ABSTRACT: Treating data assets as products helps businesses make better use of their data. This article provides guidance for finding product-market fit.

Business users don’t trust data. Data consumers can’t find what they need. Analysts spend more time shaping data in Excel than actually analyzing it. We’ve all heard the complaints. But why do they persist, year after year, even as data technology advances? Because the root cause of these problems isn’t our technology—it’s our framework for data management. 

The root cause of data problems isn’t our technology—it’s our framework for data management.

If we view data as a product, we start to see that the core of many of these issues is a lack of product-market fit. Data assets don’t align with “customer” needs. Too often data teams deliver a commodity when they should provide tailored data assets for different use cases. Take lack of user trust in data. Fundamentally, that’s a data quality problem, but different users have different expectations for “quality” data. The features that make data usable for marketing analytics vary from those that allow it to be used for fraud detection. In the first case, the data needs to be well sanitized, in the second, cleaning the data would destroy evidence. One size does not fit all. 

The same logic applies elsewhere. Why can’t consumers find the appropriate data in a sea of assets? Because no one consulted them on what data they actually use. Why do analysts spend most of their time preparing data that’s already been cleaned and modeled at least once before? Because data engineers didn’t think about their use cases. Ultimately, these groups have to jerry-rig existing products to meet their needs. It’s like a hammerless carpenter using a wrench to drive a nail. The demand and resources are present. Companies just need data teams to create new products instead of better wrenches.

So how do data teams create valuable data products? They follow these seven steps:

1. Start with consumers. Every company generates data. Who uses it? What questions can it answer? This step requires good, old fashioned communication. Data teams need to talk with down-stream consumers to figure out where the data goes after it hits the repository. In fact, these consumers might already be creating their own products on an ad hoc basis that the data team can productionize for the rest of the organization. It’s also critical to talk to business stakeholders about the kinds of questions they’d like to answer but can’t with the data assets currently available to them. What are their pain points? What don’t the assets already available address? Make a list of requirements and assign a product manager to lead development of the data product.

2. Identify the market. Once you have an idea for a data product, it’s time to start measuring the addressable market. Who all might benefit from the product you envision? What is its overall value to the company? Does that value justify the costs of creating, delivering, and maintaining it? Bear in mind, this is an iterative process, and your idea of the market will evolve during the process of fine tuning the data product.

3. Release a prototype. Using the insights from data consumers, create a new data product and make it available with the data product manager as the steward. Consider rolling it out to a limited number of consumers at the beginning and then making it more broadly available as you work out the bugs. This process will help you further judge the feasibility of creating the product.

4. Open feedback channels. Technology can really start to facilitate the process at this stage. Publishing the product on a data catalog or internal data exchange can create a formal mechanism for users to provide feedback in the form of comments or requests. These tools aren’t strictly necessary, but it’s a heck of a lot easier to keep comments and assets linked on a single platform rather than trying to remember which emails go with what file.

5. Create a collaborative space. Ideally, invite key stakeholders into the product development process. A tool that allows data providers and data consumers to work on the data collaboratively in a single, secure, virtual environment is ideal. But, depending on the organization, huddling in a cubicle together might work too. That way, analysts can show engineers what they actually do to the data once it’s provisioned.

6. Expect to iterate. As every writer knows, first drafts should never be final drafts. Repeat steps 3-5 until your product meets the demand. You may realize that what you first believed to be a single product actually needs to be multiple products. That’s okay. Fork the project and let another data project manager lead the charge to fill a different niche. At the same time, refrain from letting earlier versions of a product clutter your organization’s archive of data assets. At the end of the process, you only want a finished product to hit the metaphorical shelf.

7. Publish and repeat Once you’ve honed your product, make it available to the broader organization, and start working on the next one. Use the notes from your first attempt to save time on future products. Documenting conversations, prototypes, and feedback will allow you to develop other data products faster because you won’t have to start from scratch. The process of developing one data product may inspire the creation of several others, as your data team develops a better understanding of different consumer groups. It’s important to focus on creating one product at a time, but that doesn’t mean you can’t revisit the decisions you made later in developing new products.

Figure 1. Data Product Development Workflow

In this article, I mostly looked at data product-market fit in the context of internal data distribution, but the same principles apply to producing data products for third-party consumption. As data marketplaces grow, and CDOs take on the challenge of monetizing corporate data, strong product-market fit becomes a way to stand out from the competition. What leads data consumers to choose one asset over another on an open exchange? Well, it’s the same as in any other market—fit.