By: Eric Siegel, Founder, Predictive Analytics World
In anticipation of his upcoming conference presentation, Predictive Sales Targeting in the Energy Industry, at Predictive Analytics World San Francisco, April 3-7, 2016, we asked Nate Watson, President at Contemporary Analysis, a few questions about his work in predictive analytics.
Q: In your work with predictive analytics, what behavior or outcome do your models predict?
A: We have been lucky enough to work with a number of companies in a number of verticals. This year we have been asked to predict the price of transportation for a grain trader 90 days in advance, which potential voters would most likely favor a gambling initiative on the ballot, which potential customers would be interested in a product and what message would motivate them, which current customers would be interested in a second product, and which companies are tapped out of resources in their communities and their expansion is going to require a move or at least building a division in another city (What we are presenting on at PAW).
We have also seen an increase in the number of companies asking us to help implement the ideology of predictive analytics into their organizations. This is less of a project based relationship and more of a human resource based relationship. It has actually caused us to rethink how we might be able to interact with clients and help build their own predictive analytics capabilities. In 2016, CAN is going to offer the ability to purchase a part time data scientist and then help them manage the change management that happens when companies switch from reactionary to proactive in their decision making. It is really going to make a difference for companies going forward. They won’t have to jump into the deep end to get a data scientist. We are going to offer "toe dipping".
Q: How does predictive analytics deliver value at your organization – what is one specific way in which it actively drives decisions or operations?
A: It fundamentally changes your culture to one of innovation and "anything is possible". Because we aren't a software vendor and instead are entrusted with the hardest problems a company has been grappling with for years, we get to hire the smartest out of the box thinkers we can find. Everything is possible and solutions are just something we have to figure out. It is something we call "less wrong". We don’t have to have a 100% correct solution for it to be correct. As long as it is an improvement over what we were doing, we think it a success. It sounds rudimentary, but in most companies this is not the case. They have to have a large jump in "correctness" for something to be valuable. Because of our mentality and our lack of fear in implementation, we have implemented models into all kinds of things we do here at CAN.
Our newest success is using predictive models to find and managing talent. We found that resumes are not good predictors of cultural success or skill sets desired. Just because a data scientist has 7 years of experience doesn't mean they actually have more modeling experience as someone with only 3 years of experience. The quality of that experience has to be measured somehow. So we built predictive models measuring everything from actual skills to culture fit and have used them to find, train, and hire our own talent. It has also allowed us to manage talent better as I now understand who to put on projects, who has which actual skill set, and which data scientist will best fit what we are trying to do with a client. It's actually something that has worked so well, we are going to deploy it as a product in 2016.
Q: Can you describe a quantitative result, such as the predictive lift of your model or the ROI of an analytics initiative?
A: In September, our grain trading client asked us to solve a problem called freight variance. Freight Variance is when they actually purchase the right to grain but do not take ownership of that grain until later. In the initial contract, they have to predict what it will cost them to transport their grain at time of purchase to their facility and lock in that rate. That rate can vary widely upon actual time of delivery. The predicted rate versus the actual rate is called Freight Variance.
Our client was only using gut feel as their way of predicting price–sometimes leading to large variances and a loss of profit. CAN built a Proof of Concept model for their corn trading that used not only their data, but also weather, price of fuel, price of rubber, and a few other secret ingredients to lower their freight variance by over 35% or $3.94 Million dollars per year. This year, our goal is to improve their model, operationalize the model into their bidding process, and add their other ingredients to increase their cost savings to more than $10 Million.
Q: What surprising discovery or insight have you unearthed in your data?
A: When we started 9 years ago, we used to be shocked at the lack of data when companies told us they had a lot of data to analyze. Now, in 2015, we are shocked at the number of companies that have implemented Big Data or BI initiatives, have loads of data, but have little to no understanding of what to do with it. Last year, we spent a lot of time educating companies that their customer data probably contains really important and potentially money saving ideas in it. They just need to employ a data scientist (either internally or hire a consultant) to turn their data from noise to information.
It is one of the reasons we implemented the Staff Augmentation division of our company. We saw a need beyond doing projects and being an external vendor. Companies need help just understanding what they have, if it is useful, and what they can do with it. We see the need for companies to embrace the ideology behind what predictive analytics is and does before the implementation of a project will have any real impact to the bottom line. We see the need for a lot of beginner companies to be having a data scientist really spend time in the data with no agenda. It is only then a real proof of concept has meaning and value can be devised and implemented.
Q: Sneak preview: Please tell us a take-away that you will provide during your talk at Predictive Analytics World.
A: I want people to understand how valuable using predictive analytics and data science is in their organizations. All parts of business, not just sales and marketing, can be improved by being proactive instead of reactionary in their management. I hope to show people and open up their minds to other parts of their business that need to be revolutionized including many parts of operations and HR. Hopefully they can take back some ideas to their own businesses and lead other parts of their businesses to adopt the philosophy of proactive management through the implementation of predictive analytics.
Don't miss Nate’s conference presentation, Predictive Sales Targeting in the Energy Industry on Tuesday, April 5, 2016 at 4:15 to 5:00 pm at Predictive Analytics World San Francisco. Click here to register to attend.
By: Eric Siegel, Founder, Predictive Analytics World