At the beginning of 2020, it was estimated that humanity had accumulated 44 zettabytes – that’s 44 trillion gigabytes – of digital data. And we’re continuing to produce data at a stunning and still-accelerating pace.
Businesses across a wide variety of industries have quickly learned that making effective use of the mountain of big data at their fingertips is no longer just a way to get a competitive edge – it’s a necessity. Companies that are data-driven are more profitable than their competitors, business analysts say.
But collecting, cleaning, and analyzing all that data for actionable insights isn’t easy. In fact, for many companies, it’s posing an enormous challenge thanks to a global shortage of data science talent. According to a report from IBM, rising demand for data skills and the shortage of skilled data scientists to fill the gap are likely to result in disruptions to global product development. Filling data scientist roles is taking longer, on average, than filling other roles – far longer, in the case of senior data science positions.
And that’s despite the fact that data science salaries are, on average, quite high. Even offering excellent compensation and benefits aren’t helping companies meet their data science skills needs quickly enough.
How can companies solve this problem and ensure that they’re able to keep up with data-driven competitors? Here are a few ideas:
Not all job boards are created equal. Data science candidates know they’re in high demand, so they may be less likely to spend time on “big” job boards like LinkedIn or Indeed. Why would they want to sift through hundreds of quasi-relevant job postings when they can find better, more curated jobs on boards like icrunchdata?
Getting your job postings onto smaller, more focused job boards may mean that they’re seen by fewer potential candidates, but the candidates who do see them are much more likely to be a better fit. When posting jobs, consider a more targeted approach, and you may end up with a much better pool of candidates from which to choose.
In general hiring, it’s common practice to filter applications using their credentials. Companies can be hesitant to hire applicants who don’t have, for example, a relevant degree.
In data science, though, these kinds of credential criteria can be quite limiting. Few universities offer data science programs, and while a plethora of data science certificate options are out there, there is no industry standard. Moreover, data science is a fast-changing field. A “data science” degree from even five years ago might have focused on technologies that are now considered outdated.
Some of the best data science candidates may be career-changers with degrees and work experience in totally unrelated fields. They may hold online learning certificates or no credentials at all.
Because it’s a relatively new and fast-changing field, assessing data science candidates on credentials simply doesn’t work. To increase your chances of finding the right candidate in this very competitive field, relax your credential requirements and focus on the demonstrable skills that applicants have, regardless of how they learned them.
Still can’t find a data scientist? Depending on the specifics of your company’s data needs, it may be easier to offer training that helps existing employees learn data skills. Online platforms can make learning data skills including SQL, Python, and R accessible for even non-technical employees.
This approach does have its limitations – it would take a long time to get an existing employee to the point where they could offer skills comparable to those of a seasoned data scientist. But in some cases, that level of skill may not be necessary. As little as a few weeks of part-time study can give anyone the ability to perform some powerful analyses on large data sets.
For some companies, this can be a temporary measure that allows them to get some value from their existing data while they continue the search for a data science hire who can take things to the next level with more advanced skills like predictive modeling.