Last week, Red Ventures’ data science team attended the Analytics Frontier Conference in Charlotte sponsored by the Data Science Initiative at UNCC. The goal of the conference was to champion Charlotte as a Data Science and Analytics Hub, and the conference and agenda didn’t disappoint. Though there were lots of interesting talks and stimuli to take away, here’s our take on four that stood out:
Future-Proofing Your Job
One of the highlights of the conference was the keynote speech by Tom Davenport, a leader in the analytics industry and one of the top 100 most influential people in IT according to Fortune magazine. Though his whole talk covering the history and context of analytics was fascinating, one thing that resonated with us was his recommendations on how to “future-proof” your job in the age of automation and analytics. Several of his recommendations are applicable to us at RV:
- Step up: Take a “big picture” view of tasks and decide when to automate further. This is particularly relevant to us since employee productivity has been an area of focus at RV lately. We’re challenging lots of assumptions around which habitual tasks still require human ownership (e.g. paid search bidding and daily reporting) and which can be automated.
- Step aside: Let the humans do what the humans are good at, and let the computers do what the computers are good at. We say this all the time on the data science team, and it plays out most clearly in how we design our workflow. Rather than invest lots of time in deciding which variables are impactful for a problem, we let the algorithm sort that out while we focus on making sure the output is actionable and useful for our business teams.
- Step forward: Embrace the change, swim with the current, and build your own automated systems. We agree with this completely. Rather than build multiple one-off solutions, let’s invest in long-term data and decisioning architecture that will set us up for more championship years. (and more trips to Cancun!)
The Democratization of Data Science
Data scientists at RV work on high-impact problems with broad applications to many of our partners. For example, nearly all of our partnerships involve website optimization, so we prioritize algorithms that can boost response rate and visit conversion across multiple sites.
However, even with a growing team, it’s difficult to take on every challenge that comes our way. There are many smaller, partner-specific problems that could benefit from data science but are never addressed because they’re not broad enough. For example, a variable-rate energy product (which is only applicable to one RV vertical) is much more difficult to prioritize than the website optimization example above, which could impact multiple partnerships.
These trends are not sustainable, of course. J.D. Elliot of TIAA and Professor Jared Hansen at UNCC provided a compelling argument that tools that enable data science for a broader subset of organizations are the future, and we should start embracing them now. We couldn’t agree more, and we’ve been trying to put some of this analytical power into our analysts’ hands by providing them with tools to more thoroughly explore data, derive insights, and improve business practices. The creation of these “citizen data scientists” not only provides faster cycle times for having questions answered; it also allows RV to address more high-impact and partner-specific problems.
With more apps, devices, and on-demand content available than ever before, it’s clear the data revolution is here to stay. Being able to visualize and explore data is a powerful way to identify trends and gain insights, and it’s another tool in the movement towards the democratization of data science. One of the prevailing problems visualization faces, however, is that techniques for storing and utilizing data for modeling and decision-making are often disconnected from visualizing data and exploration. As data gets more complex (e.g., lack of uniformity, multiple sources and formats), we need to make larger investments to tie our data together to visualize it.
Visualization is often an afterthought, but given its power, it needs to be a critical part of a company’s analytics ecosystem. At RV, we’ve invested in technology that creates visualizations of our models’ performance on a daily basis, giving our data scientists a clear, concise snapshot to reference in real time.
More Computation Means Better Leverage of More Data
As the cost of computational resources continues to decline, more opportunities are opening up for data scientists. Abhishek Mehta of Tresata challenged the practice of sampling populations to tease out insights and make predictions. He encouraged the conference to instead utilize tools that can handle every single interaction with every single participant/customer in order to paint (by definition) the most complete picture of our cohort.
We take this to heart at RV, loading exhaustive datasets to build our models. With access to additional computation, we can also move away from parametric or statistical models that assume a distribution of the data. These approaches are traditionally more computationally efficient but may have biased results. Many machine learning models, such as Random Forest, Gradient Boosting and Deep Learning, don’t make these assumptions and can be very accurate but at the cost of increasing computation time. At RV, we leverage our resources to overcome these limitations and test many modeling techniques on our data in order to always optimize for accuracy.
As members of the Charlotte community, we’re excited about the prospect of making our town a Data Science and Analytics hub. We’re looking forward to more local engagement and helping to build a strong analytics community here.
This post was a collaborative effort by several members of the RV data science team:
John Thomas relocated to North Carolina from Boston almost a year ago to start the data science team at Red Ventures. John earned his BS in Mathematics and Computer Science at Gettysburg College and his MS and PhD in Computer Science at Dartmouth College.
David Crespi joined RV in June 2014 after graduating from Wake Forest with degrees in Math and Computer Science. He’s putting those degrees to good work on the data science team predicting customer LTVs. In his spare time, he enjoys going to the movies and getting ID’d for Rated R movies.
Jacob Foard joined RV in November after graduating from our Code2Hire program. After two months of on-the-job training, Jacob joined the engineering side of the Data Science team with hopes of being a hybrid engineer/data scientist.
Josh Izzard came to RV in August 2014 after graduating from Duke with a BS in Math and works primarily on website and customer experience optimization. Outside of work he’s a competitive swimmer and a movie buff (non-competitive).
Cory Locklear joined RV in November 2012 and worked as a back-end web developer for three years before moving over to lead engineering on the data science team.
Andrew Orso joined the RV team in 2015 after receiving his Bachelors in Industrial Engineering from the University of Florida and his Masters in Operations Research from the University of Michigan. When he’s not working on IVRs and Call Prioritization, he can be found running aimlessly on the shoulder of U.S. Hwy 521.
Kevin Pedde started at RV in May of 2015 with 4 years of data science and analytics already under his belt. A graduate of Western Carolina University with a Bachelors and Masters in Applied Mathematics, Kevin also graduated from the nation’s first analytics and data science Master’s program at NC State University. He’s also kind of a data science celebrity outside of RV.
Andrew Williams started at RV in July 2013 after graduating from Wake Forest and joined the data science team about a month ago after tours of duty in several different marketing and operations roles at RV. Possibly made of glass, Andrew is currently rehabbing after his 4th arm surgery in the past 8 years.
Want more from RV Data Science? Kevin and Jacob explain how they won HackathonCLT this year and offer tips for future ‘hackers.’