Hiring a team has always been a difficult task, especially when you have to deal with the uniqueness and novelty of the data science field. Here is the “who does what and for how much”.
Data science is a collective term that pretty much contains everything that is related to the data world. Specifically, when statistics and programming come together to solve data-based problems.
As a brief reminder of what data science is, here one of its definitions:
Data science is a combination of different fields of work in statistics and programming that seeks to extract insights from large amount of data
Seeking these insights requires knowledge of concepts and methods such as Big Data, Machine Learning, or cloud computing. But also curiosity, creativity, or communication skills are of much value. These skills are rarely covered by only one person. And unless you find the data science “unicorn”, you must assemble a team.
The skills required to serve a data science project vary by its size, depth, and complexity. Rather than hiring full-time employees, freelancing experts can be brought on board to temporarily close specific knowledge gaps.
Yet, finding suitable team members is not easy. The data science skills are high in demand, but short in supply. Many companies look to hire talents to run their digital transformation programs, including data science projects but often fail to do so.
According to LinkedIn’s 2017 U.S. Emerging Jobs Report, data science roles are the ones with the highest growth predictions. Moreover, Glassdoor confirms that they are among the top best 50 jobs in America (and probably worldwide, too).
In the long term, the famousness of these jobs may actually solve the demand-supply problem.
Which data science roles do you need to look for?
There are several roles in the data science universe. Each of those roles is an evolution in itself. They changed over the years to mirror the technological and market requirements. And there is no clear definition of these roles. This can be irritating not only for companies trying to hire the right talents but also for the people who try to gain a foothold in the data science field.
One of the clear distinctions, however, is the place on the skills spectrum ranging from technological, through analytical to business skills.
DataCamp has composed a great infographic on who does what in the data science world.
If you have a limited budget, consider at least three roles for your data science projects: data engineer, data scientist (aka machine learning expert), and a business analyst.
The first, data engineer, is responsible for the technical infrastructure (databases, servers) and skilled in using tools such as Hadoop or MapReduce. He or she is also mandated with getting and preparing the data (storing, cleaning, transforming). It is one of the fundamental roles.
The second role, data scientist, has ideally advanced skills in statistics, building models, and programming languages such as Python, R, or Matlab. This role is the link between a data engineer and business analyst.
The last role, business analyst, represents the business side. He makes sure the business objectives are well understood by the data science team. Skills in data exploration and visualization must be on an expert level. Often, it is the business analyst who translates the insights into actions.
Important to remember is, that a data science project relies on teamwork. Staying only within own role responsibilities and not looking beyond will only cause frustration. This is something the manager must constantly cultivate.
How much should you pay for data science skills?
The remuneration for those roles varies greatly. Factors that have an impact on the salary are usually: country, experience, and gender. The 2017 survey by Kaggle (16 000 responses) uncovers some of the truths in the data science field. In the USA, for example, being a female with more than 10 years of experience in this field pays the most. The median here is $151,000 per annum. In Europe, the median pay is around €50,000 – €70.000 (for all, no matter how much experience or which gender).
As mentioned in this post, building a data science team is a long-term budgetary commitment, starting usually at around $1 million (team of 4-6 members). That number depends on the country, industry, and experience of these members. But only looking at the table below shows the threshold is reached quickly (excluded are other costs: infrastructure, licenses, subscription fees, etc, the list goes on and on).
Hire teams not unicorns
Rather than looking for the data science wunderkind, spend your recruiting activities on building a team that covers most of the skills needed to accomplish the work. It is a better approach to look for people who excel at one thing and only feel comfortable with others. For example, hire a software developer who knows how to build stuff, even if his communication skills (at C-level) are not that well developed. Data scientists with in-depth mathematical (statistical) skills are more valuable than one with some programming and some mathematical skills. Hire teams that complement each other.
If you start building a team with a limited budget, having mostly junior staff on board, hire an external freelancing senior data scientist who will guide your less experienced members. Here, specific knowledge transfer goals must be set and constantly monitored.
Keep in mind that sooner or later you will be measured against the RoI (Return on Investment), therefore valuable and actionable outcomes must be delivered quickly. As a rule of thumb, consider a 3x3x3 model: 3 months to prove what is possible (throw some quick wins in the schedule), 3 months to produce the planned outcome, and 3 months to take it to market and integrate with the rest of architecture, if necessary. Senior management buy-in should be your main goal.
Also, before going outside of your company, have a look at your existing workforce. They may not have official data science titles, but some of them are perhaps skilled to do the required work. In some of the great tech companies such as Google, Yahoo, eBay, or Facebook, numerous “unofficial” data scientists were constantly discouraged by the work they had to do. They had to face product managers with different priorities, where the proposed solutions did not fit the product roadmap. The solution was to make these people responsible for their own products, where they could experiment and build products, and eventually create value for the company.