CDO Exchange - Lesson learned implementing a large data mesh at Payback.pdf

© 2024 Thoughtworks | Payback
Lesson learned
implementing a large
Data Mesh at PAYBACK
1

© 2024 Thoughtworks | Payback 2
A global community of technology experts and thought leaders.
10,500+
Employees
19
Countries
1993
Founded, HQ Chicago
$1.13B
2023 FY Revenue
41.8%
Women or Gender
Diverse People
2021
NASDAQ Listed TWKS
500+
Clients
3.95/5.0
Q1 overall
Glassdoor rating
North America
Atlanta
Chicago
Dallas
Denver
New York
San Francisco
Toronto
Latin America
Belo Horizonte
Porto Alegre
Quito
Recife
Santiago
São Paulo
Europe
Amsterdam
Barcelona
Berlin
Bologna
Bucharest
Switzerland
Cluj
Cologne
Hamburg
Helsinki
Iasi
London
Madrid
Manchester
Milan
Munich
Stuttgart
China
Beijing
Chengdu
Hong Kong
Shanghai
Shenzhen
Wuhan
Xi’an
Australia
Brisbane
Melbourne
Sydney
Singapore
Thailand
Bangkok
Vietnam
India
Bengaluru
Mumbai
Gurugram
Pune
Coimbatore
Hyderabad
Chennai
2

Rooted in a culture of learning and
sharing, we believe that knowledge
should be accessible to all. We are
committed to improving the tech
industry and are passionate about
sharing our expertise across
technology, business and culture.
Books written and digital publications
3
100+
books written
Perspectives
A publication for digital leaders
Learn more
Technology Radar
An opinionated guide to
today’s technology landscape
Learn more
Digital Fluency Model
Discover your digital fluency
Learn more
Decoder
The A-Z guide to tech for
business executives
Learn more
Looking Glass
The trends your business should
focus on today and in the future
Learn more

4

Why Data Mesh?
5
MESH STREET
Photo by Matt Walsh on Unsplash

© 2024 Thoughtworks | Payback 6
5 years of Data Mesh!
Data Mesh was developed
at Thoughtworks in 2019
by Zhamak Dehghani.
Data Mesh provides a set
of principles and a framework
for building a modern scalable
organisational data capability
and data ecosystem.

Significantly reduce bottlenecks
associated with centralized data
engineering teams and thus improve
implementation lead time.
Why Data Mesh
at Payback?
As the volume of transactional data we
handled and the complexity of our
product ecosystem increased, so too
did our lead time for any changes in
our data landscape.
It very quickly became clear that
something needed to change, and that
we needed to create a new foundation
for our data — one that could grow with
us.
7
Accelerate analytics delivery and the
realization of leading data use cases
such as AI and machine learning.
Increase enterprise agility and
accelerate decision-making for teams
across the business.
Empower stakeholders and product
teams to operationalize data in new
ways

What is like
to build it?
8
MESH STREET

9
Green field?

10
… more like building a big city’s new underground

…without blocking the entire company in the meanwhile
Creating a roadmap to get there…
It was clear from the start that
this process wouldn’t be easy:
migrating a complex, long-running data
estate to the cloud, requires a
combination of technical and
organizational change, to ensure the
two are effectively implemented in
tandem.
11
● The process began with the
identification of vertical business
domains, plus two horizontal
technical domains.
● Each domain was appointed a
team with a BO, a PO and a TO to
make sure that all the needs were
addressed at every stage.

12
you start with the preparatory work…

… for “eating our own dog food”
Building our own tools…
First we’ve setup 3 core teams inside
the horizontal domains to prepare
tooling and services for vertical domain
teams (and avoid to reinvent the wheel
every time).
Alongside the core teams, a data
governance council was established
to provide guidelines, oversee data
access requests between consumers
and producers, and monitors KPIs
around data quality adoption.
13
● Cloud infrastructure team:
infrastructure, environments
provisioning, security and auditing.
● Cloud Data & AI Platform Team:
foundational data & AI products
tooling and services.
● Cloud Migration Team: migration
of on-prem data assets, support
for domains teams building data
products

14
but it’s when you start digging that problems arise…

… to move forward without being dogmatic
Embracing differences and iterations…
Data duplication is a feature of the
Data Mesh — not a disadvantage:
it enables faster implementation.
To avoid ending up with a data “mash”,
strong data lineage practices are
required to ensure that data flow
information is always up-to-date.
Whenever an efficiency gain or
optimization is identified, an
intermediate data product should be
evaluated.
15
● DWH teams tend to create a
smaller number of large data
products.
● Data lake teams tend to create a
great number of data products but
with different granularities.
● Building new data products from
scratch without proper
governance, may lead to
duplicated effort across domains.

16
… and finally you have to embrace the users’ feedback

…to make them champion the new paradigm
Make users’ life easier…
Shifting data ownership towards the
business is the most challenging
aspect of Data Mesh adoption.
We first assessed the domains data
culture against the breadth and
complexity of their data assets.
We ranked them upon 4 levels (i.e.
fully centralized, contribution model,
expert support, fully decentralized) and
setup tooling and support accordingly.
17
● Communicate the benefits to
every stakeholder in terms that will
resonate with them.
● Ensure your self-service data
platform is user centered and that
domains have the right skills.
● Invest in enablement and provide
simultaneous training and
upskilling.

Lesson learned…
18
MESH STREET

Start with a thorough assessment about
how the organization operates, how the
people within it work.
Lesson learned
at PAYBACK
Data Mesh can help digital product
organizations to cut lead times,
accelerate delivery, break down
process bottlenecks, and empower
users across domains to innovate
autonomously.
But, to do that, it’s essential that the
Data Mesh is closely tailored to users
needs, existing IT estate, and culture.
19
Plan from the very beginning the
change management alongside the tech
transformation, to best support and
empower people.
Domains should be supported with
education at every stage of the journey,
and be enabled to adopt Data Mesh at
their own pace.
The biggest key to success is finding the
right balance between standards and
freedom to innovate.

Would you like
to know more?
20
MESH STREET

Data Mesh:
From Textbook to
Transformation
21
Connect with us to download the white paper
21
Norbert Wirth - PAYBACK
Alessandro Confetti - ThoughtWorks

Alessandro Confetti
Data & AI Lead - HCLS & CMT Markets
aconfet@thoughtworks.com
22
Norbert Wirth
Global VP Data
norbert.wirth@payback.net
$ tail -f questions

CDO Exchange - Lesson learned implementing a large data mesh at Payback.pdf

More Related Content

Similar to CDO Exchange - Lesson learned implementing a large data mesh at Payback.pdf (20)

More from Alessandro Confetti (13)

Recently uploaded (20)

CDO Exchange - Lesson learned implementing a large data mesh at Payback.pdf