Alphabet DeepMind and NHS

Concrete example of the use of Big Data in creating a new prediction model.

Google will be using NHS data to warn doctors of potential kidney issues.

http://www.theverge.com/2016/11/23/13726280/deepmind-nhs-data-streams-app-new-deal

Triumph of the curious

This article in the Ringer analyzes the path Theo Epstein took to guide the Cubs to their first World Series victory in 108 years. He led a revolution of the "curious" over the incurious. His easy challenge to the status quo helped solidify the value of intelligently applied advanced analytics in baseball.

Data can be powerful, but it also needs to be harnessed carefully.

Here is a link to the article:

https://theringer.com/2016-world-series-chicago-cubs-theo-epstein-analytics-war-9f1248c44eb7#.fw8ba92p0

Data analysis helps improve airline safety

Very strong article from the Wall Street Journal showing how airlines are using big data to analyze patterns and reduce injuries and errors.

Analytics and morning coffee

Data can sometimes speak very loudly. The difficulty comes in interpreting what it means. As we become inundated with more and more raw information, the threat of "analysis paralysis" becomes all too real. Despite those pitfalls, there are signs that marketing teams are becoming much more skilled in the use and interpretation of analytics.

This article in Fortune discusses the rapid growth of analytics. Companies are expected to increase their spending by as much as 66% over the next three years. One of the main beneficiaries are companies like Quantifind--a new startup focusing on marketing analytics. 

Fronted by longtime marketing executive David Karnstedt, Menlo Park, Calif.-based Quantifind tripled its customer list last year to 30 brands, including beverage makers Pepsi  PEP -2.29%  and Heineken. Its specialty is sorting social media “signals” and correlating them with sales data. “(Our) platform allows marketers to explore, understand, and change their impact on revenue by serving up the most likely correlations and letting human judgment ultimately discern what to study further and where to take action,” he said.

Mr. Karnstedt goes on to give a demonstration of how their software works:

How does Quantifind’s software work? Without disclosing the name of the customer, Karnstedt offers the example of a quick-serve restaurant that was seeking ways to increase breakfast sales among teenagers. Despite the addition of several menu items catering to that demographic, the company saw very little impact on results.

Using Quantifind analytics, the company dug deeper and discovered a pattern of negative feedback—not from teens, but rather from their mothers. It turns out they didn’t care for the coffee on the restaurant’s menu. “More data also revealed that teens were relying on moms for a ride for breakfast. So the answer to the question ‘How do we sell more breakfast to teens?’ was to offer moms better coffee,” Karnstedt said. The strategy worked. Coffee sales, in particular, doubled.

Sometimes you find things where you are least likely to look!

Here is the link to the full article:  Marketing Analytics Specialist Quantifind Doubles Backing - Fortune 

Depth in analytics

We often use terms like "big data" and "analytics" as if we mean the same thing for all applications and uses. That is not true. During one of the presentations at a meeting I attended last week, there was a slide that helped explain some of the major differences along a spectrum.

The presenter discussed three main types of analytics:

1. Descriptive:  What happened and why
2. Predictive:    What could happen
3. Prescriptive:   What will happen

As you can see, analytics starts by describing and reporting about past events. It helps explained what happened and then goes into the number of times specific events occurred and where. Over time, as you build your competency, you start to get into predictive and prescriptive analytics.

In the speaker's opinion, the best organizations are weaving analytics into the fabric of their organization. Data analysis becomes a key part of the basic decision-making and strategy of the organization.

I am a big fan of analytics, but I still think there are issues with this interpretation that require clarification. While I generally agree with the model, I still am unclear about how you decide whether or not you are asking the right questions. You might have tons of data, but if it isn't helping to solve your business issues, it may not be all that valuable.

Starting off the New Year

Original post:  Jan 5, 2015

Over the holidays, I was reading an article on sports analytics. While it may seem as if the use of statistics to improve the performance of sports teams has no bearing whatsoever on the production of medical devices, I would challenge you to think about this excerpt:

At the end of the day, it boils down to this. The information is only as good as it is to the person receiving it. I’ll take a C+ piece of analysis communicated perfectly over an A+ piece of analysis that’s not communicated well. Only a small portion of the work is the analytics itself. The rest is putting it in a practical format so the salary-cap person and the coach can appreciate it and use it. Instead of trying to go overboard with analytics, focus on the practical: Focus on the things that have the highest impact on your organization.

Full text of the article here:  NFL Teams Are Analyzing Everything From The Salary Cap To Fan Loyalty | FiveThirtyEight

This one paragraph really highlights my own personal goal for 2015. I'll be searching for those few great pieces of analysis and insight that can truly impact our organization. I will then challenge myself to communicate it as perfectly as humanly possible.

What's your goal for the New Year?

Spurious correlations

Original post: May 16, 2014

I personally believe analytics is a tremendously powerful tool. That said, it is just a tool. It has to be combined with informed analysis applied toward a specific outcome in order for it to be useful in a business sense.

A little knowledge can also be a dangerous thing. It is especially so when we think we are more capable than we actually are. There is a famous saying that "correlation does not imply causation". In other words, simply because one trend looks like another does not mean that the two are related.

There is a new website that helps to demonstrate this principle. Spurious Correlations lists a number of trend lines that have a 99% correlation. Unfortunately, they also have exactly zero relationship to each other. It would be madness to say that one causes the other. We ought to smile and remind ourselves of this as we start to examine our own data.