A while back (July 2015), I was fortunate to speak at PyData in Seattle about the pitfalls of overreliance on data. You can find the slides here.
My talk centered around my belief that the term "data-driven," when taken at its face value, should not be something we strive for. Instead, we should seek to be "data-informed." Pedantic, I know, but for those that do not work within the field, I think the distinction is important.
Here's its abstract (pardon my snark):
Companies can't stop gushing about how "data-driven" they are - how they're using "big data" and "data science" to synergize and streamline all the things. But being driven by data alone is a flawed approach. Instead, companies should seek to be "data-informed" - interweaving designers, UXers, and data scientists so that each side is able to perfectly complement the one another.
This talk will discuss the importance of allowing data and user research to complement one another, in addition to the pitfalls of being driven by data alone (for instance, the cons of A/B testing).
While the actual talk isn't one of my best - I sound like I'm reading from cards (I kind of was) - I'm still a big believer in the overall message.
We shouldn't be surprised that being data-informed is ultimately a better approach. Simply, we're just adding more information - quantitative and qualitative - to our existing dataset and weighing that information appropriately.
The best decisions make use of all relevant information, not a limited set, much like the best algorithms are those developed with the best data and features.