I love how life works and things work out! One quick conversation with someone, led to a quick series of emails, and I got introduced to Dr. Michael Housman, who is the Chief Analytics Offier at Evolv, and what a great conversation! He is being very humble below, but he has his PhD in Applied Economics and Managerial Science from Wharton! Plus if you go to Youtube – he has been on tons of news shows being interviewed (e.g. CNN, Fox, etc.). Basically, it was a pleasure to meet Michael.
1. Michael, I may be using the wrong term – but how did you decide on and then become an industry leading Data Scientist? Additionally, Data Scientists are in high demand for almost every industry – so why did you decide to focus on HR?
I started off down this path by studying organizational behavior and safety culture in hospitals. I learned that when doctors and nurses work well together, there is a significantly lower incidence of infections, readmissions, medication errors, and even “never events” like wrong site surgeries and retained instrument post-operation. So improving workplace culture through the use of data science can literally save lives!
At Evolv, we use data and analytics in order to optimize workforce profitability at all stages of the employee lifecycle: (1) hiring; (2) development; and (3) plateau / separation. Even as I’ve moved afield from health care, I’ve found certain themes that are present regardless of industry: when you bring in better raw talent, train them well, build positive workplace relationships, and create a strong organizational culture, you inevitably see significantly better workforce outcomes that can be a significant driver of workforce profitability for large companies.
I love my job and I’m pretty sure that I have the coolest job in the world because I get to study what makes people engaged at work, what keeps them longer, and what allows them to reach their full potential. Through the use of big data and predictive analytics, my ultimate goal is to create a happier and more engaged workforce.
2. If you don’t mind, can also explain a little about Evolv and your role there?
At Evolv, we use data and analytics in order to optimize workforce profitability at all stages of the employee lifecycle: (1) hiring; (2) development; and (3) plateau / separation. The idea is that we inject data into human capital management in order to replace decision making based on gut instinct and intuition with a data-driven and evidence-based approach to talent management.
I’m the Chief Analytics Officer at Evolv so I run the Analytics team, which is responsible for applying econometric techniques to analyze client data, performing academic research, producing White Papers, and engaging in R&D and new product development. So my team is constantly sifting through our massive data warehouse in order to identify interesting trends and insights that we can deliver to our clients in order to create value.
3. It seems amazing to write this, but it seems like predicting turnover, and quality hires are now becoming “normal” capabilities. So what is the next “frontier” for you or HR to tackle?
Although more organizations are engaging in predictive work, I will just say that not all predictive models are created equal. For example, the majority of clients and vendors in this space still use turnover models whereas we believe that there is a class of survival models – derived from epidemiology and econometrics – that is far more accurate and robust than dated turnover models. This is something we often education our clients about.
Likewise, we find more and more of our clients moving beyond turnover and thinking about the performance of their workforce on various KPIs. But as with turnover, there’s a wide range in terms of the models that they’re using to analyze performance. We find that traditional BI dashboards with trend lines and moving averages are staring to pave the way to more advanced econometric methods like panel data models that can predict performance in the future.
4. I am sure you have many case studies of where your systems generated great and accurate predictions, but if someone doesn’t do something with the numbers, it is for naught. So can you share a story where you saw a HR team take the numerical outputs you provided and did some amazing “HR” to address the issues?
We had a client approach us once with questions about their overtime policies. They found that they were asking more or less overtime from their employees – dependent largely on holiday season ramps – and wanted to know the effect on employee tenure and performance. We did an analysis of their overtime data and found that there was a sweet spot: 1 to 5 hours of overtime per week was ideal but any more or less than that led to people leaving more quickly and performing poorly.
We delivered this insight to the client and they re-vamped their overtime policies to ensure that their employees stuck between these guidelines. They gave their employees opportunities to engage in some overtime but then capped it unless they had permission from their supervisor. We’re still gathering data on the outcomes of these changes but they’re on track to save a total of $10M annually from this insight alone.
5. Last question, what is the coolest, wildest or funniest analysis or outcome you have worked on?
We stole an idea from Google and give our Analysts “20%” time to allow them to work on projects that are interesting to them but also have potential to yield value to the company. One of our Analysts was studying the impact of someone’s technology footprint on their tenure and performance so he looked at browser and social media usage and we released those results to media outlets.
I think he had a couple hours to spare on Friday afternoon and decided to parse people’s e-mail addresses. So he looked for words like “sexy,” “crazy,” and “boozy” in the e-mail handles that applicants use when applying for jobs and found that all of those terms were associated with shorter tenure and poorer performance. The lesson: don’t use slang or colloquial terms in your e-mail address when you apply for a job.
For the record, we don’t use any of that information in our scoring algorithms because our hiring criteria needs to be job relevant.