Don't Hire a Data Scientist (Yet): Speaker Series Event Recap

Insights / 04.29.2019
Red Door /

7/26/2023 7:29:30 PM Red Door Interactive http://www.reddoor.biz Red Door Interactive

How do we make use of the abundance of data generated through new technologies? Are the revolutionary capabilities of data science evenly distributed across the business landscape? Why are so many companies still struggling with business intelligence, predictive analytics, machine learning, and tracking their data? What do data scientists actually do?
For marketers to get an edge or even simply to stay competitive, getting a handle on data, business intelligence, and machine learning is essential. It’s not enough to simply believe in the mysterious power of big data. You have to understand it. If you hire a data scientist, or work with one, you have to create the necessary conditions for their success. Before you’re ready to use state-of-the-art tools, you need to get your own house in order.

We addressed these and other important concepts and questions in our "Don’t Hire a Data Scientist (Yet): Prepping Your Data for AI, BI, and Personalization" panel discussion, presented in partnership with Tealium, an industry leader in real-time customer data orchestration. To make sense of what’s new and what’s next, we assembled a group of experts with skin in the data game.

  • John Faris, President, Red Door Interactive (Moderator)

  • Andy Batten, Sr. Director of Data Analytics, Red Door Interactive

  • Ron Hadler, Sr. Director of Marketing Technology, Red Door Interactive

  • Ted Sfikas, Director of Solutions Consultants, Tealium

A few years ago, “big data” was the buzzword du jour. How are marketers most commonly using big data today?

“Marketing is really familiar and loves buzzwords,” said Ron Hadler. “Big data, data scientists, AI. The idea here is it reminds me of a movie, The Force Awakens. Han Solo and Finn are on the Death Star, like, ‘How are we going to get into this facility?’ And Finn goes, ‘Well, use the force.’ And Han Solo says, ‘That’s not how it works.’”
Hadler explains that it’s not useful to think of big data technologies as magical divining rods - they’re more like a funnel for information, the effectiveness of which depends on the skill of the marketers using it and the quality of the information flowing through it. “Big data is what drives data science, what drives artificial intelligence. And we have a saying here at Red Door that bad data equals dumb robots. If you don’t have the data, you don’t have quality data, you end up with bad predictions, bad personalization, and bad porting.”
Andy Batten went into specifics, suggesting reasons why some companies still struggle with managing and tracking their data. “Really, the whole idea around personalization and BI is to be able to hit users with the right messaging at the right time. If you can’t effectively know that it’s the same person that was on your mobile device and desktop, or on this site and this other site that you own, if you don’t know that’s the same person and they’re interacting with your site in these different ways, then you can’t effectively message to them and move them further along that funnel.”
Takeaway: Big data by itself is not necessarily a boon to your business. The power of data relies with those who know how to track, manage, organize, and interpret it.

What is a Customer Data Platform (CDP) and how does it differ from a Data Management Platform (DMP)?

Ted Sfikas of Tealium touched on a possible solution for this problem, using as an example a component in Tealium’s universal data hub called a Customer Data Platform, or CDP. “It distinguishes between the mountain of big data that’s coming into your organization and it splits the stream,” Sfikas says. “It is meant to detect data that is distinguishable from a customer perspective,” including psychographic data distinct from customer data.
Customer data, Sfikas said, is “very raw: page view, first name, zip code.” By contrast, psychographic data is “very rich: high-value customer, high-value prospect, frequent visitor. It depends on the dimension of time. So, this is something that needs intelligence to process it... That second type of data is what personalization technologies absolutely need to work.” This is first-party data, drawn straight from a user interaction.
Hadler again encouraged marketers to apply some skepticism and look beyond buzzwords. “I think the difference between platforms that are true CDP and platforms that may be acting like CDP or using that marketing buzzword and a customer data platform is, a customer data platform is really going to collect all the information that Ted was just talking about.”
Batten added a further cautionary note on applying first-party data. “You have to be careful on how you use it because it is going to be the most accurate, but then how you use it can veer towards creepy.”
The panel examined in depth some recent controversies around data breaches, the arrival of Europe's GDPR restrictions and California's impending CCPA compliance demands. Looking at such news stories from an interested, analytical perspective can be a tremendous help in making sense of the data landscape and determining how you want to marshal your resources.
Takeaway: A CDP collects rich psychographic and first-party data, which is what is needed for personalization to work properly. This can vastly improve applications for data - it can get a little creepy!

Why is there such a high demand for data scientists these days? What do they do all day?

Hadler highlighted the ubiquity of data science concepts in the modern world and explained how almost everyone has some familiarity with applied data science and machine learning from using platforms such as Gmail and Outlook. He mentioned that many companies will avoid deep investments in hiring experts and enduring months of modeling trial and error by instead taking advantage of platforms that use their own machine learning.
“We use machine learning any time we place an ad on Google or Facebook,” he said. “They use machine learning to ingest in your data for customers and it will look like audiences, putting those ads in front of the right people. We have a platform here at Red Door where we can basically make changes to your website via for SEO, and then within 12 hours understand how it’s going to affect your Google search ranking. Normally that would take 90 days. Those are some of the things in the way that machine learning is really being used by marketers today.”
Moderator John Faris summarized this as “feeding the robots data and then applying an algorithm that’s constantly updating in real time.”
Sfikas agreed that this is an exciting time in the field of data science and added that he’s optimistic that today’s data scientists are placing a strong emphasis on clean data, thereby addressing one of the chief issues holding back many companies. “In the last two years that I’ve been touring the United States and talking with customers and agencies about machine learning, the number one problem of data scientists is that they spend 60 to 75 percent of their time doing something called data wrangling.”
Takeaway: Data scientists are expensive and their projects are complicated, but it is crucial to understand some of the details of the work that data scientists are doing, particularly in regards to the platforms you may use.

What are some of the typical issues with getting data prepped for Predictive Analytics?

Whether or not it’s time for your company to hire a data scientist of its own, today or next month, is an open question depending on numerous factors. However, the time to prepare your company’s data for predictive analytics and all the applications of machine learning was yesterday - the next best time is today.
“Even if you can’t hire that data scientist today,” said Batten, “you’re going to need good data for your models once you’re ready to apply modeling to your data. We typically do tracking audits, and campaign tracking audits, and campaign data audits, and things like that to understand what’s going on… Getting your data, your house in order, from the web analyst perspective is what I would recommend.”
Sfikas encouraged us to get serious about tracking and cataloging our own data, starting as soon as possible. “Build a data catalog and write down, as a human being, all of the data that each department needs. Not just your digital marketing teams, but also your warehouses, your legs, your machinery, write it all down. It’s a catalog, and it’s a service.”
From there, Faris suggested getting to work on a solid data layer strategy. “A lot of this stuff is going to be automated, and so it’s about, upfront, setting the right strategy with really smart people. Setting things up, configuring them appropriately, and then letting the machine just take it from there.”
Takeaway: There’s no shortcut around a rigorous data audit. It’s the first step toward preparing yourself for success with data science and machine learning, no matter what that looks like for you.
Ready to operationalize your data to strategically plan for things like personalization, predictive analytics, audience segmentation, and other areas of business intelligence? Start getting yourself “data-ready” by downloading (and answering) our 10 Questions to Ask for Data Readiness below.
Still have additional questions about data science, machine learning and their place in today’s marketing landscape? Feel free to drop us a line on social, or reach out to us on our site — we’d love to chat.

  • Insights
  • Data Analytics & Visualization
  • Data Quality Management
  • Marketing Systems Integration
  • Marketing Technology
  • Predictive Analytics