Time for a Mesh Start? How Vistaprint Reimagined Its Data Architecture as a Data Mesh

Sebastian Klapdor, EVP and Chief Data Officer at Vistaprint, explained how Vistaprint scaled the benefits of its data and analytics capabilities using data mesh architecture at Corinium’s CDAO Fall conference in Boston

In his years advising clients about data and analytics for management consulting firm McKinsey and Company, Sebastian Klapdor noticed a recurring challenge for enterprises. They struggled to scale the benefits of their data and analytics initiatives.

This is a profound problem. After all, you can’t become a data-driven organization if data is siloed, and insights are bottlenecked beyond the reach of decision-makers.

Speaking at Corinium’s CDAO Fall Boston summit last week, Sebastian Klapdor, Executive Vice President and Chief Data Officer for e-commerce marketing company Vistaprint, argued that switching to a data mesh architecture is essential to break through these barriers.  

When Klapdor joined Vistaprint in November 2019, his vision was to transform the $1.5 billion USD company into one of the world’s most iconic data and analytics-driven organizations. And while some have called data ‘the new oil’, Klapdor believes this is a poor analogy. After all, oil is difficult to extract, expensive to store and highly toxic. In Klapdor’s vision, data is more like water. Easy to store, available to everyone, and free-flowing.

“That again comes back to democratizing data ,” Klapdor said. “I want every decision-maker at Vistaprint to have the right information at the right point in time to make the right decision.”

Making Data and Analytics Available at Scale

When Klapdor joined Vistaprint, he observed many of the same architectural issues that had limited the impact of data and analytics for his former clients.  

“The technology and the organization around it with respect to data at Vistaprint had clearly reached a ceiling; it did not scale anymore,” Klapdor recalled. “Why was that? It was because there was a central monolithic team working on monolithic hardware in a data center in a basement. So basically, velocity was zero.”

This centralized team was building tables on top of one another in an ever-growing pile. And because the team lacked domain-specific expertise, it was tough for them to answer questions from analysts about the context of any of the data. They had, in effect, created the mother of all data siloes.

What’s more, the whole operational model and mindset around data was highly centralized and unidirectional – the antithesis of a self-service model. You ask a question in an email, and we’ll send you a PowerPoint presentation.

Breaking down many of these structures would be essential if Klapdor was to realize his data-driven vision for Vistaprint.

The Objectives of a Data Mesh Architecture

Klapdor identified five core changes that he would need to make to ensure the benefits of data and analytics could be realized at scale and at speed across Vistaprint’s global footprint:

  1. Build cross-functional teams consisting of data engineers and data scientists focused on end-to-end data product delivery
  2. Put these teams close to the respective business problems by dividing the company and its data into ‘domains’, like ‘marketing’, ‘products’ etc
  3. Divide the centralized data center into a distributed cloud-based stack
  4. Embrace ‘data as a product’ thinking. Essential to this is linking data to the business value it creates
  5. Move to a self-service operational approach – this isn’t about ‘we do data analysis for you’

By making these five changes, Klapdor was able to put problem-solvers close to problems and data at the fingertips of decision-makers. The results for Vistaprint were transformative.

“[The teams] can move fast without being dependent on other teams ,” Klapdor said. “And the second big thing is that they are domain-oriented. They have all the context for their data and their business problem. So, they can build much better data products than a team that doesn’t have that business context.”

Getting Started with Data Mesh

Implementing a data mesh architecture requires a paradigm shift in thinking about data and analytics. Of course, changing the way people think is a real challenge.  

“Data mesh is actually an architectural paradigm on how to organize data in your company ,” Klapdor said. “It is the distribution of data away from a central, monolithic team and monolithic units into data domains.”

He continues: “Then, cross-functional data product teams build products for our stakeholders, who then use these products in a self-service way.”

For data leaders looking to get started on the road to data mesh, Klapdoor has five key pieces of advice:

  1. Create a domain map. Stakeholders need know to which domain to go to if they want a certain data point. The map will also help you to establish single sources of truth. It’s important to make specific domains responsible for specific data points
  2. You need a strong data platform. The more features you have, the easier it will be for your data product teams to innovate
  3. Develop a data catalog. To make data discoverable in the decentralized data mesh it needs to be properly cataloged
  4. Promote ‘data product’ thinking. Being part of a team that is responsible for creating data products will be a big shift for engineering teams and analysts
  5. Develop processes for effective communication. This is especially important to avoid the duplication of efforts in a decentralized environment

Going forwards, Klapdor expects his cross-functional teams to produce more data products that will have a measurable impact on Vistaprint’s performance.

“I hope that, in a year from now, we’ll have many more of these data products, at the scalable stage, but also many more on that overall landscape, driving more value for Vistaprint and our customers,” Klapdor concludes.

Vistaprint isn’t the only company exploring data mesh concepts. To discover what the executives at CDAO Fall Virtual 2021 had to say about data mesh, click here.