In anticipation of his upcoming Predictive Analytics World for Workforce conference presentation, Finding Top Data Scientists for Your Organization: Optimize the Hiring Process with Analytics, we interviewed Michael Li, CEO at The Data Incubator. View the Q-and-A below to see how Michael Li has incorporated predictive analytics into the workforce of TheData Incubator. Also, glimpse what’s in store for the new PAW Workforce conference.
Q: What is the specific business problem you are solving?
A: There is global demand for highly skilled data scientists across multiple industries and use cases. While data scientist roles are diverse, the skill-sets required are quite standardized — how to leverage free open-source technology for predictive analytics. Analytics tools can have a steep learning curve leaving managers to struggle for ways to deeply and quickly train their data science employees. We’ve found practical, extensive mini-projects force students to get their hands dirty and produce tangible results.
Q: If HR were 100% ready and the data were available, what would your boldest data science creations do?
A: The biggest untapped big data use case in HR is around hiring analytics. Most hiring is done solely based on the judgement of a single manager’s hiring experience even though making a bad hire is one of the most costly mistakes a company can make. Leveraging the data from the thousands of similar hires a company has made over the years can help inform managers to make better hiring decisions.
Q: When do you think businesses will be ready for “black box” workforce predictive methods, such as Random Forests or Neural Networks?
A: The predominance of certain models has more to do with a cultural familiarity (linear regression is taught in business school) than with how hard to understand a model truly is. One of the reasons companies invite us in to train their employees (both managers and individual contributors) is to broaden their organization’s familiarity with more advanced modeling techniques like Random Forests and Neural Networks. In the end, these supposed “black box” methods are not actually that much more opaque than so-called “white box” methods. Firstly, using tools like differential analysis and feature importance can give a lot of visibility on how more advanced models work. Secondly — and more importantly — our supposed understanding of how “white box” methods like general linear modeling work is more limited than many realize. In a multi-factor model, it’s hard to keep track of the hundreds of potential correlations amongst inputs — a fact that is often lost in the beguiling simplicity of a linear model. We emphasize both these lessons in our curriculum.
Q: Do you have suggestions for data scientists trying to explain their work to non-technical stakeholders?
A: Thanks for asking this question, Greta! The best advice about explaining data science to non data scientists I can give is to not explain models but to tell stories. Humans relate to stories — not math. Business stakeholders will often accept your analysis (as long as your results are reasonable) and do not care about the subtleties of statistical modeling. However, they will probe to make sure you have thought “outside the data” — looking at the limitations of your model or thinking about factors you were not able to consider — to make sure that your model makes sense in a broader business context.
Q: How does business culture, including HR, need to evolve to accept the full promise of predictive workforce?
A: HR is still a very gut-driven decision-making culture. To a certain extent, this is absolutely appropriate — there will always be a healthy room for human intuition when dealing with individuals. However, individual hiring managers need to realize that they are only directly involved in a relatively few number of hiring decisions at an organization and that those cases might not be representative of what will come. Ultimately, we have to be humble about the limits of our own personal experience and look to the data for more.
Don’t miss Michael’s conference presentation, Finding Top Data Scientists for Your Organization: Optimize the Hiring Process with Analytics, at PAW Workforce, on Tuesday, April 5, 2016 from 11:15 to 11:35 am. Click here to register for attendance.