I still like my OK/Cancel partner Tom Chi's take on this in 2009 (and also the last OK/Cancel ever posted). It's the best articulated and still relevant. okcancel.com

Data is really problematic because data can be manipulated and is rarely conclusive. I think of intuition as a "point of view" and a point of view is not right or wrong it is just a lens at how you look at the problem. Data can help me refine my point of view or sometimes it changes it, but I always lead and am guided by intuition, over data.

Kevin, great article. I really love this part:

The interplay of all disciplines (engineering, design, research, marketing, sales, QA, product, legal, customer care, etc) is where the magic happens. Metrics are an absolutely critical interface between disciplines, but when wielded and controlled by only one discipline they can greatly limit the potential of the others.

Diversity is another issue, but I've always found that bringing different types of mind to bear on a problem (in a way, different sources of data) and allowing them to contribute in meaningful ways is always a good thing.

Data has a systemic bias.

It is great tool to show how to do what you are doing better, like how to row a boat better. Having begun my career as a data analyst, I love the power of data. The problem happens when you want it to be the lighthouse.

It's a fine balance between intuition and data, you just need to know when to tilt which way to keep the sword balanced on it's tip.

Personally, if it ever comes to data vs intuition at an early stage, I generally trust my intuition especially in domains I have experience. That is again a steep slope if this is data after 6 months of usage of product. Intuition should not be a masquerade for ego.

i just got off a plane on which i started reading nate silver's signal and noise. there's a part where he delves into baseball and the "battle" (at one point) between scouts and stats. scouts relied primarily on intuition, human characteristics/observation/empathy (to a point) + some stats, while the new stats-driven models were historical numbers driven. both were good, but together — they were better.

some things can be measured with data — e.g. how many branches people started, how many highlights are made, while others cannot — whats stopping people from using branch more? for example. thats where designer's empathy/intuition can provide value (as well as talking to Real People)

"The worst thing you can do is let that data stifle you"

So true. Just because a problem is easy to measure does not make it valuable to solve. I've found that often you can only get really big results from doing things that are hard to directly tie to value (however you're measuring it).

It doesn't mean that data doesn't matter, just that it's important not to fall prey to the Streetlight Effect en.wikipedia.org

At the end of the day aren't all great decisions made on conviction and belief? The biggest mistake I see clients make is thinking data can make a decision for them. That there will be clear inconvertible in data. Even Nate Silver only gives probabilities of outcomes and he can be wrong (he predicted San Fran to win the Super Bowl).

One other thing, is that data is almost always about the past. There is no data about the future. Making forward looking decisions using data is like "driving a car looking in the rear view mirror" (W. Deming). Separating leading from lagging indicators is difficult and most organizations data is lagging.

I feel like a product evolves more quickly when decisions are based on intuition/instinct. Data and Metrics are great for validating decisions or shoving infront of stakeholders.