Schedule

This session discusses the data lakehouse, which is the new kid on the block in the world of data architectures. In a nutshell, the data lakehouse is a combination of a data warehouse and a data lake. In other words, this architecture is developed to support a typical data warehouse workload plus a data lake workload. It holds structured, semi-structured, and unstructured data. Technically, in a data lakehouse the data is stored in files that can be accessed by any type of tool and database server. The data is not kept hostage by a specific database server. SQL engines are also able to access that data efficiently for more traditional business intelligence workloads. And data scientists can create their descriptive and prescriptive models directly on the data.

It makes a lot of sense to combine these two worlds, because they are sharing the same data and they are sharing logic. But is this really possible? Is this all too good to be true? This session discusses various aspects of data warehouses and data lakes to determine if the data lakehouse is a marketing hype or whether this is really a valuable and realistic new data architecture.

• The importance of combining the BI use case and the data science use case in one architecture
• The relationship between the data lakehouse architecture and SQL-on-Hadoop engines
• Comparisons of the data warehouse, data lake, and data lakehouse are biased
• Missing components of the data lakehouse
• Storing data in open file formats has practical advantages
• Is the data lakehouse a business pull or technology push?

In this session we will touch on the most used models. How to apply them in context. Also the need to choose a model fitting the rhythm and purpose of your data.

A lot of discussions are going around about what would be the best data model. Guru’s fall over each other to proof their points. Depending on the background of IT professionals, they might not even know that there are several types of models available. Due to the focus on data science the models behind it are ignored or overlooked. What model you choose has implications for your applications down the line. So, you need to choose a model fitting the purpose of your data, through the life cycle of the data.

Why is it relevant now? GDPR, the new EU Data Act and AI Act ask organisations to know what data they have got, with what purpose, when they started collecting the data and how long are they going to keep it. And those are just the minimal demands. In a lot of industries data is recorded as an asset in the balance sheets. Data has moved from a supporting role to the main act.

A data model determines how data elements relate to each other. If data can be combined, how it can be retrieved. When designing a system most of the time it is overlooked that data is not just saved for registration purposes, but also for analysis.
With the rise of Data Science, Artificial Intelligence and Machine Learning data quality is not a minor concern anymore. There are no people interpreting the registered data. Data is interpreted on a large scale by (mathematical) models in computers. Computers are used for what they do best: calculation. If data is stored in the wrong way, the models will give wrong results, often with disastrous consequences.

The main types of models, relational, dimensional, ensemble and graph are explained.
The focus when choosing a model is: which concern of the organisation needs to be addressed. Is it important to register as accurately as possible what happened in interactions with a customer? Is it necessary to look back at how business was run to improve? Or should we look forward to the future, based on historical data?

Based on the objectives, we discuss the advantages and disadvantages of each type of model. There is no ‘one size fits all’ in data modelling. Choices made during development have long-term consequences for the possible applications of the data. These concerns have implications for the data architecture. In this session we focus on data models.

• Why data and data models are moving from a supporting role to a leading role
• Type of data model, relational, dimensional, ensemble and graph
• Relationship between data model and organizational objective
• Rhythm and purpose of the data captured in a model
• Capture history in a model.

Most firms today want to create a high quality, compliant data foundation to support multiple analytical workloads. A rapidly emerging approach to building this is to create DataOps pipelines that produce reusable data products. However, there needs to be somewhere where these data products can be made available so data to be shared. The solution is a data marketplace where ready-made, high quality data products that can be published for others to consume and use. This session looks at what a data marketplace is, how to build one and how you can use it to govern data sharing across the enterprise and beyond. It also looks at what is needed to operate a data marketplace and the trend to become a marketplace for both data and analytical products.

• The need for a high-quality data foundation to support decision making
• Incrementally building a data foundation using DataOps pipelines to product Data Products
• Using an enterprise data marketplace to share data
• What is the difference between a data catalog and a data marketplace?
• Challenges in establishing a data marketplace
• What processes are needed to operate a data marketplace?
• Governing the sharing of data using a data marketplace
• Trends – publishing analytical products in a marketplace
• Progressively shortening time to value using a marketplace.

Data management encompasses a broad spectrum of dimensions and focal areas which come to play when an organization needs to adapt to an environment that becomes more and more digital. But where to start? And how do you develop a successful strategy that will be adopted throughout the organization?

Evides is already 5 years underway in rolling out a data management program to grow towards a data conscious and mature data driven company. In retrospect there are a few lessons to draw which proved of key importance. And with that, the future perspective becomes more clear every day. From the perspective of people, process, technology and organization a few insights will be further elaborated:

• One single truth in data: one can speed up growth in data maturity, but one cannot skip steps towards growth
• Align data management on the appropriate organizational level and ensure proper sponsorship
• Continually show the value of data for stakeholders and grow active data management from the inside
• How to design and develop a sustainable BI-platform which delivers value from the start
• What is required for your workforce to make the data strategy work in terms of skills?
• How does one know when it’s the right time to invest in data governance tooling?

Since Google announced its Knowledge Graph solution in 2012 the paradigm has found its way into many real-world use cases. These are mostly in the analytics space. The graph database market has exploded over the last 10 years with at least 50 brand names today. International Standardization is coming – very soon SQL will be extended by functionality for property query queries. A full international standard for property graphs, called GQL, will surface in late 2023.
The inclusion of graph technology dramatically enlarges the scope of analytics by enabling semi-structured information, semantic sources such as ontologies and taxonomies, social networks as well as schema-less sources of data. At the same time graph databases are much better suited for doing complex multi-joins analyzing large networks of data, opening up for advanced fraud detection etc. The Panama papers is the best-known example. Finally graph theory is a mathematical discipline with a long history, which among other things have created graph algorithms for many complex analytics, such as clustering, shortest path, page rank, centrality and much more.
This presentation will cover what a Knowledge Graph is, how it is different and yet complementary to other technologies. Furthermore, Thomas will cover:

• Why do semantics and relations matter?
• What kinds of data architectures and pipelines?
• Which are the vendors and the products?
• Which standards exist?

