Professor Florian Zettlemeyer from Kellogg School of Management did a great job of teaching the Leading with Big Data & Analytics last week. I sketch noted the course and gleaned 5 key takeaways.
1. Analytics Requires Managerial Judgement
While analytics itself can be performed by experts, the decisions that it drives and that are needed for it to thrive have to be made at a strategic level.
2. The Run The Business Data you’ve got May Not be the Data You Need. This was a real aha for me. Run the business data is optimized for profit and returns. It’s not objective. It is not unbiased. Analytics experiments require random data or an equal chance of each data point being chosen. They require objective data. Most of the time we try to use the data we already have to make decisions rather than understanding the data we need and generating that instead. For example, what data would help solve a churn problem? It isn’t the data we already have.
3. There is a Checklist for Bad Analytics! Any time you see a chart, ask yourself and out loud:
- Are there pre-existing differences between groups?
- Is there a common driver of decisions & outcomes?
- Can you reverse the causality?
- Is there a plausible coincidence that could also explain the result?
4. You are always sitting on an assumption! Make sure the facts really support that assumption. If you can say “For two years in a row, the difference between the price of these 2 cars has been $500 and we expect that difference to continue (all else being equal.)” then you’ve based your assumption on data.
5. Make Friends with a Data Scientist. So you can run your crazy ideas by someone and get friendly feedback.