The PhDs are hired, the software installed, the data collected and the rest of the company waits eagerly for profit to climb – why shouldn’t it, now that we have advanced analytics? Why can’t we hit a homerun just like the Oakland A’s did using sabermetrics? If Billy Beane had one Paul DePodesta, shouldn’t we do better with an entire geek squad?
We all secretly wish for a magic weapon to vanquish competition. Where better to place our faith than in a model churning out intimidating, neat lines of output? A perfect deus ex machina to get out of a sticky situation. However, as with anything complex, we forget that it’s one thing to own analytical infrastructure and another entirely to be able to use it well. If you’ve ever gotten your parents a smartphone, you know what I mean.
Regardless of whether you’re advanced enough to implement agent based modeling or just taking baby steps beyond bar charts, a few ground rules remain the same:
- Set a target because your analysis is only as good as your data. Whether you’re pulling data from the CRM system or fielding a new survey to gauge consumer loyalty, you should know up front what you’re looking for, starting with a list of hypothesis to prove or disprove. Which metrics, markets and consumer segments are relevant? Gather the quant and non-quant people together at this stage will help you get at what is both interesting and feasible to test. If you’re pulling from existing data, be generous in allotting time to data cleaning, especially if they come from disparate business lines and geographies, something as simple as an erroneous currency or unit conversion could derail your conclusion.
- More/ fancier analytics does not equal higher “analytical maturity”. For many companies, the arrival of analytics elicits two common responses – “yay-for-hard-definitive-proof” and “who-cares-I-use-my-gut”. In their survey of 5,000 employees at 22 global companies, our sister program for IT executives, the CIO Executive Council, found that 43% of all employees are “unquestioning empiricists”, while another 19% are “visceral decision makers” – overly reliant on or lack appreciation for data. Only 38% are the “informed skeptics”, people who apply judgment to data, are aware of its limitations, their own biases and are ready to teach. Which group do you belong to? Take their quiz to find out.
- Build your team of “T”s. “T-shaped” talent describes people with deep knowledge in one area (the “I” part) and comfortable operating less deeply in a host of others (the “―” part). For a quant person, that means statistical expertise complemented by a healthy dose of business acumen and communication skills. Depending on one’s level, the shape of the “T” should vary. An executive “analytical champion” (a la Thomas Davenport) would be something of a “fat T” (broader rather than deeper) whereas a “data scientist” would fit the profile of a “skinny T” (deeper rather than broader). The remaining 80% of the analytical taskforce falls somewhere in the middle.
It is easy to get lost in the numbers and forget all about intuition and judgment. In my daily encounters with data, I’m often reminded of this sobering fact: “The combination of some data and an aching desire for an answer does not ensure that a reasonable answer can be extracted from a given body of data.” Tip of my hat to John Tukey, the statistician who championed exploratory data analysis.
Related posts:
on 17 October 11
Respond
[...] readers of Wide Angle may have noticed that I am not the first MLCer to wax poetic about the power of data recently. Not surprisingly, this is not a coincidence. We are starting to map out our 2012 [...]