5 skills key to be a good Data Product Manager

The demand for Data Product managers is on the rise. As data becomes central to more and more applications, these PMs form a key glue between the data ingestion, and its usage by applications and users alike. After spending more than a decade in product roles for data intensive applications including personalization, chatbots, ad tech, consumer data platforms and insights products, my few learnings about what it takes to be a good Data PM.

1. Get intimate with your data

It is very important for Data PM to understand how the data is being acquired, what transformations is it undergoing and what the data is being used for in the end. In one of my positions as Data PM, my boss had suggested that I spend the first month creating a data map of all the sources, data elements and reports of the advertising data platforms I was managing. I did this begrudgingly. But I soon realized the value of it, as I could immediately give context to the business partners about the results they were seeing, identify anomalies and recommend models and visualizations. Now, in any new role I take on, I spend considerable time in understanding the data stores, and transformations in the products I manage.

2. Get ready to get your hands dirty

As a data PM, you should be comfortable writing queries and pulling data from all kinds of databases. Being able to query data to either measure a behavior, quickly understand the data attributes or to diagnose an issue seen in prod is paramount. No you don't have to be a Ninja Data Engineer. But it is important to be able to write and execute queries for different kinds of data stores whether relational or Hadoop or Nosql. While the datastore architecture might be vastly different, the query structures are similar and easy to learn especially with myriad tools available right now. And let me tell you a little secret - for more complex queries, I usually ask my dev counterparts to give me sample queries and modify them as required. Frankly speaking, they will be happy to do so to keep you from troubling them with all data requests.

3. Be comfortable asking questions about data and output 

I have worked with many data scientists, analyst, engineers and people lot more smarter than me for building AI applications, creating predictive models, or even estimations. In my experience it is important to trust your gut. If something doesn't seem right, then don't be afraid to ask questions. This goes back to 1 & 2, if you know how your data is collected and stored, and you are not afraid to get your hands dirty, it will give you confidence to either double check or ask the right questions to verify if the output is correct. Taking a course or two in either analytics or ML will also help in understanding the process of model building to see if there is model over fitting or biased sample data. The truly good data scientists and engineers always appreciate another set of eyes as the cost of a mistake in production can be fairly high. But be humble and nice while doing so.

4. Don't forget visualizations

Another key aspect of data PMs is having an understanding of how data should be visualized to make sense to the users. Analytics products are different from typical transactional products as there is no set workflow for a successful transaction. Analytics products are required not just to provide succinct summaries in the form of charts and trends but also allow exploration of the data by allowing drilldowns, follow relationships between data objects etc. Easier said than done.

5. Keep an eye on your end customers

Sometimes being a data PM would mean that you are deep in the platform team where the consumers of data could be other applications. But it is imperative to understand the usecases the data is supporting. It could be for showing the right ad for maximizing yield using click stream data to exploratory analysis with normalized data for internal teams or charts and trends for KPIs using summarized data. Knowing the end usecase will help making the right decisions for all stages of data flow.

Of course, all of these skills are in addition to key product management skills of prioritization, stakeholder management, communication and problem solving skills.

Last but not the least, below are my favorite resources as Data PM. May the insights be with you.

Designing data intensive applications by Martin Kleppmann

Story telling with Data by Cole Nussbaumer Knaflic

Machine learning by Andrew Ng

KD Nuggets

Do leave your feedback if you found this helpful and add any other relevant points in the comments!