Most Data-Driven Cultures… Aren’t
Business + ManagementConferencesCultureWest 2018posted by Elizabeth Wallace, ODSC July 31, 2019 Elizabeth Wallace, ODSC
For Cassie Kozyrkov, Chief Decision Scientist at Google, reducing the instances of errors in statistics is the top priority. Many organizations think of themselves as data-driven, but in reality, it’s at the mercy of good leadership at your organization. If your organization isn’t good at getting and using data, you can’t be sure your data science team can contribute something of value. Let’s take a look at some of the ways she believes your organization could be failing in data rigor.
[Related Article: Meet the Leader of a Data-Driven Work Culture: The Data Science Manager]
Data-Inspired or Data-Driven?
Imagine you want to stay at a hotel. The hotel has 4.2 stars, and you say “Yes, this is great.” But in reality, you wanted to stay at the hotel in the first place. If you didn’t, a 4.2 wouldn’t be good enough. Many times, our decisions are driven by our gut, and the numbers fall close to where we wanted to be in the first place. Until we are data-driven and not just data-inspired, you won’t have a data-driven business.
Tip 1: Decision making is a skill. You must know the difference between data-driven and data-inspired. Answer for yourself, “If I don’t learn anything else with data, what am I going to do?” This is your default action. Understand your default action to see how the data affects it or doesn’t.
Tip 2: Don’t forget the point of data science. You must operationalize data science. Let data science be the discipline of making data useful. If the data isn’t actually beneficial to your decision, you’re working on the wrong thing.
Tip 3: Inspiration is cheap, and rigor is expensive. Conclusions about the data are easy to come by, but moving beyond the data to the real insight is difficult. Everyone is qualified to read the data, but not everyone is qualified to take those insights beyond the conclusion.
Tip 4: Split your data. Don’t trust insights without this concept. You must split your data to see if the data that inspired you also exists in data that doesn’t. Give one part of your data to your entire organization and see where it leads. Teach them how to look at data.
Tip 5: Rigor begins with the decision maker. We don’t need rigor for rigor’s sake, but having the leadership that allows rigor to inform decisions and build a company culture of data. You must understand how rigor applies to business applications, i.e., the most useful thing and not the most complicated thing.
Tip 6: Understand how decision making is delegated within your organization. A misalignment can break trust. Data science leaders must know how data flows through the organization to become true decision makers.
Tip 7: Harness large data sets. The history of data science is a history of data splitting. In the old days, there wasn’t any data. We made decisions from our gut. Now, we have so much data that we have the option to choose the right kind of data in our data sets to uncover inspiration and test that conclusion.
Tip 8: Do things in the right order. Start with what’s worth doing, move to how it works, and finally, what is the method that gets the job done.
Tip 9: You’re at the mercy of the data quality. You must take the data quality seriously, or you’ll be guilty of “garbage in and garbage out.” You must have a good relationship with your data analysts to understand how your data can be the best and not just the “most.”
Tip 10: Testing is the basis for trust. You must test your data for accuracy and to find insights that work. It’s vital to produce results for the business and not just fun new models. Build a culture of testing and rigor to make sure your stakeholders can trust the results you’ve put out.
[Related Article: A Manager’s Guide to Starting a Computer Vision Program]
Data-inspiration is excellent for getting things moving, but ultimately, it’s a shallow victory if you don’t understand how to build a true data culture. Make sure your data is rigorous and builds real results for your organization, or the breakdown will be very real. It’s possible to become data-driven, but we must accept what data can do for us and how we should change our perceptions of it all. Data isn’t magic. Make sure you’re headed in the right direction.