In recent years, the advent of the data mesh architecture has popularized the concept of “data as a product”. The reality is this principle has been used for years in product-led organizations and it is independent of which data architecture you have.
In this article, you are going to discover what it means to treat data as a product and how to apply it regardless of how centralized or decentralized your data organization is. We will also cover its benefits and the guidelines your data teams can follow to embrace it.
What is data as a product?
Simply put, “data as a product” is the consequence of applying product thinking to data assets.
On one hand, we understand product thinking as the ability to identify what makes something useful based on capturing user needs first.
On the other hand, we consider a data asset - any piece of data that can be used to gain insights from your business. It could be a data table, a group of tables, a dashboard, a report, notebooks, etc.
In summary, data as a product considers data assets as standalone products that have value by themselves because it helps answer the questions decision-makers have.
Now that we have defined data as a product, we don’t need to mistake it with data as a service or with data products. Let’s understand the differences.
Data as a product vs data as a service
Treating data “as a service” means your data analysts directly deliver information and insights to decision-makers normally in the shape of a spreadsheet, a presentation, or an interactive dashboard fit for that purpose.
This makes customers of data to be really dependent on data analysts. It is a common trait of companies that treat data teams as a service function instead of a growth function.
On the other hand, data “as a product” advocates for discoverable and documented data assets. These allow well-trained customers to use them to at their own pace and come back to the data at any time without having to rely on a data analyst.
Data as a product vs data products
A data product (note the lack of “as a” in the wording) is a more generic term in the industry to refer to any product that facilitates a goal through the use of data - as defined by DJ Patil, former United States Chief Data Scientist in his book Data Jujitsu (2012).
For example, a personalized home page based on previous browsing behavior is a data product, a recommendation to buy a product after I put something on the shopping cart is a data product, autopilot capabilities are a data product, so on and so forth.
For further reading on this topic, you can check Data as a product vs data products. What are the differences?
Benefits of treating data as a product
Companies that embrace this principle benefit from:
Data is one of the only assets in the world that does not go away when it is consumed. For too many years companies have been neglecting this fact by creating lots of adhoc extracts, spreadsheets, and visualizations with limited scope. By treating your data as a product you will be creating a solid foundation of core datasets that are used across the organization.
It is ok for these core datasets to be reused multiple times for adhoc aggregations in different teams –actually, high adoption of these datasets is a sign of maturity and something you can use to measure the success of your transformation.
Autonomy for decision-makers
When treating data as a product, decision-makers are autonomous in informing a decision based on data. This is key for those organizations that envision self-serve analytics and want to use data as an asset and not just as a bypass for building a visualization.
Empowering analysts to do more
Your data analysts don’t want to be folks who just extract data and hand it over to business teams to make informed decisions.
They want to be closer to decision-making and contribute to the goals of the organization by performing advanced analysis and experimenting as much as possible. In order to do that, you need to make the underlying data layers easily accessible and documented by treating them as a product.
Building data teams as product teams
If you wish to change how your organization treats data, you should use these guiding principles to build data teams that treat data as a product:
Add a product management role
As your data infrastructure grows in complexity and as you increase the number of stakeholders being served by the platform, you will need to dedicate more time to product management activities. These usually entail understanding customer needs, prioritizing solutions, owning a roadmap, working on a mission and vision for the team, etc.
Add domain product owners
You need people who understand the data the business is managing and its potential. Your data analysts in domain teams will usually hold the role of domain product owners, so they need to be familiar with dimensional modeling and star schemas to do a great job here.
Get senior management on board
You need the senior management on board to transition into a product-led organization for data. Treating your data analysts as providers of data in spreadsheets will not provide a competitive advantage in the long term so you need to inspire the perspective of those who request those spreadsheets in the first place.
Level up analytics across the organization
You will need to teach decision-makers some basics about data visualization or even SQL. Prepare training materials that suit their needs and expectations so they don’t shy away from actually using the data. Teach analysts how to use and transform the data and your number of data-informed decisions will skyrocket.
Prioritize autonomy with alignment
Autonomous teams are great and necessary, but that does not mean that you can down-prioritize alignment. Working on and getting everyone aligned on governance and processes will make a positive return on the investment in the long term.
In this article, we’ve explored the main benefits of treating data as a product and the initiatives that companies have to prioritize to successfully form data product teams.
Organizations that have the right processes, people, and technology in place to create reusable datasets as products will have a competitive advantage - because all the time saved can be invested in more advanced data usage.
Written by Xavier Gumara Rigol
Want to maximize the value of your data infrastructure? Learn more about the future of the modern data stack