Data science develops with a complex process. There are other software products that may be developed through a simple process that starts with a brainstorming stage and eventually become prototypes and launched into the market. However, a completely trained math model is much more intricate.

There are new features in the field of data science, being the Data Science Product Manager the main of them, since it is the connection between research and Return On Investment (ROI). Plenty of companies are aware of the importance of the results for their data science endeavour, and they are strengthening their focus on this aspect.

The reason is that the amount of projects that succeed is very discouraging, although the potential of data science and machine learning is very auspicious.

Combining abilities

Product Managers (PdM) work with product lines, since they are allotted one of them and they deal with its profitability. In the particular case we are approaching, the PdM will manage the growth in profitability of technical applications regarding product lines. In this field, the data science PdM requires technical knowledge.

This technical knowledge does not mean they need to be data scientists, but they are expected to recognize and categorize the different kinds of business with ease, together with the technical challenges which could be settled through data science or machine learning.

The requirements that a PdM must have are a deep understanding of mathematic modelling and firm familiarity with former applications. By way of example, in the case of a product line with an image recognition element, the data science PdM would be expected to know that convolutional neural networks (CNNs) are a useful solution in these kinds of business issues.

On the other hand, there is no need to know all about the most recent advances in generative adversarial networks (GANs) or in abilities to implement a CNN, just an insight on identifying the right types of problems so that the data science team of professionals can deal with them.

However, the solutions that the data science team makes must be endorsed as well by the stakeholders and executive decision-makers. In the case of data scientists recommending the use of GANs for image classification, the PdM requires translation abilities in the aspects of reading, understanding and interpreting the supporting research underlying in the recommendation.

Then, after translating the underlying information, a data science PdM will need to transfer all the research into a comprehensible presentation for non-technical people, so that they can be able to make decisions about the recommendation.

Communication Skills

A data science PdM has to interpret and translate all the information that has to do with data science and business necessities in order to present a plausible situation and the reasons to carry out a project to their audience: the executive staff. For this reason, every aspect must be taken into account: market assessments, the suitable moment for products on the roadmap and even budgeting.

Research and development are equally important in data science. The nature of the research phase is essential so as to achieve an approval to carry out the presented plan.

However, this required research cycle is not easy to adjust in standard projects and product management. It is a demanding sequence that may go through failures or might need expansions to analyse the latest advances or further areas of research. The research outcomes are effective but they also pose new issues to be explored in further work. Businesses need productive results at the end of projects to keep on working in a successful way, but research does not always work from the prospects of the researchers.

This situation entails a struggle between the requirements of businesses and the research demands. Therefore, a data science PdM requires communicative and facilitating skills to balance and assess the research process at the same time as overseeing the prospects and expectations from the executive staff.

There are meetings between the data science PdM and the executive staff where the PdM shares all the information about projects in their presentations and discussions. In order to have successful meetings, the data science PdM must have excellent translation skills to express the business requirements to data scientists and to convey research to stakeholders.

Further skills that are important in the data science PdM role are planning and creative abilities. Data science products have different kind of requirements since their models have a different nature as well. The models in this field clean inputs in order to produce the awaited production.

Information about the expectations of customers must be gathered in order to analyse the level of adequate accuracy. Another aspect to take into account is customers’ notion of tolerable and non-tolerable breakdown. As a conclusion, expertise is needed in the role of data science PdM with a view to take data science projects on the way of production.

On the way of production

Many prototypes do not success in the development from the prototype stage to production. As it is mentioned before, data science has a complex processs of development and prototypes are still far away from entering the production stage. It is the role of data science PdMs to use their abilities in order to turn prototypes into produced projects.

Non-technical users do not notice or take into account the models operating in the backdrop. This situation causes prototype failure, since they do not operate as users presume. Data science is extremely important in the field of product prospects, but it is not the same case from the point of view of the users’ prospects.

Hence, we can state again that the PdM has the role of a translator, since they need to plan the productization that approaches both utility and user perspectives. The output must be presented in a way that users perceive as useful and valuable. All aspects regarding visualization or the layout of the interface, together with accuracy, are extremely important at this stage. Expertise in achieving prototypes’ arrival on the market is needed.

In short, the job of a data science PdM is highly demanding and semi-technical. It is mainly based on the translation role in order to convey information about research to ease the decision-making process, convert business needs and succeed in the prototype-production stage to satisfy the customers’ expectations.

Otherwise, without the good practice of a data science PdM, some projects will fade and weaken on the way of production instead of turning data science potential into earnings.