s analytic team at OFA. You give them a bunch of servers and storage to play with, something that should be trivial for HP. And then you pose a lot of problems for them to solve. You add a few product-centric people in the mix to turn the team’s results into salable products and you let the sales department start selling.
This has two huge advantages. First, it’s cheap. Second, it’s how you get great products.
If I Were King I’d be sorely tempted to abdicate and show them how it’s done.