Benefits of data mesh might not be worth the cost
Data mesh can improve an organization's data quality and insights, but significant challenges can make these benefits difficult to achieve.
Data operations and capabilities have become more agile, which allows organizations to empower employees with data at all levels. It also gives rise to new decentralized data approaches, such as data mesh.
Organizations continue to grow their data capabilities, spending more year over year on technologies and initiatives such as data mesh. Some 93.9% of surveyed organizations plan to increase data investments in 2023, according to the "Data and Analytics Leadership Annual Executive Survey 2023" from NewVantage Partners, a Wavestone company. That's up from the 87.8% who reported increases in data investments in 2022.
Data mesh and fabric is a category with a higher spending forecast in 2023. Some 41.5% of data leaders identified data mesh and fabric as a primary area of investment for 2023 in the NewVantage Partners report.
Organizations can save costs, increase returns on data investment and improve efficiency with data mesh. However, low data maturity or lack of consistency in data governance and management can hinder this approach.
Benefits: Collaboration, flexibility, quality
Data mesh offers diverse benefits such as improved collaboration and data quality. Top benefits include the following:
Improved data sharing. Data mesh focuses on a decentralized approach to data and can help improve sharing with internal and external partners, said Robert Thanaraj, research director at Gartner.
With data mesh, each data owner manages their data and establishes governance based on the requirements for the specific data it holds. The owner can share that data with others while controlling the management and governance levels.
Better alignment of data solutions to business needs. Data mesh puts data-related decisions in the hands of the business users knowledgeable about that data, said Phanii Pydimarri, former head of commercial products, AI and advanced analytics for industrial and household tool maker Stanley Black & Decker.
"Data mesh lets the subject matter expertise stay with the data, putting it in their hands to come up with solutions," Pydimarri said.
Chief data officers and their teams work with the business users in these scenarios. This means data solutions and the digital products that use data are all more likely to align with the actual business needs.
Faster access to critical data, faster path to insights. Data mesh enables self-service access. Each business team can quickly access critical data.
"They can be much more agile," Pydimarri said.
With data mesh, data leaders and data teams still advise and guide, but no longer serve as intermediaries or gatekeepers to data -- roles that can slow down the whole process.
Improved data quality. Data mesh assigns responsibility for data management, governance and quality to the domain users.
"You have a clear responsibility owner," said Dan Sutherland, senior director of technology consulting at Protiviti.
The domain users, as owners, can then enforce the requirements for data when they share the data with others.
Increased visibility into the data. Similarly, data mesh creates better visibility, Sutherland said.
"Data mesh forces you to define and classify data into domains, and that gives you a clear view of the data from when it's created to when it's consumed," he said.
Cost-savings potential. The decentralized approach keeps the data with the domain owner, rather than moving it and replicating it through multiple systems and applications, as is often the case in a centralized structure. Data infrastructure where the data isn't touched multiple times for multiple reasons can save companies money.
"Data mesh is a complex framework, but once it's defined, you have cost savings," Sutherland said.
Data monetization from the beginning. Data mesh sets organizations up to estimate the value of data products, said Michele Goetz, vice president and principal analyst at Forrester Research.
"You can monetize your data from the beginning -- you know the value you're going to generate when you create a data product," Goetz said.
Data is user-driven and designed for problems at hand, she said. Organizations can calculate the cost of data, outcomes and ROI.
"Everybody can create a data product to solve their problems or support their objectives," Goetz said. "Eventually, you create a flywheel: If everyone can create a data product, then they can share them, and you start to see data not just impacting your operational metrics, but it starts to be seen at the bottom line of your organization, and you truly do become insight-driven."
Improved likelihood of data delivering business value. Organizations that modernize data architecture and processes within a data mesh and achieve such benefits have a higher likelihood of deriving business value from their data.
"It's higher data quality, faster business response, and it's a better understanding of the data landscape so [the organization] can look for new opportunities," Sutherland said.
Don't overlook data mesh challenges
Organizations typically face challenges when they adopt data mesh, with the following issues:
Consistency in data governance and management. Data mesh is a decentralized data management architecture, so organizations often face challenges ensuring all owners adhere to centrally required data governance and management standards.
"The first challenge that comes up is, how can I create consistency when everybody can define the data however they want?" Goetz said.
She listed the following common questions that can stymie efforts:
- When you are decentralized and distributed, how do you govern?
- Where do you govern?
- What does governance look like?
- How do you control and protect things?
- How do you ensure privacy and regulatory compliance?
The lack of required maturity. Only 18% of surveyed organizations said their data and analytics governance was mature and scaling across the enterprise, according to Gartner's 2021 "Data and Analytics Governance Survey."
Peter AikenAssociate professor, Virginia Commonwealth University School of Business
That lack of maturity presents a significant obstacle for organizations looking to implement data mesh. Domain-level subject matter experts in an immature organization won't be ready to create and support the data governance and management they're required to handle.
"An organization without a high level of maturity, or with bad data, can't even have a conversation about data mesh," said data management expert Peter Aiken, associate professor at Virginia Commonwealth University's School of Business and past president of the data association DAMA International.
A lack of a product-centric culture. Similarly, organizations without a product mindset or the skills required to maintain and mature that product-centric culture will struggle with a data mesh approach, Pydimarri said. Data mesh is meant to spur and support data-fueled products and solutions.
Getting it right. Organizations have their choice of tools, architectures, frameworks and methodologies they can apply to their data enterprise.
This multitude leads to overlapping approaches, confusion and implementations that miss the mark. In other words, some might try to adopt data mesh, but fall short in implementation and, thus, anticipated benefits.
Goetz said to consider whether data and data governance teams are interpreting data mesh into what they know -- or if they are ready to see data from the point of view of business value and use cases.
"If the former, it isn't data mesh, and they will not make progress to making data work and all the data challenges that exist today," she said. "If the latter, they embark on a path that creates organic and iterative data outcomes that eventually are visible to the bottom line of business."
An uncertain future for data mesh
The benefits and challenges of data mesh leave its long-term future uncertain.
Data mesh might be a good framework for businesses that acquire companies but don't consolidate with them, thus wanting a decentralized approach to most or even all of the individual companies' data, Thanaraj said.
It might also be a good option for large organizations that operate in multiple countries. These organizations' leaders might want to -- and are sometimes required to -- maintain local data autonomy.
"That's where I see data mesh being a much more appropriate data architecture to apply," Thanaraj said.
Still, questions remain about the long-term value of data mesh. In fact, Gartner labeled data mesh as "obsolete before plateau" in its 2022 "Hype Cycle for Data Management."
Moreover, organizations could more readily use other better-defined and more easily implemented approaches to improve their data programs, Aiken said. Organizations have DataOps, existing data management frameworks and data governance practices at their disposal. If a data program doesn't follow best data management practices, data mesh won't improve it.
"Those improvements could be achieved by other practices that don't have a buzz around them like data mesh," he said.
While other experts agree with Aiken, they haven't ruled out data mesh as a potentially valuable approach. But some feel it might be a long time before most organizations overcome the challenges and reap its benefits.