Parse, quantify, calculate, predict: The new metrics

Anyone with agency experience will tell you how sick they are of the phrase, “It’s like comparing apples and oranges.” Why has this idiom become the standard excuse by which so many people defer making a comparison? Let’s say I like apples twice as much as oranges. Now, if you offer me an apple or an orange, my comparison and subsequent choice doesn’t seem too difficult to make. Sure, we haven’t exactly compared the apple to the orange, but we have successfully made a decision by comparing my preference for each. (I’m apparently not the first to make this argument, and Wikipedia does a great summarizing.)

No, the comparison above isn’t too useful, but we can learn to apply the practice of adding a preference value as an attribute to make a whole range of comparisons possible. Consider how Google finds the “most relevant” sites. Rather than making a decision based on one person’s preference like we did with apples and oranges, Page and Brin concluded that the best way to judge a site’s relevance was to compare the number of times other people linked to, or endorsed that site. After building in several other attribute values, they came up with the Google algorithm.
Only after joining my current company, Zapoint, did I begin to truly consider the extent of potential applications for this model. We applied this logic to the resumè to develop an algorithm that we now use to compare individuals for employment. Comparing 10 resumès by hand for a single position is a daunting task, but objectively comparing 1,000 is both unmanageable and unrealistic. Our algorithm basically compares resumès based on skills (e.g. leadership, brand marketing, project management) and achievements (promotions, job changes, employment consistency). When given a search interface, this algorithm can be used to instantly and objectively automate candidate selection.
Comparison algorithms are appearing in increasingly relevant applications, and are usually applied to similar, seemingly unquantifiable items. Take image comparison for instance. Same idea, same process, but a different set of calculations. Like the resume which is parsed into skills and employment terms, you first must parse the picture into small bits or pixels. Once the image is broken into pixels, there are a host of values that you can apply to begin to compare two pictures. Comparison is only the tip of the iceberg when it comes what image processing algorithms can do.
With this immense power at our collective fingertips, how then do we continue to be so reactive when it comes to some of our most important issues? As far back as 2004-2005, the same people that were burned by the “tech bubble” were buzzing about the impending burst of “housing bubble.” Rather than considering where everyone was getting the money to buy these houses, and who would be the real victims of the collapse, people continued to creatively finance assets well beyond their reach. Borrowing was maxed, home values declined, and the balance of assets to liabilities outstanding shifted enough that individuals and institutions couldn’t meet their obligations.
How did modern economists and bankers miss in their calculations badly enough to bury the financial system? Maybe the industry and regulators monitoring it are as corrupt as the media portrays them to be, but I have to believe that we should have the ability to foresee this type of economic situation. Following my usual logic, I have to believe that Google can do it better and luckily they’re heading in this direction. Google.org currently has 2 open initiatives that key in on this issue, Predict and Prevent and Inform and Empower to Improve Public Services. No, fixing the economy isn’t explicit in these initiatives, but Google is applying the predictive methodologies outlined above to many of the issues that we are facing today, and will continue facing in the future.
Just as Google is currently monitoring national health two weeks faster than governments can, new algorithms will be born to inform us about, and help us manage, an infinite range of issues. We will come to rely on this type of hard data and precise calculations rather than the wild speculation and media hype on which we seem to currently use as a basis for our important decisions.