Source: itproportal.com
In 2005, the National Science Board began advocating for a data science career path in an effort to ensure experts could manage the inevitable onslaught of future data. While the practice of data science reaches back to the 1960s, and includes several older disciplines like statistics, it wasn’t until these past 15 years that companies began to integrate data science into their daily operations and decision-making in a more comprehensive way. While integration has grown exponentially, data science is still an emerging field, and enterprises around the globe are now competing to find ways to transform data into profit.
As the new industry emerges and higher-education scrambles to adapt, the demand for expertise significantly outweighs the supply, resulting in a competitive job market with some of tech’s highest salary expectations. Although some top tech companies have the luxury of operating multi-tiered robust data science teams, most companies rely on lean groups of individuals to produce insights relevant across all sectors of the business—research and development, sales, marketing, operations, customer service, HR, and the list goes on.
Most companies have outgrown the myth of the 10x engineer, but it is also crucial to foster realistic expectations of data scientists. To maximise the output of a small data science team, companies must embrace the structures and resources necessary for them to thrive and generate impressive ROIs. Here are some simple tenets to follow.
1. Hire generalists, not specialists
Most small businesses can’t immediately afford to hire a specialist for each position—engineers, analysts, data scientists, visualisation specialists, etc.—and also don’t want to risk damaging their output by understaffing. While the ideal data science team is made up of specialists, the solution for small teams is to either hire (or become) generalists, not specialists. Larger teams may employ a Director of Analytics or a Chief Data Officer to communicate findings and run the team, members of small teams must have communication skills dispersed throughout the group. Furthermore, without a visualisation specialist, small teams must familiarise themselves with the necessary programs and software.
Creating a generalised team may actually increase job satisfaction by creating pathways for autonomy, mastery, and purpose. Many data scientists are drawn to the field by the satisfaction of solving problems and creating an environment that allows end-to-end mastery satisfies this need. Of course, companies will have to seek out this type of candidate: one who feels comfortable working independently on a smaller team.
2. Stick to the (user) story
Businesses have long focused on putting the customer first, with the philosophy being that comprehensively meeting a customer’s needs generates positive financial results. Data science teams don’t directly interact with consumers; their “customer” is the business unit looking for help with a specific problem.
To improve efficiency and output, focus on user stories. Stakeholders are often on the front lines, dealing with dilemmas directly impacting the bottom line, and by maintaining a sharp focus on partners’ goals, data science teams become intertwined with a company’s success and accountable for the impact.
Whether user stories are written by the stakeholder or the developers, it’s essential that they are documented. Similar to how nothing gets done without a deadline, powerful insights won’t be generated without first identifying a clear business problem.
3. Encourage inter-team mentorship
Training is expensive, timely, and completely necessary in this rapidly-expanding field. High-functioning, high-output teams don’t rely entirely on outside sources for their training; they rely on each other. Openly encourage knowledge-sharing, create the necessary time, and develop infrastructure to support this goal. Offloading certain technical skills can create opportunities for team members to expand into new skill sets without needing to hire new talent.
4. Develop scalable processes
Early data science projects are exciting for any company, and achieving initial results may have team members working frantically with occasionally reckless abandon. Winning the day is important, but high-output data science teams must operate at scale or else growth—the core goal—becomes burdensome.
Processes should be documented both in real-time and after the project’s completion during post-mortem meetings. The aim is to create repeatable, measurable processes that answer relevant questions like: Which features were added at different points in the production? Or, at what point does the model need to be updated?
5. Rely heavily on SMEs
It’s troubling enough that data scientists must be experts in analytics, programming, statistics, big data, cloud computing, IoT, artificial intelligence, and machine learning, when it comes to topics like inventory management or international marketing, let other experts step in. Don’t overburden a small team by requiring experience outside of the data world. While it’s important that the candidates can work with the marketing and sales team to disseminate information and offer data-driven insights, they do not need to know how to run a marketing campaign. Subject matter experts provide invaluable information that’s not obvious to the untrained eye. Partnering with SMEs is critical for small data science teams to create meaningful impact. Actively seeking in-house experts and outside partners significantly improves possibilities for success.
The never-ending feedback loop
Technology experts are no longer siloed away from the general population; they are integrated, working across all business sectors and tiers within an organisation. Recent trends have companies of all stripes labelling themselves at tech companies—fashion, dining, finance, and beyond.
Successful small data science teams integrate feedback from colleagues and customers in order to hone their craft continuously. Just like other functions within a business (customer service, marketing, or sales), the more responsive a data science team is to feedback from within their own team and other outside groups, the more resilient they become. Foster this feedback by making it part of the culture. Companies that keep the data team siloed, especially when it’s a small data science team, may find it difficult to put insights into action. Those enterprises that allow for this never-ending feedback loop, however, will find more success and more output from a small data science team.