In anticipation of his upcoming conference presentation, The Sprint for Teaching Data Science: LinkedIn Learning, Analytics and the New Era of Just-In-Time Skills Training at Predictive Analytics World for Business New York, Oct 29-Nov 2, 2017, we asked Steve Weiss, Content Manager, Data Science and Business Analytics, at LinkedIn, 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: At LinkedIn, (among other things) we track employment supply and demand, toward predicting where market opportunities will be. We want to help people with their career prospects, and we want to help employers find the best candidates to fill job openings. At LinkedIn Learning, we track specific skills-demand in order to gain insight we can apply toward creating the online learning courses that will—in aggregate—help the most people. That can mean a fairly broad variety of course topics—some courses will be extremely high-demand and others will be fairly narrow and focused on very vertical skills or topic coverage—but the overall aim is to provide a robust set of very helpful job skills for the present and near-future. Provide the training people need now, but also what they’ll need over the next 18-36 months. Predictive analytics help us skate to where the puck will be, not just where it’s at this moment. Things just move too fast in job markets to play it any other way.
Q: How does predictive analytics deliver value at your organization – what is one specific way in which it actively drives decisions or operations?
A: Given the high production values we use in our training courses, we don’t have unlimited resources with which to create our content at LinkedIn Learning; we’re continuing to add production resources and personnel to aggressively stay ahead of demand of LinkedIn members, including all of our Enterprise clients (we have accounts with virtually the entire Fortune 500, along with thousands of corporate clients globally). So we need to choose which course topics to cover very carefully. Predictive analytics helps take the guesswork out of the process, although to be clear, we definitely rely on the hard-won market knowledge and network-based wisdom of our content managers to flight-test the theories borne of data-driven insights. And vice versa…
As one example: when I came aboard as content manager for data science and business analytics two years ago, tasked with building a course library where none had existed in that overall category, the common wisdom at Lynda.com (this is right before we were officially, operationally integrated with LinkedIn as LinkedIn Learning) was that we were over-represented with Excel courses, and perhaps under-represented in topics like Qlik and Mathematica. But analyzing the top 50,000 listed skills pulled from LinkedIn’s 400 million users (now over 500 million, BTW), and measuring those against the most in-demand skills sought by all the companies doing recruiting and in-house training via LinkedIn, showed us the opposite: That Excel, perhaps unsexy as it might seem to hardcore data scientists, was still easily the tool of choice for a surprisingly large number of people doing data analytics work. The data showed us that—as always in an ever-dynamic tools market—certain tools, languages and platforms were doing sometimes surprisingly better than others. Which doesn’t mean those products and topics are on the way out by any means, but the data can help us refine our product roadmap strategies.
Q: Can you describe a quantitative result, such as the predictive lift of your model or the ROI of an analytics initiative?
A: We use our analytics to test and verify emerging topics, so say, two years ago, Apache Spark was only beginning to emerge as a listed skill being used by LinkedIn members or as a skill being sought after by employers. I knew that was about to change, but wasn’t going to show up in all the data we use. So what’s interesting is to go ahead and line up a course on Spark for Data Science, as a strategic content development decision, and wait to see Apache Spark appear and then grow in our skills/demand analysis. When, one or two quarters later you see the demand manifesting in the numbers, it’s a great feeling: you release the course, you see the numbers reflecting heavy course usage (in terms of views and number of viewers), and customer feedback thanking you for being ahead of the curve and providing forward-looking skills training for people to grow their careers. And you hit the Go button on more courses in the same area—as we’ve been doing for a while now for Spark—and you check against your competition, who maybe haven’t jumped into this part of the market yet, and you’re receiving feedback from enterprise clients who are telling your corporate sales team that they’ll adopting LinkedIn Learning due in part to the fact that you’ve got the kind of Apache Spark coverage they need, right now… and it feels great. You’re using analytics to win for the user.
And if it doesn’t… if that demand you suspected was going to materialize doesn’t actually arrive—and fortunately this hasn’t happened yet—it provides a sanity check. Makes you examine your on-the-ground resources and other data-gathering techniques and troubleshoot them. Predictive analytics can make you look really smart, but it also keeps you humble.
Q: What surprising discovery or insight have you unearthed in your data?
A: As mentioned previously an awful lot of people are using Excel to do limited-scale data science. That—along with the increase in numbers of people listing “data science” and related topics among skills they’re offering, pursuing, or recruiting for—suggests the entire field is growing. Put another way, it suggests that data science isn’t just for data scientists any longer. So this in turn suggests that there’s a growth market for entry-level data science topics across the board, and that where you used to focus on filling topic-area needs (for emerging areas) beginning at the practitioner level (intermediate and up, through advanced-level course), you now need to build out those learning paths at the lower end as well, since it’s beginning to appear that many people are refocusing their IT skills (or are entering anew) on the so-called “data science and analytics” career market. That’s part of what makes this job so fun: you get a chance to learn where things might be headed—from the data you’re collecting and analyzing—and then you get to develop and test hypotheses about what it all means, and how you adjust your product strategies to improve the lives of your customers.
Q: Sneak preview: Please tell us a take-away that you will provide during your talk at Predictive Analytics World.
A: We’ll learn more about the ways LinkedIn and LinkedIn Learning are constantly developing analytics tools and reporting, such as our monthly LinkedIn Workforce Reports. And I’m looking forward to providing insights about the first round of results for the LinkedIn Economic Graph Research program.
Don’t miss Steve’s conference presentation, The Sprint for Teaching Data Science: LinkedIn Learning, Analytics and the New Era of Just-In-Time Skills Training on Monday, October 30, 2017 at 11:20 am to 12:00 pm at Predictive Analytics World New York. Click here to register to attend. Use Code PATIMES for 15% off current prices (excludes workshops).
By: Eric Siegel, Founder, Predictive Analytics World
Eric Siegel is the founder of Predictive Analytics World (www.pawcon.com) — the leading cross-vendor conference series consisting of 10 annual events in New York, Chicago, San Francisco, Washington DC, London, and Berlin — and the author of the award-winning book, Predictive Analytics: The Power to Predict Who Will Click, Buy, Lie, or Die – Revised and Updated Edition, (Wiley, 2016).