Data is collected everywhere, all the time. At the CUE Spring conference, Dr. Zev Eigen, the Global Director of Data Analytics at Litter, showed us how data collection and analysis has evolved and shared best practices.
In the past, statistics were the best available option for predicting human behavior. However, humans are unpredictable, and today’s workplace is diffuse and rapidly changing. For example, let’s assume that someone eats hot dogs every day for six days. Statistics would predict that the person would eat another hot dog on the seventh day, but better data and more robust analytics could predict that the person would grow tired of hot dogs and choose a different meal—anything BUT hotdogs!
We used statistics because we didn’t have a lot of available data, computing power, and analytical tools. But computer power and analytical tools are getting better all the time. The field of data science is advancing rapidly. Even five years ago this area was akin to comparing a horse and buggy to a car. Now data analytics tools are available that are more like the car. And, just like you don’t need to know how to build a car to drive it, it’s not necessary for you to know how to build algorithms in order to effectively use available data analytics software.
There are four distinct data-driven services:
We collect a lot of data in companies, but we don’t always know how to use it. For example, let’s say that you already know who the good/bad performers are on your sales team. Now you want to be able to predict who is a good hire. You can take sales and performance data and create a machine learning model based on certain output criteria.
For example, Zev shared that he can weave together data sources from individuals, publically available data and internal company data to predict employment outcomes.
There is so much data available in the workplace that it’s often hard to identify an area on which to focus. But this is important since it’s usually easier to choose one area for improvement before moving onto others. Zev recommends picking the biggest “pain point” and working on that first. For instance, does your company do a good job at preventing turnover of high-performing employees? Does your company do a good job at retaining diverse talent? Etc.
We collect a ton of data on our employees. By mining data such as badge swipes, email or calendar meta-data, and digital communication, we can learn:
Data analytics can show how often employees interact with each other, the types of interactions they have, and improve communication and collaboration. So-called “relational data” measure networks of individuals within organizations. “Organizational Network Analysis” software like Syndio measures this information and makes the output actionable. You don’t even need to be a sophisticated technologically forward company to benefit from this!
Data science has proven useful for improving the hiring process, identifying high-performing employees, and even identifying employees who are more likely to quit. But it’s a double-edged sword. On the one hand, there’s a myriad of public data available via social media, that can be part of a data science initiative. On the other hand, privacy laws in Europe, and US anti-discrimination and equal employment laws may pose legal risks. Gartner, Inc. predicts that by 2018, 50% of business ethics violations will be related to data. Littler predicts that by 2020, most privacy causes of action will be related to data.
Data science isn’t as simple as plugging data into some formula. It requires multiple data sources that are carefully examined, cross-validated, and rigorously tested. It’s important to engage a data science expert who can help you execute the project and navigate the available software options.
Dig into the discussion with a software vendor. For example, if you ask what data analytics software does, the answer is usually that it uses artificial intelligence to make predictions. If you ask how the software does that, you are likely to be told that it uses a proprietary system or algorithm. But that’s not always a satisfactory answer. Many algorithms are not proprietary; they are often publically available—which is sometimes a good thing! You might prefer standardized methods over something built by a startup for the purpose of being unique, but perhaps not necessarily the best for your organization.
Ask about data validation, too, by asking if the system was validated by a third party, and reviewing the study. Finally, you can ask for indemnification, but it’s unlikely that the vendor will grant it to you.
Dr. Zev J. Eigen combines his expertise in labor and employment law with his deep experience in complex data analytics and social scientific research to handle three categories of work: (1) predictive analytics using artificial intelligence or “machine learning” algorithms as applied to HR and related business decisions; (2) statistical analysis and econometric modeling of issues arising out of class actions and patterns and practice matters; and (3) statistical analysis of labor, employment and HR related data. He is a nationally recognized expert in these fields appearing in the media frequently and is a frequent speaker on matters pertaining to workplace data analytics as well as a recent recipient of three awards for innovative data scientific solutions in the last year. Follow him on Twitter: @zevdatascience or contact him at email@example.com
Guest blogger Liz D’Aloia founded HR Virtuoso to help companies optimize their employment application processes. HR Virtuoso creates customized, company-branded short form employment applications that work on any mobile device. This allows companies to get far more applications, and also keep their existing applicant tracking system. Prior to launching HR Virtuoso, Liz rose through the ranks of transportation, retail, and mortgage companies as a Senior Employment Attorney and VP of HR. Liz is a nationally recognized blogger, speaker, and HR practitioner.