Today we’re featuring insights from Shiftgig’s Chief Data Officer, Sean Casey, to share his perspective on the evolution of the Data Science field, and how it factors into Shiftgig’s continued success.
When Shiftgig evolved to an on-demand labor model in 2014, we set out to establish ourselves as a company that could grow and scale using our own data, but also knew that we would have to wait for enough of it to accumulate to really unlock that potential.
By the end of 2015, we had a large enough data set to begin leveraging machine learning and we assembled a team of data scientists. Today, Shiftgig’s Data Science team is a strong mix of PhDs and self-taught data scientists, with a focus on harnessing this data and helping make our business stronger, faster and better.
As we begin to kick off 2017, we’re expecting to see the work we’ve put into place really come to life in the first half of this year, so it’s a great time to look back on what our team has accomplished so far, and what’s next – both for Shiftgig and for the field of Data Science.
Laying the Groundwork
The primary goal of our work is to help make both the product we’ve built and our business operations more successful in a data-driven way.
With traditional temporary staffing, there is a short-sighted focus: filling a shift in advance. What this model fails to take into account, however, is whether this shift will actually be worked successfully, often leaving this unknown until it is too late to intervene.
We quickly realized, however, that just showing up is not enough – for both our clients and our Specialists – and we really needed to define success more clearly. So Shiftgig set out to build ourfirst predictive model to tell us if a shift was going to be successful or not, meaning our client had their need filled, our Specialist got paid, and the shift had been worked.
We wanted to begin with risk mitigation to better predict the number of back-ups needed for any given shift. To assess risk, our model takes into account a wide range of data, including how Specialists use our app, past shifts, geography, pay rates, and every other data point you can imagine. The model then helps us predict these different risk factors as places to intervene.
For example, it might be able to show us that certain shifts are most successfully filled when our Specialist doesn’t work the night before, or that for some events, we need to automatically include a backup Specialist.
When we’re able to predict some of these potential risks and flag them for early intervention, it allows us to communicate with our clients more effectively and help them plan ahead for what to expect.
Our model can also look for variations in the cities in which we operate. For example, a Specialist in New York City may only want to travel two to three miles, whereas our Specialists in parts of Texas are often willing to travel much farther.
Every shift is different, so we can’t treat them the same way – a Friday evening shift is going to be very different from one on Monday morning. The duration of the shift also makes a difference, so perhaps we might advise a particular client that a single 12-hour shift should be split into two six-hour shifts to help drive a higher fill rate.
It’s been an important shift for our business to prove that this is possible – that we can predict human behavior in certain cases and do it in a way that will help support our internal teams for even better results.
Expanding Our Reach
Generally speaking, Data Science is difficult to hire for. There are far more jobs than there are people who have the ability to make an impact at a company. We’ve put together a strong internship program where we can bring on people who come in through non-traditional routes, like boot camps.
Cultivating our own team and giving them hands-on experience helps us solve this problem in a way, but we’re always looking for great people to join our team. More specifically, we look for individuals who like to ask questions, and who have a lot of integrity.
When you’re dealing with large quantities of information, there’s a lot of power in that – you can make it say whatever you want. A good Data Scientist should be someone who can follow the data, rather than forcing it to tell a certain story.
Shiftgig’s team is very collaborative. We work in an agile way by picking each other’s brains for what we think a trend will mean, the best way to approach a hypothesis, and giving and receiving peer feedback regularly.
It’s also important to me that our team consistently collaborates with other parts of the company, learning where we can be useful and what business challenges we’re trying to solve. Many times there are ways to address those challenges with our work, which is not only rewarding for us, but also helps us continue to push toward our goals of positively impacting the business.
Looking Ahead Through 2017 and Beyond
Many companies are still figuring out what Data Science means for their business, perhaps assembling data teams simply because it was the “thing to do.”
In the coming year or two it will become table stakes for a business to use advanced analytics. We’ll see that companies who adopt this mindset create real value and move to the top of their industries, while others will be left behind.
As our influence is growing and expanding, I’m inspired by the impact Data Science can make to help Shiftgig to empower millions of people to find meaningful work, and make the experiences of our clients so much better.