A data product is a solution that takes advantage of data assets to provide tangible value (removes a pain point or delights with benefits) to end-users.
The emphasis is on the value generation for product users, not the “solution” aspect.
Traditionally, data teams focused on the “solution” part, tweaking data pipelines that extracted volumes of raw data from data sources, optimizing data models for analyses, or building SOTA machine learning algorithms for predictive analytics. Often, this focus overshadowed the primary goal of addressing specific business problems or delivering additional benefits.
Zhamak Dehghani was the first to argue that the data mesh architecture puts data teams into a product development role:
“For a distributed data platform to be successful, domain data teams must apply product thinking with similar rigor to the datasets that they provide; considering their data assets as their products and the rest of the organization's data scientists, ML, and data engineers as their customers.”
Does that mean a dashboard can be a data product? Absolutely. To further understand this, let's delve into the various types of data products.
What are the different types of data products?
With data productization emerging as a new field, experts have various ways to classify them:
By functionality: Such as recommender systems, visualizations, and anomaly detectors.
By deliverables: Like source-aligned datasets, customer-aligned tools, or aggregate metrics.
We believe there is a better way to understand the different types of data products.
The types of data products should be organized by the value they bring to the end user:
Dataset as a product: When you serve raw data to end users, you equip them with the ingredients to develop their data products. Examples include APIs for data export, data warehouses, data lakes, or even spreadsheets detailing competitor pricing.
Data insights as a product: Instead of sharing raw data, you already extract insights from the data as metrics or KPIs. Examples are self-service dashboards for data-driven business decisions, Google Maps suggesting the quickest route, and smartwatches converting sensor data into heart rate metrics.
Action as a product: These data products take insights and turn them into the next best action. Examples include product recommender systems, self-driving cars, and optimization machine learning algorithms that pack cargo shipments.
Product generation as a product: Products like ChatGPT and DALL-E fall into this category. They act as "product factories", generating non-data tools valuable to users. ChatGPT, for example, can draft business plans, write poetry, or offer legal advice.
This might all sound abstract. So let’s look at concrete use cases of data products in real-world industry case studies.
Leading industry examples of data products
Example of a dataset as a product: Price discounts to lower costs for consumers while lowering product waste
Rohlik, a major e-commerce food retailer, faced a challenge common to those selling perishable items like fruits and dairy: the products' short shelf life.
Using Keboola, Rohlik integrated various data sources to create a dataset. This dataset took into account current stock, historical sales, expected sale duration, and other factors.
This information was then utilized to introduce dynamic discounts, where products nearing their expiration date were priced lower. These discounted products were showcased separately for the benefit of the buyers.
The outcome? Rohlik minimized waste, consumers enjoyed better prices, and there was a positive environmental impact.
Example of an action as a product: recommender algorithms boost sales by 720% for Olfin Car
Olfin Car, a prominent car seller in the Czech Republic, offers not only new and used cars but also services like financing, car maintenance, and insurance.
Facing stiff competition in a saturated market, Olfin Car turned to competitive intelligence to gain an advantage. Partnering with Keboola, they automated the collection of competitor pricing and product offerings, ranging from other resellers to manufacturers.
By integrating this competitive intelligence data with historical supply and demand trends, online behavior patterns, and more, they trained advanced recommender algorithms. These algorithms intelligently adjusted product recommendations on Olfin Car's website to maximize conversions and sales.
The outcome? A staggering 760% revenue surge in just one quarter.
These are just some of the examples of how data products helped companies grow. Let’s look at the wider advantages of data products.
The benefits of data products
Data products offer many benefits to their consumers and the company that develops data products.
Benefits for business teams (data consumers):
Informed decision-making: Data products empower business teams with access to relevant, real-time data, enabling them to make more informed and data-driven decisions.
Cross-functional collaboration and alignment: Data products facilitate collaboration among different domain teams by providing a centralized location (aka the product) for data sharing, analysis, and reporting.
Streamlined workflows: Teams can use data products to solve tailored business problems faster. Leaving them more time for revenue-generating work.
Independent problem-solving: Before data products, business users had to rely on IT and data experts to provide them with the data and insights needed to do their jobs. Data products are self-serving, making business users independent and liberating technical resources for more revenue-generating work.
Benefits for the business itself:
Revenue growth: Data products can uncover new revenue streams, improve pricing strategies, and increase customer retention, ultimately leading to business growth.
Data monetization: Businesses can monetize their data assets by creating data products for external customers or partners, generating additional revenue streams.
Cost reduction: By automating data-related tasks and optimizing operations, data products can lead to cost savings in areas like marketing, supply chain management, and customer support.
Competitive advantage: Leveraging data products can provide a competitive edge by enabling businesses to adapt quickly to market changes and deliver better customer experiences.
How to start building your first data product?
Building data products is a continuous process. The product development lifecycle goes through these stages:
Understand the value that needs to be delivered: Analyze the problem for the potential product user and evaluate the value the user would get if the product was delivered.
Build (a partial) product: It’s important to move fast and build the minimal possible product that would deliver value.
Test value delivery with end users: There are many methodologies to assess value delivery including user interviews, user experience tests, and product analytics. Ensure to understand if and where the product you built offers value. Identify the gaps and re-address new user needs.
Sadly, most companies become bogged down in the initial stages of product development, such as creating data science algorithms, designing a metadata schema to increase data quality, or ensuring their federated access control complies with data governance policies. As a result, they aren't able to iterate and progress quickly.
That’s why Keboola built tools to get you building data products faster:
Data apps allow you to instantly deploy data insights as products and data actions as products with a couple of clicks. For example, launch a digital marketing dashboard, collect feedback with a customer survey app, or predict e-commerce sales.
Data templates allow you to launch pre-built ETL pipelines that generate datasets as products with a few clicks. Data templates generate datasets for digital marketing platforms (Facebook Ads, Google Ads, etc.), organize your CRM customer data in one data model, integrate project management (Jira, Asana) tickets in one place, etc.
The data science workbench offers integrations and features to build, optimize, and deploy machine learning and AI models into your product development pipeline.
The sandbox environment allows you to play with data without breaking down production pipelines. Giving you the freedom to test and experiment with novel data product ideas.
Both data apps and templates come pre-built with out-of-the-box data governance, data quality controls, data access management, automated deployment, and all the back-office work. So you can spend more time focusing on the value of product development and less time on the overhead.
Or, you can use the pre-built solutions as a blueprint and tap into the wide ecosystem of Keboola’s features (Streamlit integrations, data science workbench, CDC, etc.) to build your data products.
Build high-quality data products with Keboola. Faster.
Keboola helps you set up and deploy data products in minutes with its data apps, data templates, and ecosystem of product development features.
So you can spend more time on value delivery and less time on the engineering overheads.
Oops! Something went wrong while submitting the form.