Big Data might be the big catchphrase, but Nate Silver wants to talk about applied statistics, probability, and Big Bias. If you read The Signal and the Noise, you know Silver. He’s known for, among other things, successfully calling the 2012 US presidential election. He was celebrated for his accurate predictions, despite a sea of conflicting and confusing data points throughout the campaign.
Somewhat surprisingly, Silver’s keynote on Monday night at SAP TechEd in Las Vegas – the night before Vishal Sikka’s morning keynote kicking off tons of SAP HANA announcements and technical sessions - wasn’t about statistical minutia. Instead, Silver focused on lessons from the 2012 election and more recent real-world examples of misunderstood statistics. We are all swimming in data. So how do we figure out what matters?
Here are some takeaways from Silver you can take back to the office:
- Silver took a pointed look at how, in politics particularly, voters’ data consumption is becoming more fragmented and specialized with the rise of candidate-friendly media channels. As a result, voters are getting less exposure to ideas they disagree with than in the past. His point: if you’re only getting the stats you want to hear or getting cherrypicked information, can you make a truly informed decision?
- Another caution: Humans are trained to look for patterns where they don’t exist. And the more data you are getting, the more random events can look like correlations (example: football wins affecting the stock market). Applying some human skepticism to these increasingly complex data relationships can help keep them in perspective.
- The “smarter” the technology gets, the more it can generate error. Again, common sense applies. Investigate those outliers or head-scratching findings. Because sometimes a random event can actually be a random bug.
- Don’t discount iterative change. Over time, you can create increasing change with less and less effort. Silver specifically mentioned this in the retail context, where a 3-5% shift in sales could mean significant impact in revenue.
- Be wary of guarantees based on forecasts. Get a grasp of probability. To really make an accurate risk analysis you have to understand your margin of error.
- Some findings show that the less you admit bias, the more bias researchers find (in hiring, for example). Admitting your potential blind spots can be a big help here.
Silver and SAP’s Steve Lucas also commiserated about hiring – Lucas admitted it was challenging to find data scientists with the right skills for this new world of data analysis. In addition to technical data skills, old-fashioned work ethic, entrepreneurialism, and creative problem solving count. It looks like the door is open for those folks who don’t come in as data scientists but have these traits could build on their technical skills.
Silver admits that daily business decisions don’t always have the clear-cut results you get when counting votes in an election. But faced with massive quantity of information, there’s big opportunity if we become educated consumers of our own data.