Most people know what product management is. Data Product Management is a brand new discipline in the workplace. When it comes to creating data-driven products, data product management is the best approach. This new approach led to creating a new position in tech companies, the data product manager. This article will explain data product management and how it differs from regular product management. What are the typical projects that data product managers manage? What skills are required for this job? Is your company in need of a data product manager? These are the questions that this article seeks answers to.

What is data product management?

Data product management is the strategic development, launch, and support of data products. To design the best data products, data product management must focus on users and not just data products.

What’s a data product?

You may be curious about what exactly these data products are. Data products provide value to the end-user through intensive use of data analytics, machine learning, and other data science techniques. A product management project at Spotify might be to create a personalized playlist that suits each user’s tastes. A visual search engine available as a SaaS via an app could also be considered a data product. The product could also automatically filter job applications by the human resources department.

The role of the data product manager

Without a data product manager, a project in data product management cannot be completed. Because he facilitates collaboration between people from different backgrounds and skillsets within the company, the data product manager is crucial. A typical data product team includes data scientists, data engineers, and data scientists. To design products that meet users’ needs, the data product manager must also represent them.

Structure for a data PM team

Two poles can be used to divide a data product team. The first pole is responsible for model performance. The second pole is responsible for activation performance. The first pole is responsible for transforming the raw data, building a model from the refined data, and tuning the model to obtain the best forecast. This pole is composed mainly of data specialists. The second team must create the most activation possible, i.e. The mechanism by which the prediction will be useful to the user. This includes the delivery of the model’s output, the display of recommended videos, and the activation of that output (clicking on one of those videos to view it). This is primarily the responsibility of the developers.

The two poles of model performance and activation success can be broken down into a data product team.

They cannot be used in isolation. The model chosen determines the activation. The activation is the reciprocal of the model. A platform that sells e-commerce products must display product recommendations almost immediately. The model must meet certain requirements immediately after activation (or time constraint). A good data product manager can deliver high-quality products by understanding and coordinating both of these poles.

Product Manager vs Data Product Manager

After reading the section, you might be curious if there are differences between data product managers and traditional product managers. The truth is that the roles are very different.

Similarities Between the Two Jobs

The duties of a data product manager are similar to those of a regular product manager. Prioritization is the number one challenge for both product managers and data product managers. In another way, both PMs must decide which problems need to be addressed first to create the best product. Both PMs must create roadmaps, compile backlogs, and present release plans. They also need to develop business cases and define metrics to determine the product’s success.

Two very different jobs

A regular product manager works at the intersection between engineering, business, and UX. However, a data product manager works at the intersection between engineering, data, and business!

Regular product managers are focused on satisfying the needs of their users through their products. A regular product manager will seek to bridge the empathy gap with his users to design the best product. Data is rarely used twice to validate features and clarify insights. Product managers do not have to be experts on data. Knowing how to interpret user feedback, understand common visualizations, and perform A/B testing are enough.

On the other hand, a data product manager will tell you that data wins arguments. A data product manager will make decisions based on experiments and quantitative assessment. Data product managers’ products are also fuelled by data, as we have already mentioned. Data product managers must keep up-to-date with the latest data science developments to effectively manage their teams. Here are some examples of the many tasks that a data manager can oversee:

Finding new data sources: identifying data sources that can be helpful for your business (e.g. competitor data).

Data serving: Creating new data pipelines for other departments

Optimizing existing process: switching between manual ID recognition and automated recognition

Cross-selling product recommendation: New feature implemented

Introducing a new internal metric: providing a new forecasting model for the finance department to advance analytics

Watch for news and research: staying on top of new models and trends in data science.

You can also get detailed information about the Product Manager Roles and Responsibilities.

Example of a Data PM-led project – Cross-selling product Recommendation

You should now have a better understanding of data product management and its differences from regular product management. It might be worthwhile to search for a project where a data PM can lead. This section shows an example of an eCommerce project where a data PM could implement cross-selling product recommendations.

Cross-selling product recommendations: cross-selling is a sales technique that allows customers to purchase additional products similar to those already purchased (i.e. products in the shopping basket). Studies have shown that customers spend more on e-commerce platforms when given cross-selling product recommendations. This type of recommendation is essential for e-commerce platforms. This is an example of a product management project. This project involves building a recommendation engine. Recommendation engines heavily depend on data science techniques like matrix factorization or deep learning tools.

A typical roadmap to building a product is broken down into sprints. A sprint is simply a time when ideas are transformed into tangible action. A sprint usually lasts for one month. An interaction phase is usually followed by the end of a sprint. This phase updates the goals and sets new goals for the next sprint phase. Now you will see how a data PM develops a product in five sprints.

Sprint 1:Analysis work for fundamental questions

  • Are current product recommendations efficient?
  • Are cross-selling recommendations likely to significantly increase the platform’s net revenues?

