Most data science projects are doomed to failure before they even start.
There are a couple of reasons. The aspiring data scientist and management may be drawn to a sexy problem rather than an important problem. The full range of data required to do a complete analysis may be inaccessible or even non-existent. And even when a demonstrably performant statistical model can be built, existing business processes may be too expensive to modify to take advantage of the model.
These are just a few issues that come to mind and these seemingly “obvious” problems are, somewhat frustratingly, much more common than one would think. The solution is one familiar from enterprise sales strategy: build a large pipeline of opportunities and then qualify hard before investment of effort.
With thanks to Sky, a former colleague from my Mumbai days, here are three sets of readiness studies one can perform to qualify data science opportunities and hopefully cheaply rule out those that just won’t cut the mustard.
The output of an opportunity qualification exercise can take many forms. One that makes a lot of sense is an Investment Logic Map.
As Charlie Munger said: “All I want to know is where I’m going to die so I’ll never go there.” Hopefully the above will help young aspiring data scientists avoid a few death traps. Here’s a useful catchphrase from a former Pivotal colleague to help us remember this mental model: Don’t ever let your statistical model die in a Powerpoint presentation.
Mental Models 4 Life 