Culture eats data science for breakfast
Read time 3 min
“It’s not enough to introduce big data technology, fancy algorithms, and some unicorn-like isolated data scientists to any organization and work culture, and expect them to be able to change things on their own. In order to get the most out of data science, the culture has to change to support data-driven work,” writes Johan Himberg, Data Scientist at Reaktor.
A LinkedIn post, The Inconvenient Truth About Data Science by Kamil Bartocha, recently popped up in my feed. For a Data Science professional, the list is funny – because it is true.
In just a few lines, it says something essential: the academic and technical excellence is but a fraction of our daily work, and there are a lot of buzz words that flutter, like miraculous butterflies, around it. I could add a few personal favourites to the list, starting from “marketing will ask for a segmentation, no matter what the problem is”. You’ll find more additions in the comments of the article.
However, Bartocha’s piece brought into mind an inspirational meeting from a while back: half-a-dozen seasoned data scientists talking about data science training. How should you do data science in business? What to tell the people? From a group of data scientists with doctorate degrees, you would expect advanced lessons on methodology, technology, business acumen, visualization – or at least on the importance of data wrangling (yes, it will take some 80% of the project time).
However, our conversation kept circling back to organizations’ culture and communication. This is how I’ve summarized our conclusions:
Principles for a data-driven organization
- Be Bayesian. Understand uncertainties.
- Be curious. Seek evidence and learn new things constantly.
- Be active and agile. Test; don’t just gather observational data. Fail fast – but not too fast. Collect enough evidence.
- Be courageous. Act on the evidence and analyses.
- Be truthful. Don’t abuse data: don’t bend evidence upon your wishes or “internal politics”.
- Be transparent and helpful. Always cooperate and work across silos. Team up around the information path: action–decision–analytics–data. Advocate data democracy, within legal, ethical, and business critical limits.
- Be wise. There is a time to be analytical and a time to be intuitive and far-sighted.
The first virtue, being Bayesian, is of course, the essence of data science. The rest of the principles listed may sound “soft” at first. But, for example, being active is not just about personal characteristics. In data science, active means controlled, empirical tests instead for passive data gathering and blind “correlation hunt”. Courage is needed when you actually start relying empiricism instead of “always done” and “gut feeling”.
Of course, we need to understand what the organization is aiming for and what the business is. Of course, professional excellence and technology are important. But having said and acknowledged that, we have learned, many times, that by themselves they are not enough. Being data-driven is about doing things based on objective information, that is, evidence. Maybe we should organize the way of doing things accordingly?
It’s not enough to introduce big data technology, fancy algorithms, and some unicorn-like, isolated data scientists to any organization and work culture, and expect them to be able to change things on their own. In order to get the most out of data science, the culture has to change to support data-driven work. If culture eats strategy for breakfast, it’ll definitely do the same for data science.