To say that businesses are dealing with a data mountain is an understatement. Data generation has increased exponentially over the last few years, increasing by at organizations with at least 1,000 employees. Customer data is no exception. Behavioral data, marketing data, stock levels, engagement — vast amounts of information are constantly collected and stored. So, what should you do with this glut?
Many companies have an almost dragon-like approach to data collection, hoarding as much as they can, based on the idea that at some point in the future it will prove useful. Others simply monetize their data by selling it on to third parties. But the smartest businesses know that data is only an asset when it’s used to achieve strategic business goals. These businesses are adopting a new mindset — one that moves away from perceiving data as a static, often inaccessible asset, towards seeing it as a product.
Data as a product? Or data products? You’ve probably heard both terms, and they’re closely linked. The ‘data as a product’ approach borrows a simple principle from product thinking: start with a need, and then work backwards. In essence, if you’re creating a new product, you need to know that somebody wants to buy it. A product that doesn’t serve a purpose or overcome a problem has no value.
Historically, data use has been dictated by technical demands, limitations, and processes, rather than business needs. This leads to data being difficult to access, hard to consume and tough to share. Treating data as a product requires establishing a clear need for the data to serve from the outset, and then organizing and managing it to meet that need.
Data products are the manifestations of this approach — the culmination of processed, analyzed, and curated data sets designed to provide solutions or insights. They could take the form of an analytical warehouse, an AI model, a user-friendly dashboard, or any system that transforms raw data into an actionable asset.
What about data mesh?
Another term you’ll often hear in relation to data products is ‘data mesh’. This term describes a federated approach to data management, as opposed to the more traditional method where data is centralized, and often controlled solely by the IT data department.
Data mesh seeks to reopen the gates, giving data access and ownership to people across the whole organization, and encouraging a more collaborative approach to how data is managed, used, and shared.
To adopt data mesh, most businesses would need to instigate change at an operational level, with a shift in culture, structure, and the creation of new roles. Data products are a key facet of data mesh, and usually the easier principle to start on that journey. Even for organizations that are not willing to embrace the operating model and structural changes required for data mesh, it is still valuable to apply “data as a product” thinking for how the central data team manages their data.
Data products: key principles
To distinguish a good data product from just another data dashboard, there are certain principles you should apply:
Data products put the customer first. Whether for internal or external use, delivering value should be your guiding star. Before tackling the technicalities, you should be asking yourself, ‘who wants or needs this product?’.
Discoverability is also key. While your data product may be created to answer a problem within a specific business function, once published, it should be readily available to anybody within the organization.
Scalability is closely related. Data products shouldn’t be single-use. Ideally, they will serve multiple use cases, so while you may start off with one goal or use case, data products should be able to flex, grow, and continuously serve business needs.
A data product should be usable. Presenting data in an intuitive, consistent, understandable way makes it easier for a non-technical audience to consume.
Users need to be able to trust the data they’re served. So, all data products must follow a common governance standard within your organization. At we apply the DATSIS principles (discoverable, addressable, trustworthy, self-describing, interoperable, and secure) to ensure quality and consistency.
Where to start with data products?
If you want to introduce data products to your business, start by finding an existing use case. You may already have an aspirational idea in mind, or a burning problem, bottleneck or customer sore point that needs attention. If you don’t have a use case in mind, reach out to your business teams — particularly those that are more advanced in terms of their data usage — and ask them 'if you could achieve something today through better data use, what would it be?'.
You’ll also need to assemble a product team. A data product’s success relies on breaking down the silos that typically divide teams by different functions. By bringing together a cross-functional team of people with complementing skills, you bridge the gulf between those who collect and analyze data, and those who need to put it to use — uniting them towards a common goal.
So, once you’ve identified your value driver, kick off with what’s known as the 'thin slice' approach. This is a term that describes a way of working to achieve end-to-end functionality faster. Technical solutions typically involve multiple layers of functionality, and the tendency is to work on one layer at a time, perfecting it before moving on to the next. Thin slice means focusing on a narrow cross section that passes through every layer to bring an end-to-end product to delivery sooner. By slicing the development process into manageable slices, organizations can quickly test and iterate on feedback. This approach minimizes risk, maximizes learning, and ensures that the data product is aligned with the actual needs of users.
Data products and AI
Making use of AI to improve the customer experience, increase the operational efficiency of your business, or generate new value is likely to be high on nearly all CXOs’ agendas. If your goal is to create an AI model and embed it in your business, data products can help in two ways.
Firstly, they will give you high quality data on which to train your AI models. Remember that what you put into it is what you get out, so quality is key for better results. Data products help improve the input into your AI models.
The second aspect addresses operationalizing your AI models into production. Data products can come in different shapes or flavors, with multiple types of inputs and outputs. One such type is a data product that embeds an AI model. They take data inputs from other data products, encapsulate the pipeline for training and evaluating the model, and serve the AI model for real-time inference as its output. The platform you build to support data products accelerates the path to production for your AI models, enabling more advanced predictive and prescriptive analytics use cases.
Unlock business value
CXOs find themselves increasingly at the forefront of technical change and innovation in their organizations. To be the agent of change for your business, make data products the tools that move you forwards:
Turn raw data into useable data — that can generate new business value or solve a business challenge.
Improve discoverability — if data is designed to be self-service, users no longer have to rely on data specialists to access or understand it.
Boost trust and reduce risk — everybody is using the same high-quality, up-to-date data. This also enables collaboration.
Offer flexibility through reuse — data products ROI increase when they support multiple use cases, accelerating processes and providing insights faster.
Ensure compliance — with governance that's consistent, and built-in from the start.
By embracing a modern data product strategy you can improve the quality, trustworthiness, security, and accessibility of your data, and ultimately unlock more value from it. Data products help you become a data-driven organization.
