Introduction
In a previous article, we saw howthe Medallion architecture enables the technical structuring of a data platform into Bronze, Silver, and Gold layers. This foundation is essential, but it says nothing about a crucial point: who produces the data, for whom, and according to what common rules ?
Having clean data in a Gold layer isn’t enough if no one knows it exists, if it remains confined to a central team, or if each department reworks it on its own. This is where three complementary concepts come into play: Data Democracy ( the vision), the Data Product ( the value building block), and the Data Mesh ( the organizational model). Together, they answer a simple yet central question: how can we ensure that data truly creates value, at scale, across the entire organization?
1. Data Democracy: putting data accessible
Data Democracy refers to an organization’s ability to make data accessible and usable by as many people as possible, regardless of their level of technical expertise. The underlying idea is simple: if data remains the exclusive domain of a few specialists, it cannot become a true driver of decision-making.
In practical terms, the democratization of data rests on three pillars:
- Easier access → The right people can access the right data at the right time using self-service tools like Power BI;
- Skill development → Business users are trained to read, interpret, and cross-reference data on their own (data literacy);
- A culture of curiosity → The organization actively encourages decision-making at all levels, rather than just within the data or IT departments.
However, we must be careful not to interpret this too simplistically: democratizing data does not mean giving everyone access to everything, all the time. Rather, it is about ensuring that “the right people have access to the right data, at the right time, for the right purpose,” while maintaining a balance between accessibility and governance.
When implemented effectively, Data Democracy delivers tangible benefits: faster decision-making, greater autonomy for business teams, and the emergence of initiatives driven directly by frontline staff. However, it also presents real challenges, particularly regarding security, data quality, and the consistency of metrics.
2. Data as a product: the concept of the Data Product
While Data Democracy sets the course, we still need to define what we’re sharing. That is the crux of the Data Product concept, popularized by Zhamak Dehghani as one of the pillars of the Data Mesh.
A data product is the application of “product thinking” to data: we stop viewing it as a technical byproduct of an operational application and instead treat it as a product in its own right, designed for users (the “consumers” of the data), whether they are internal or external. In practical terms, a Data Product brings together datasets, models, pipelines, and their metadata to serve a clear use case.
The Characteristics of a Good Data Product (DATSIS)
For a data product to truly fulfill its purpose, it must meet six key properties, grouped under the acronym DATSIS ( Discoverable, Addressable, Trustworthy, Self-describing, Interoperable, Secure).
A concrete example
To make the concept clearer, let’s consider an example: an “Overtime” data product designed for HR teams. This is not simply a table in a data warehouse, but a coherent and governed set of resources that typically includes:
- several Gold tables organized into a dimensional model: dimensions (entity, period, overtime reason, etc.) and one or more fact tables (number of hours, associated amounts, reference periods);
- a business interface that provides several ready-to-use views based on these tables;
- a clearly designated owner (HR Data), responsible for ensuring compliance with business rules;
- designated users: managers for team oversight, payroll for calculating compensation, HR for overall analysis;
- Commitments to quality: data freshness, completeness, traceability, and accessible documentation.
This captures the essence of the concept: a clear deliverable, a designated owner, clearly defined users, and quality commitments. Data is no longer simply “stored somewhere”; it is packaged as a standalone product that business teams can use with confidence, without having to understand the inner workings of the model.
3. Data Mesh: Scaling Up organizational
Data Mesh is an organizational and architectural model introduced by Zhamak Dehghani in 2019. It is based on a simple observation: as data sources and use cases multiply, the central data team becomes a bottleneck. It
is unfamiliar with the job, has to juggle dozens of tasks, and ends up falling behind.
The Data Mesh approach turns the problem on its head: responsibility for data should lie with the business teams that know it best. To achieve this, it is based on four fundamental principles: decentralized domain ownership, data as a product, a self-service platform, and federated governance.
This approach has been adopted by several leading companies such as Netflix, Zalando, Intuit, PayPal, and VistaPrint. Zalando, for example, implemented a Data Mesh to move away from a centralized data lake that had become unmanageable, giving each business unit responsibility for its own data in order to accelerate its personalization use cases. Netflix, for its part, has built a standardized self-service platform, which they themselves call a Data Mesh and which aligns with one of Dehghani’s pillars.
Advantages and limitations
The Data Mesh offers several advantages: organizational scalability ( the central team is no longer the bottleneck), business alignment ( data is managed by those who understand it), and team autonomy ( each domain delivers at its own pace).
It also has very real limitations. Above all, it is a cultural and organizational shift rather than a matter of tools: without strong executive sponsorship, it often fails. It can generate resistance—both from business units, which are saddled with data responsibilities, and from central teams, which fear losing their roles. If poorly managed, it can lead to costly silos and ineffective governance. Finally, it is not suitable for small organizations or contexts where few departments handle data at scale: the added complexity would then outweigh the benefits.
