Dr. Arif Wider discusses the concept of data mesh at the Data Innovation Summit 2022, emphasizing that it focuses on people rather than technology. He highlights the limitations of centralized data ownership and suggests that data products should be organized within domains led by cross-functional teams. The approach fosters a mindset shift in organizations, enabling more scalable and interconnected data products over time.

1. Data Mesh
It’s not about technology, it’s about people
Dr. Arif Wider
Data Innovation Summit 2022, May 5-6, Stockholm, Sweden
Dr. Arif Wide

2. Who am I
Dr. Arif Wider
● Former Head of Data & AI
at Thoughtworks Germany
● Many years hands-on consulting,
often as a lead engineer in
Thoughtworks’ client’s data teams
● Now tenured professor of
software engineering at HTW Berlin

3. SCALE
~ 2005 ~ 2007 ~2010+
Volume Velocity
Variety
What problem are we solving?

4. SCALE
~ 2005 ~ 2007 ~2010+ NOW
Volume Velocity
Variety Getting Value in
face of complexity
What problem are we solving?

5. Typical situation with centralized data teams
5

6. Somehow, we didn’t dare to touch the centralized
paradigm of a “single source of truth”
6
Warehouse Data lake Cloud

7. Centralized data ownership does not scale well
7
Big data | AI
platform
Ubiquitous Data Innovation Agenda

8. The source of the problem
8
central
data platform

9. The source of the problem
9
checkout
service
checkout
events

10. The source of the problem
10
checkout
service
checkout
events
f
r
i
c
t
i
o
n

11. Data Mesh is a synthesis of existing practices
11
Domain-
Driven Design
Product
Thinking
Platform
Thinking
Data
Mesh

12. Data as a product
Data Product
Manager
Domain Data
Product
What is my market?
What are the desires
of my customers?
What “price” is justified?
How to do marketing?
What’s the USP?
Are my customers happy?

13. Applying domain-driven design…
13
Data products belong inside
domains. A domain will
usually contain many data
products that can be used
both within and outside its
domain.
Types of data products:
source-aligned, consumer-
aligned, or aggregate data
products.
Data products don’t operate
in isolation.
Data Products

14. …to structure your organisation, i.e. your people
14
Data products belong inside
domains. A domain will
usually contain many data
products that can be used
both within and outside its
domain.
Types of data products:
source-aligned, consumer-
aligned, or aggregate data
products.
Data products don’t operate
in isolation
Data Products
Domain Teams
Teams own one or more
data products, depending
on the complexity. They
usually sit within a domain.
Teams have long-term
ownership for data
products.

15. From
● THE data team
● Data pipelines
● Producing data
● Data engineering team
● Data lake / warehouse
It’s a mindset and a language shift,
not an architecture (well, mostly)
To
● WHICH data product team
● Data domains with data experts
● Owning and serving data
● Cross-functional data product teams
● Data product experience platform

16. Spreading a mindset
Incubate
16
To begin, incubate
a small slice of the
overall operating
model to establish
some of the
foundational
capabilities and
learn from them.

17. Spreading a mindset
Sustain
17
As the foundations
become
sustainable further
data products can
be introduced.
With each new
products coming
on line the teams
can also begin to
grow.

18. Spreading a mindset
Sustain
18
As teams learn and
the platform begins
to mature, the
development of
further data
products begins to
accelerate

19. Spreading a mindset
Scale
19
As data products
scale and mature
the ecosystem
establishes with
increased
interconnectivity
creating new
possibilities for
deeper insight and
collective
knowledge across
the organisation

20. Thank you!
@arifwider
wider@htw-berlin.de
Dr. Arif Wide