Thoughts about Product Management and Data Science

In the recent year's topics like data science, machine learning (ML), Artificial Intelligence (AI) and Predictive Analyzes gain much popularity, and probably each tech company is looking on what it can do in this area to be up to date.

While trying to learn and adapt all these new fields, I would like to suggest and share a bit of different angle about some analogy between data science domain and Product Management.

This article brings some thoughts about the similarity in phases and process between modern data science and Product Management work developing a new product or exploring new markets.

"The first commonality between Product Management and data science is data "

The first step building and suggesting a model in data science is data, and when you start working on a new Product, you need data. You need to collect data. You need to analyze it. Data "cleaning", looking for anomalies, filling missing items, Normalizing the data. Data and Data Processing are typical jobs for both data science and Product Management, and to be successful in both area, you need to learn and know how to do it.

While trying learning and building some data science model, I realized how this step is similar to the data step in Product management. While data science can use tools for data visualization, show graph, calculate and present mean and variance, manipulate data, use statistical tools as they have in Python or simple excel. The Product Manager can use powerful tools or models to visualize the data about the market and customers. An example can be Porter's five force's (analyzing business and Competitors) or RWW (Real Win Worth). In both fields, processing, and visualization of the data is an essential step for a better understanding, and to reach the right conclusion, but also that you can share this and communicate It with your peer and managers.

"Data Visualization is an important tool for both Data science and Product manager." 

Visualization gives you the option to see things differently. It can inspire you to develop a new feature or a product (if you are a Product Manager) or direct you to the right direction of selecting your model if you are a Data science. It is vital that your Visualization will be truthful, functional and insightful but also you want it to be beautiful.

Finally, you want your data Visualization to be Enlightening, this is probably a combination or the rest of the other factor, but with some social and ethical ingredient. 

Nowadays communication is essential. No matter if you are a Product manager or a data science you need to communicate and market your ideas to others, colleagues or managers, if you create and develop a great data visualization and keeping the above guidelines not only it will help you to the next step, you will be able to communicate your ideas efficiently and successfully.

The next step is starting to design, develop and build your model (Data Science) or your Product. In this step I find a great commonality too, the terms are common, both disciplines use the name features if you build a product or a new version you need to select your features and prioritize them, and if you are a data science the input to your model is a set of features. 

"Selecting the features and prioritization is one of the keys to a successful product or a model." 

Product management deals a lot with the features, arranging them in a backlog and selecting which features to do first, and which will be delayed or will not be implemented at all. It is one of the cores and the art of Product Management. While learning a bit about machine learning and deep learning, this seems to be also a critical factor in building your model, right selecting of features can help you create a successful model and avoid spending time and resources, on a feature that will provide small contribution or no effect on the model.

Another place of commonality is the design and implementing phase, in both fields, it is essential to build and introduce the first product or model. In Product management it is usually called MVP (Minimal Viable Product) in data science I did not find an equivalent term, but based on what I have learned, it is recommended to start and iterate and build your first model quickly, to get preliminary results. 

The last commonality that exists in these two areas is managing and releasing new versions once the product is released. The first version of the product, as well as the first version of a Machine learning model probably is only the starting point. There are more iterations to come to improve and release, better versions, it is a journey of try and error, to create a new and better version of the product or the model.

My key take away from limited initial experience with data science is that the two worlds are close, I think that every product manager should be familiar with these new fields of data science, and even if your current role doesn’t immediately can make use of them it can highly contribute to your work as a product manager.