Of course, every concept has its strengths and weaknesses, and the data mesh is no exception.
In this article, we will shed light on the conditions under which data mesh makes sense and what framework conditions a company needs to create for successful implementation.
What is the data mesh approach suitable for?
The concept is essentially based on the fact that all data in the company is the responsibility of cross-functional teams that understand both the specialist nature of the data and have the technical skills to process it. The teams form interfaces with each other and make their data available to other teams in a high-quality form. This creates a mesh of teams and coordinated data streams. The advantage is obvious: when teams assume overall responsibility for their data and make it available to others, a high-quality basis for data-driven decisions is created.
A data mesh approach is therefore a very good tool for companies in which the sharing and reuse of data is of central importance. Be it to develop data-based business models, equip products with intelligent functions or optimize internal processes based on data. The decentralized data mesh approach is particularly advantageous when a large number of data sources and users are available, as regulations and roles bring together data supply and demand at a technically structured level.
What requirements must a company fulfill for Data Mesh?
The data mesh approach requires thinking in terms of (data) products instead of projects and disciplines. In particular, data responsibility in cross-functional teams has a deep impact on the organizational structure. The roles and responsibilities of these teams are at odds with the strict disciplinary thinking of traditional organizational structures. For example, projects are traditionally planned as projects with a fixed start and end date, after which project members return to their line tasks. Managing a data product, on the other hand, is tantamount to a permanent line task. In the best case scenario, the roles and duties of the data product teams are anchored directly in the company processes so that a corresponding level of commitment becomes clear. The (gradual) establishment of data product teams is a task for the entire company and must be accompanied by a strategic decision and support from all management levels. This means that the management must be aware of the implications and be willing to do so. It takes a long time for the awareness of creating value from data to bear fruit. And it may take even longer before revenues or savings from data products become measurable.
The basic steps to becoming a digital company, as described in the Blogpost "Step by step to a digital company" can also be considered with regard to a data strategy. Step 4 in particular looks at the development of a "data culture" as part of a digitalization strategy and corresponds to the changes in awareness within the company that are relevant here.
How can you start a data mesh approach?
Implementation can take place step by step, when a data product team starts and other data products from different areas gradually follow. Binding roles and processes should already be defined for the first data product team in order to ensure its ability to act.
Start with a simple data product
It is important that the hurdles for the change towards a data mesh are as low as possible. It is advisable to start with a data product that requires little explanation and is used repeatedly in numerous company processes, such as customer and supplier master data. As experts, the product team can disseminate their knowledge and experience to other parts of the company and thus serve as multipliers.
Consider learning effects and involve users
Further data products can be successively developed according to this model. Learning effects from existing data products should be taken into account in further developments. It is particularly important to introduce potential users to the data offerings during the development phase. This can be done through training at various skill levels, advice from experts and the provision of information about data products, processes, roles and guidelines. The presentation of best practices in data use increases awareness of the value of data in the company.
Create motivational incentives - demonstrate advantages
In the beginning, it can be helpful to set certain motivational incentives for the development and use of data products in order to make the sharing of data a matter of course.
Continuous monitoring of data usage shows which data products are particularly in demand and can serve as best practices for further developments.
Build up the necessary know-how in the company
In terms of staffing the data product teams, it is necessary to provide skills in data architecture and data processing from the outset so as not to overload IT capacities. If these skills were not previously available in the company, they must be provided at an early stage and with foresight so that the development of the data mesh concept does not fail due to a lack of expertise.
Access authorizations: Can anyone and everyone simply see all the data?
The need-to-know principle usually prevails in companies, meaning that data is only shared and accessible if knowledge of the data is demonstrably necessary for the fulfillment of a task. In the data mesh concept, the focus is on making data available for more comprehensive use. This paradigm shift must be incorporated into the company's data strategy and include guidelines for its use.
The basic attitude of sharing data with others does not mean that all data products can be used and combined by everyone in any form without control. Processes for regulating data access are important in the data mesh approach and allow that not everyone can see every data product or that not every data product can be combined with another data product. This is important because different views of data may have been separated into different data products for data protection reasons and the subsequent combination may otherwise undermine this regulation.
It must be possible to map transparency regarding access authorizations throughout the company's entire tool landscape. This is the basis for trust and therefore the willingness to share data.
How do (potential) users cope with the data mesh approach?
As the data mesh approach is essentially based on decentralization, the involvement of potential users is essential. Information must be provided on regulations, roles and, above all, on the data products. Training is advisable at different skill levels and for different perspectives. Advice from experts can also be made available on a low-threshold basis.
Creating transparency about existing data offerings in order to utilize synergies
Existing data products can only be used if they are easy to find and access can be requested without any problems. Transparency about an existing data offering, including up-to-date documentation on the technical content of the data products, e.g. by means of a data catalog, is recommended here. Data preparation steps and data quality indicators should also be provided transparently in order to increase trust in existing data products. This prevents new, very similar data products being built on the same source data for very similar questions instead of using existing data products. In this sense, support during the development of a data product can create awareness that a data product that is more generic in case of doubt can also answer questions for other users.
Data Mesh is a promising approach, but it has its pitfalls, especially in the non-technical area. The basis for the successful introduction and use of the approach is above all strategic decisions at company level and the implementation of the strategy over a period of years. This includes comprehensive paradigm shifts in the organization of the company, e.g. in terms of cross-functional teams.
In addition, it takes a lot of time to change the awareness of those involved at all levels. Training, motivational measures, information and advice can be used to promote awareness of the natural sharing of data, the value of data as an asset and for a long-term data product instead of a short-term project.
The clear responsibility of cross-functional teams for a data product is a key characteristic of the approach and the basis for trust and use of the data products. With guidelines and roles derived from the data strategy, this responsibility is channeled into uniform channels and comparable data products are created.
This blog post was created in collaboration with Nicky Grassmann.
