From Concept to Mesh: Choosing the First Data Product

So you’ve embarked on the journey towards decentralised ownership and have secured Data Mesh buy-in? Then, you are probably already thinking about driving value quickly to maintain the momentum you have hopefully garnered thus far. And what better way to maintain momentum than to produce the first data product and tie it up with business value to show impact?

But how can you start creating Data Products? One of the first things you need to do is make sure everyone is on the same page and has the same understanding in your company as to what you mean by a Data Product. I’m not talking here about coming up with a formal definition of what a Data Product is. This can help but don’t get too stuck on that. Instead, explain what a good Data Product is versus a bad one. Team up with your Product people and let them help you list some real Data Product examples from the different domains. Do you have different categories of Data Products? List them and explain the difference. While some companies choose to have a very general concept of a data product, others like to detail it, e.g. operational Data Product, Analytical Data Product, AI/ML Data Product. Flesh out the answers to all the main questions and points of confusion you hear about — more on that in upcoming articles.

Let’s circle back, however, to the original question: How can you start creating Data Products?

The question opens up many more questions. Which Data Product should you start with? How can you begin creating Data Products? What is the minimum of technology, governance, process and literacy you need in place to start? Should you already start thinking about interoperability? All this and more.

In this article, I’d like to address the first notorious question.

Your First Data Product: How to Choose?

Fact is, we are not starting from Zero! Every company already has data that is creating value and is being used for critical business use cases. The Data has “only” yet to be productised. The word “only” here, being an understatement!

Picking the first Data Product is quite a strategic move that you should not take lightly. Your key stakeholders and those who might be sceptical of your push towards a Data Mesh will be looking very closely. So you need to show value that is tied up to the Business quickly and having that early success story is key.

So how do you choose? Here are some of my lessons learned and what I recommend when choosing that first Data Product.

1. New or Existing Data Product

Existing “Bad” Data Product: We are discussing our first Data Product here, so how come we have an existing Data Product? The reason why I say “existing bad” Data Product builds on the premise that we are not starting from Zero. I found this to be helpful to my pitch. An existing “Bad” Data Product is data that we know should become a Data Product in the future and has not yet been productised.

A New Data Product from Scratch: This is a data product that will be created for a new Business use case . That is, we don’t have a baseline we can draw upon when it comes to existing friction points or past bad experiences with this particular data format.

Guideline #1: To gain momentum and maintain buy-in, you are better off starting with an existing data that is already been in use and serving a business usecase and where there are already some existing pain points, which you can alleviate.

2. Upstream or Downstream Data Product

Attaching a business value and outcome to a Data Product can be challenging. If you start downstream at those final Analytics/AI/ML Data Products where the decision-making takes place, and the direct business value can be elicited, you can trace your path back to the source data and start there. This way, you are sure that the Data Product you are creating influences the final Business decision-making. From that point, you can have two approaches for the next Data Products you create:

1. Focused Approach: Work all your way downstream from that first Data Product, building all the data products needed along the way till that final analytics/AI/ML Data Product. The advantage of this approach is that you will have one solid example end-to-end, and you can show a very strong impact.
2. Holistic Approach: Work on the important upstream Data Products across the board. The advantage of this approach is that you include many Business use cases, and many stakeholders can already witness some value rather than hearing about your fantastic progress with another use case.

Guideline #2: Trace the way from the Final downstream Analytics/AI/ML data products to the Producers and start with that Data Product. i.e. Start Upstream.

3. Partnering Domain and Team

Partnering with the right Team is essential for the success of your first pilot Data Product and is a strategic move that lays the necessary groundwork for success. Having a motivated team can significantly influence the outcome positively. This is because an enthusiastic team can:

1) Absorb new concepts quickly, which is pivotal to an endeavour such as Data Mesh and Data Products, all of which are still shrouded with many literacy gaps.

2) Embrace challenges and collaborate on finding solutions rather than raising their hands in defeat.

Guideline #3: Partner with a motivated Team to work on your first Data Product Pilot.

4. Type of Data

Building your first Data Product will be a challenging feat. You will be testing things out, and you will have to retweak things. There will be lessons learned. And that’s fine. It’s almost like failure is the price of entry. With all this effort's complexity, I recommend staying away from the extra layer of complexity of having Sensitive data, as you would need to handle this on top of everything else. You can delay that to the third or fourth Data Product you are building.

Guideline #4: Avoid Data Products that would include sensitive data in those first two pilots.

Summary

Choosing that First Data Product is critical to maintaining momentum and securing your buy-in. When choosing the First Data Product to create, select 1) a Data Product where there are already some existing pain points that you can alleviate. 2) a Data product that can be tied up and bring a tangible business value. To find those products, one way is to start tracing your path back from downstream analytics/AI/ML Data Products to the producers, i.e. Start upstream. 3) Partner up with a team who is motivated and enthusiastic about creating a Data Product and who is willing to take chances with you and embrace Challenges. 4) Finally, start with a Data Product that is void of sensitive data.

Now that you have a better idea of how to choose your first data product. The next question is how to start building these Data Products.
In the next Article, I will share a Framework I developed for building data products one step at a time. Such a framework enables us to show business value quickly and maintain this much-needed momentum to retain that initial buy-in.