Data Science is still a relatively young field, especially compared to Software Engineering and Product Management and I would argue that the community is still experimenting the best way to structure and organise data science in organisations of different sizes.
During the time we have been building Signal AI, and also while meeting many founders, I have seen people struggling with the reality that you need completely different profiles of people in different stages of a technology company. The requirements of a pre-PMF (Product Market Fit) company are completely different than those for an established international company expanding to new markets. This blogpost will provide my thoughts on the opportunities and challenges for three specific company stages and the profiles in the Data Science function that I think will maximise the success of the company.
These reflections will not be applicable for all companies out there as they are more relevant for VC-backed, product companies. Nonetheless, I hope there are some potentially useful tips for other types of organisations. Also, I will not discuss any aspect related to hiring or how to structure Data Science in an organisation because I have already shared my views on this in the past.
A company in a pre-PMF stage (usually a pre-seed or seed startup) is an exhilarating environment to work in. Everything moves extremely fast with constant iteration under a high level of uncertainty. At this point, the team tends to be focused on ensuring they are asking the right questions, solving the right problem and identifying the correct potential customers. The team will probably spend a large amount of time rephrasing their problem definitions, creating POCs and prototypes and validating feasibility of different technology solutions. The team is unlikely to have a rigorous and clear problem definition or experimental framework that will allow Data Scientists to optimise and improve a potential solution.
Given this chaotic environment, the most suited profile for people related to the Data Science function would be self-sufficient individuals with a relatively full-stack background who can own the whole process “well enough”. This includes understanding the potential clients and product offering, shaping (and reshaping) the relevant questions to ask and providing different potential ways of solving the problem. These people should have a product-first mentality, as well as high EQ to be able to navigate the communication challenges and constant changes in direction and while they need to be competent on Data Science fundamentals, they probably do not need, at this stage, to be specialists with a deep knowledge of one particular area.
PMF and Scaling
Companies will know they have found the beginning of PMF once selling their product suddenly becomes ‘much easier than it used to be’. At this point, there will be higher certainty about the team solving the right problem for the right audience, at least at a basic level. At this point (usually Series A/B for VC-backed companies), companies tend to spend some time improving their infrastructure and operations across the business with the goal of limiting the bottlenecks that might be limiting exponential growth. However, at the same time, the organisation is likely to be looking at new initiatives to provide even more value to the users. At this point, the company will be simultaneously working on high-iteration and problem definition (similarly to the pre-PMF stage) for some initiatives while working on improving and optimising others. From a Data Science perspective, this is a very interesting moment because it requires different profiles to work in the company. On the other hand, you will need some people with a similar profile as the ones described in the previous section to continue the exploratory work. On the other hand, for the more established offering and components, it is likely that you will build a reliable experimentation framework to be able to optimise your AI components and the derived value to users.
For the optimisation side, you should look for people who have a better system-thinking perspective and who prefer to work on a reasonably well-defined problem where they can use increasingly sophisticated solutions. At this point, specialised knowledge is critical in order to continuously improve the end-to-end product solutions. At the same time, this is likely the best moment to invest in Data Science infrastructure and support skills in order to improve the performance and impact of each person in the function.
As a company grows in headcount (especially if there is also a geographical distribution), communication, focus and alignment quickly become the most challenging aspects for an organisation. Related to this, people’s happiness, while being critical from day one, becomes more complex as the perceived impact and context about the organisation as a whole tends to get diluted. How to structure the company and the different functions plays a pivotal role and so does certain processes related to people. This is specially relevant for career progression and personal development paths.
This is when you need to have strong managers and leaders in the Data Science function (and across the organisation as a whole) who are effective at leadership as well as cross-team and cross-function communication in order for company-wide initiatives to be effective and for everyone to understand how their efforts contribute to the company goals. At the same time, given the (potentially very quick) headcount grow, people with high EQ will be extremely important to help others grow and navigate the necessary and potentially difficult conversations on career progression and personal performance.
Take Away points
My main takeaway is that you need to be clear what the goal for the next stage is and always be adaptable as things change. The goals, challenges and expectations will change significantly every two years or so and the company and the Data Science function must change accordingly.