The data PM and his team hypothesize each question. Each question is then answered by the team using past data. Each answer may also generate sub-questions. The team should answer these sub-questions until they are clear.

Sprint 2: The first sprint’s analysis usually yields significant results. Let’s suppose that the team discovers that the platform could benefit from better product recommendations and product cross-selling recommendations.

The data PM decided to create an MVP (a small recommendation system that allows cross-selling) to gather as much information and feedback as possible. Also, the team tests it with current users to determine where these new recommendations should be placed on the website. Data PM develops new hypotheses that can be tested. Data engineering provides all data required for the recommendation. The recommender is built by the data science team (MVP). The frontend team conducts A/B testing with actual users to find the best place for recommendations.

Different teams build and create metrics to optimize recommender usage during this sprint. This optimization can be achieved by great analytics of these metrics. It may take time and multiple iterations to reach the right product features.

Sprint 3 – Let’s assume that the results of the previous sprint were significant but not exceptional. Cross-selling recommendations increase net revenue on average. The increase is not significant. The frontend team also identified two locations where recommendations would have the greatest impact. The product page is the first, while the checkout page is the second.

Due to the improved recommendation system, the slight increase in net revenues is not enough to make the data PM confident that the sprint phase will see a greater increase. Data PM asks data scientists for their final recommendations by the end of the next sprint. The data PM also asks the frontend team to focus on the two most important hot spots for A/B testing: checkout pages and products.

Sprint 4 Both locations (product pages and checkout pages) will see significant increases in net revenue. The team also has sufficient data to calculate user retention. The retention curve measures how many users are losing over time as they use the product. Analytics clearly shows that users who access the new recommendations have higher retention rates than those without access. The recommender is now fully functional.

Now, the data PM can decide whether to make cross-selling recommendations for all users.

Sprint 5 The data science team monitors the model’s performance and performs additional checks.

This example shows how data PM can go from an easy investigation (are my product suggestions efficient?) to the complex development of a new data product. The development of a completely new data product (cross-selling suggestions). It was also evident that sprint 3 saw the data PM continue product development, even though the results weren’t as impressive as expected.

Each feature in data products is often the result of multiple iterations. The data PM was able to help its team achieve the best product during sprint 3. He focused on a feature that would activate at a specific website location.

A Data Product Manager’s Skillset

I hope you found the following sections interesting and useful in understanding data product management. Perhaps you are already interested in a career shift and want to be a data PM and would like to know how to do that. This section outlines the skills required to become a data pm.

  • Learn the life cycle of data products

You must first understand how to manage the lifecycle of data products. Keep in mind the impact your products will have on the business. You should also know which teammate you should ask for and what actions to take during each sprint phase. Sprints require data preparation, hypothesis, modeling, evaluation, interpretation, deployment, optimization, launch, and maintenance.

  • High-tech technical knowledge

Second, you need to improve your technical and analytical skills. Data PMs should be knowledgeable in statistics, computer science, data analytics, science, and the tools associated with these fields. A data PM should be able to explain the pros and cons of each machine learning model. As mentioned, data PMs should constantly learn new techniques to provide the best products. Data PMs are also problem solvers. To solve the problems you face, you’ll need to have great analytical skills. This knowledge must be updated regularly. It’s best to plan time to learn new information if you want to be a data PM.

  • Develop emotional intelligence

Third, you need to improve your emotional intelligence. Emotional intelligence is a combination of self-awareness and self-regulation. It also helps with motivation, empathy, social skills, and self-regulation. Your products and your team will benefit greatly from great emotional intelligence. Emotional intelligence will help you close the empathy gap between your user and you. This will lead to better products that truly meet customers’ needs. Second, emotional intelligence can positively impact your leadership, collaboration, interactions, and leadership. This will lead to a better team and a healthier work environment. This skill allows you to give and get quality feedback. Feedback is essential for improving the user experience and learning.

Data product managers require a wide range of skills, leading to a steep learning curve.

Data Product Manager or a Data Product Designer?

A data PM might be necessary for you to make a decision. This section will provide guidelines for determining if you require a data PM.

First, you need to ask whether data fuels the products you sell to customers. Products or features fueled by data may include voice assistants, personalized recommendations, and fuel consumption optimizers for Formula One cars. Hire a data PM if your products/features fall into this category.

If data is not driving your product, you need to ask a second question. This is a question about whether data heavily drive your internal operations. Is your business generating a lot of data that must be stored, processed, and analyzed? Do you need forecasts for a department in your company? Is your company facing many logistics challenges? If you answered yes to both of these questions, your company is highly data-driven.

A data PM is needed if you have data-driven operations. A PM can do the job if you don’t have data-driven operations.

You might also be interested in understanding the ethics of product management and its needs.


Write A Comment