4. Data Democracy vs. Data Mesh: Differences and Complementarities
These two concepts are often mistakenly confused. Data Democracy is a vision: it describes what we are striving to achieve ( data accessible to everyone, within a controlled framework). Data Mesh is an architectural and organizational model: it describes how to practically organize data and teams to achieve this vision at scale.
The two are complementary: the Data Mesh is an ideal way to implement Data Democracy within a complex organization, because it combines local autonomy with shared governance. Conversely, a Data Mesh without a vision of Data Democracy can easily devolve into mere technical decentralization, without any real benefit to the business.
It is also helpful to contrast the Data Mesh with another concept: the Data Fabric. Whereas the Data Mesh is primarily organizational ( it changes who does what), the Data Fabric is
primarily technological ( it automates and unifies access to data through an intelligent layer, often powered by AI). In reality, many companies combine the two: a Data Mesh for organizational purposes and a Data Fabric for technical integration.
5. How do I get started in practical terms?
Adopting these concepts isn’t something that can be imposed overnight. It’s a gradual process that combines vision, organization, and tools. Here are a few key steps to get started:
- Clarify the Data Democracy vision → Define what “making data accessible” means in your context: which audiences, which uses, which rules;
- Identify your initial areas → Map out the key business areas (sales, finance, product, operations, etc.) that stand to benefit most from a Data Mesh approach, and clarify their scope;
- Form multidisciplinary teams → For each domain, establish a core team consisting of a Data Product Owner and one or two Data Engineers/Analysts, working directly with the business;
- Launch one or two pilot projects → Rather than aiming for a complete transformation, start with a high-value data product to validate the approach;
- Invest in the self-service platform → Leverage standard cloud building blocks to enable departments to become self-sufficient;
- Establish robust governance → Work with the departments to define common rules that everyone follows, ideally through automation;
- Train and support → Invest heavily in data literacy: without building the skills of business teams, neither Data Democracy nor Data Mesh can deliver on their promises.
In addition, the Medallion architecture discussed in the previous article remains fully compatible with this approach: each domain can adopt a Bronze/Silver/Gold structure for its own data and expose Data Products from its Silver and/or Gold layers. The two approaches complement each other: Medallion technically structures the data within a domain, while the Data Mesh organizes the domains in relation to one another.
Conclusion
Data Democracy, Data Product, and Data Mesh are not interchangeable buzzwords: they are three different approaches to the same goal—making data a true driver of value. Data Democracy sets the course, Data Product defines the unit of value, and Data Mesh provides the organizational model for scaling up.
When properly understood, these approaches transform data into a shared corporate asset: visible, reliable, well-governed, and, above all, actually used by business teams. They do not replace a sound technical architecture (such as the Medallion model); rather, they complement it by answering the essential question: “How should we organize our teams and our
"What practices ensure that data truly creates value?" In data, as in product development, it is this organizational discipline that makes the difference in the long run.
Are you looking for experts who can guide you in implementing a data strategy that integrates architecture, governance, and analytical insights?
Learn more about our Product Data/AI Practice: https://www.5degres.com/product-data/
FAQ: OKR-Based Management
What is the difference between Data Democracy and Data Mesh?
Data Democracy is a vision ( making data accessible and usable by as many people as possible within a controlled framework). Data Mesh is an organizational and architectural model that enables this to be achieved at scale by entrusting responsibility for data to business domains. The two are complementary.
What is a data product, in a nutshell?
A data product is a reliable data asset that is documented and managed like a product: it has an owner, users, measurable quality, and a clear purpose. It combines data, models, pipelines, and metadata to support a specific business use case.
What are the 6 characteristics of a good data product (DATSIS)?
DATSIS stands for a Data Product that is Discoverable , Addressable , Trustworthy , Self-describing , Interoperable , and Secure .
What are the four pillars of the Data Mesh?
Decentralized ownership by domain, data as a product, self-service platforms, and federated and automated governance. These four principles, established by Zhamak Dehghani in 2019, are inseparable.
Does Data Mesh replace the Medallion architecture?
No, the two are complementary. Medallion technically structures data within a domain (Bronze, Silver, Gold). The Data Mesh organizes data at the enterprise level. Each domain can certainly apply Medallion logic internally while exposing Data Products within a global Data Mesh.
Do you have to be a large company to implement a data mesh?
Data Mesh is particularly valuable in multi-domain organizations with a central data team that is already stretched thin. For smaller organizations or those with few data use cases, its complexity often outweighs the benefits: it’s better to start with the basics (Medallion architecture, initial data products, simple governance) before considering large-scale decentralization.
Nicolas Jacob Peres
Data Engineer
