There’s a growing mismatch between the growth of data and the growth of data skills and knowledge. The former is increasing at a healthy rate, while the latter is struggling to keep up. One outfit that’s hoping to close the data education gap is Data Society. The company, which provides tailored education and training sessions to companies and organizations (as opposed to individual users), seeks to give users a basic foundation of core skills in the areas of data management, data analytics, and data science.
The Washington, D.C. company employs 50 full-time educators, who present a data science curriculum that was created by a team of professional content creators and data scientists. The company also has a roster of hundreds of part-time instructors it can tap into, particularly for specific industries.
The company started out with the business-to-consumer (B2C) model, but shifted course several years ago to the business-to-business (B2B) model, which enables it to be more flexible in meeting clients’ needs, according to Data Society co-founder and CEO Merav Yuravlivker.
“We completely flipped the switch and said, no more B2C, we’re focusing on our industry clients,” says Yuravlivker, a former teacher who founded Data Society in 2014. “We’re one of the only players in the space.”
Over the years, Data Society has trained upwards of 10,000 individuals for more than 100 clients, many of which are employed by financial services companies, healthcare firms, and federal agencies. The company’s client roster includes names like NASA, CDC, Department of Treasury, Discovery Financial Services, OptumHealth, and IQVIA, among others.
The students fall into two general groups: practitioners who need the technical skills to be able to work with data and apply the techniques, and executives who need to know what data is capable of and how to build a team. “Beyond the technical piece, we found a big need to train executives, mangers, and general staff in the understanding of how to use data science in general, how to think about that strategy, how to staff up,” Yuravlivker says.
The ability to tailor the Data Society curriculum to the specific needs of a client gives the company an advantage over education providers that have more rigid curriculums. This customization takes the form of industry-specific capstone projects in some cases, while in others, the real-world projects begin to resemble consulting engagements. (All training is now conducted virtually, due to COVID-19).
For example, Data Society recently helped the National Park Service with a project to determine which parking lots it should install electric vehicle chargers in. “They actually created a whole new revenues stream for their agency,” Yuravlivker says. “It was really fun to see that.”
Another customer, a Fortune 50 company, engaged Data Society with a plan to teach students how to develop text mining systems. “So we built a course that goes from entry-level programming through text-mining techniques,” she says. “That is just an example of how we’ve used their own data to pull out insights.”
Yuravlivker detects a shift in how organizations are thinking about data, particularly since COVID. They’re more interested in getting results from their data as opposed to investing in big data and data science for its own sake. But they’re still not sure how to go about getting there, and so they need a helping hand to guide them.
For Data Society, that means a return to teaching basics–even if her clients have other ideas.
“We have seen more of the introductory-level programs,” Yuravlivker says. “Everybody loves to know that we teach deep learning, but very few organizations are ready for it. So we’re working with a lot of foundational skills to build up that continuous culture of learning, building up that community of practice, and that common understanding of what it is.”
Yuravlivker is struck by the fundamental misperceptions that many of her clients and prospects have about the nature of data and data science.
“Honestly, there’s a lot of misconceptions about what data science can and cannot do,” Yuravlivker says. “There’s a lot of managers who just want to throw data science at everything without understanding the reason behind it.”
A lot of managers love the idea of applying machine learning and deep learning techniques to their particular business challenge, Yuravlivker says. But quite often, their challenges don’t warrant using those technologies.
“A lot of times, we’ve heard ‘Well, we need deep learning to solve this challenge,’ when maybe it’s just a classification problem, which is a bit of a different level,” she says. “Maybe it’s just, how can we have the appropriate data governance and standardization of data collection to help facilitate the process of data collection?”
The first step in data science is getting one’s data in order, and that tends to occupy the lion’s share of the data scientist’s time. Ensuring the data is consistent, clean, and high quality isn’t as sexy as, say, building a deep learning model. But it’s a critical step that cannot be skipped.
Those core data management techniques also form a core part of Data Society’s curriculum.
“As you know, garbage in, garbage out (GIGO),” Yuravlivker says. “If we collect bad data, if we put it in an algorithm, it doesn’t matter how good it is–the results are going to be wrong.”
Instead of transforming her students into deep learning experts or math whizzes, Yuravlivker would much rather they have a strong grasp of the fundamental issues impacting the data ambitions of real-world companies.
“The biggest lift, that biggest transformational value, is from that 0 to 1,” she says. “It doesn’t mean somebody needs to learn about neural networks. I would argue most people don’t need to know that. But what they do need to under and is how to use data, what it can do. They need to understand that data collection is 70% to 80% of any data science project.”
For companies and agencies predisposed to buy into the hype around big data and data science, Data Society’s approach is a dose of strong medicine. But if they follow the company’s educational approach, they’ll find themselves on a firmer footing when it comes to building their own data products.
This article originally appeared in Datanami.