It is a non-technical presentation, focusing on business requirements and architecture. More technical information will be covered in the workshop Understanding Graph Technologies on the 5th of April.

The concepts and practices of Data Mesh and Data Fabric are data management’s new hot topics. These contrasting yet complementary technology and organisational approaches promise better data management through the delivery of defined data products and the automation of real time data integration.

But to succeed both depend on getting their Data Quality foundations right. To work, Data Mesh requires high quality, well curated data sets and data products; Data Fabric also relies on high quality, standardised data and metadata which insulates data users from the complexities of multiple systems and platforms.  

This session will briefly recap the main concepts and practices of Data Mesh and Data Fabric and consider their implications for Data Quality Management. Will the Mesh and Fabric make Data Quality easier or harder to get right? As a foundational data discipline how should Data Quality principles and practices evolve and adapt to meet the needs of these new trends? What new approaches and practices may be needed? What are the implications for Data Quality practitioners and other data management professionals working in other data disciplines such as Data Governance, Business Intelligence and Data Warehousing?   

This session will include:

• A brief overview of the main concepts of Data Mesh and Data Fabric
• A review of the current state of Data Quality Management – its successes and failures
• An analysis of the impact of Data Mesh and Data Fabric on Data Quality Management – will they improve or worsen the Data Quality status quo?
• Practical guidance on how Data Quality Management needs to evolve to support these new data management approaches
• A suggested roadmap of actions which Data Quality practitioners and other data management professionals should implement to ensure they remain relevant in the new world of Data Mesh and Data Fabric.

[On April 5th, Nigel will run a full day workshop on Data Strategy, check the conference schedule for details.]

We come from a world of algorithms and the focus of AI is still largely on optimizing models. In this contribution by Jan Veldsink, we are going to focus on data centricity, putting the data in the centre and turning it into multiple analyses/reports and applications.

Datacentric AI is a form of artificial intelligence that focuses on working with and using data to solve problems. This type of AI typically involves using machine learning algorithms and other techniques to analyze large amounts of data and extract actionable insights from it.

Some key points of datacentric AI are:

1. Datacentric AI focuses on working with and using data to solve problems.
2. This type of AI usually includes the use of machine learning algorithms and other techniques.
3. Datacentric AI can be used to analyze large amounts of data and extract actionable insights from it.
4. This type of AI is often used in a wide range of applications, such as image and speech recognition, natural language processing and predictive analytics.
5. Datacentric AI is an important tool for businesses, organizations and individuals who need to understand large amounts of data to make better decisions and improve their operations.

This session looks at the emergence of Data Observability and looks at what it is about, what Data Observability can observe, vendors in the market and examples of what vendors are capturing about data. The presentation will also look at Data Observability requirements, the strengths and weaknesses of current offerings, where the gaps are and tool complexity (overlaps, inability to share metadata) from a customer perspective. It will explore the link between Data Observability, data catalogs, data intelligence and the move towards augmented data governance and discuss how Data Observability and data intelligence can be used in a real-time automated Data Governance Action Framework to govern data across multiple tools and data stores in next generation Data governance.

• What’s happening with data in business today and why there may be problems ahead
• Requirements to avoid problems and strengthen data governance
• What is data observability and what is it trying to help you do?
• What is it that can be observed?
• The data observability process – what are the steps and how does it work?
• Vendors in the market and what they are capturing
• The link between data observability and data catalogs
• Data observability, prescriptive analytics, and real-time automated data governance.

After a brief introduction of Vektis, Herman Bennema will use practical examples to show what kind of information can be extracted from the claims data (with a total value of over 850 billion euros) managed by Vektis. He will also discuss the legal bounderies within which Vektis operates and outline the dilemma we face in the Netherlands: how do we ensure a sound balance between, on the one hand, the privacy risk when using healthcare data and, on the other, the potential to keep healthcare affordable, accessible and of high quality based on data analysis.

• Introduction Vektis
• Data architecture
• Examples of information based on claims data
• Dilemma: what is possible versus what is allowed
• Practical approach for navigating between GDPR and Data exploration.

Engaging Stakeholders and Other Mere Mortals

Interest in Data Modelling, especially Concept Modelling (Conceptual Data Modelling) has increased dramatically in recent years. That’s great news, but our modelling can still be improved. When it’s done well, Concept Modelling is a powerful enabler of communication among different stakeholders including senior leaders, subject matter experts, business analysts, solution architects, and others. Unfortunately, the communication often gets lost – in the clouds, in the weeds, or somewhere off to the side. Sometimes the modeller has drifted too quickly into abstraction, sometimes the modeller has taken the famous “deep dive for detail,” but the outcome is the same – confusion, frustration, and detachment. The result – inaccurate, incomplete, or unappreciated models.
It doesn’t have to be this way! Drawing on over 40 years of successful modelling, this session describes core techniques, backed by practical examples, for helping people appreciate, use, and possibly even want to build data models.
Topics include:

• Unclear on the concept – how to think about concept modelling
• “Role induction” for clients – skip the “tutorial” on data modeling and Just Do It!
• Get a sense of direction – guidelines for data model graphics
• “Scripts” for extending the model – the value of consistency
• “Plays well with others” – make data modelling vital for analysis and design.

[On April 5th, Alec will run a half day workshop on Concept Modelling, check the conference schedule for details.]