Ep. 41: Ivo Sokolov - Data Engineering

FULL EPISODE TRANSCRIPT
Adam: (00:00)


Welcome back for episode 41 of Count Me In, IMA's podcast about
all things affecting the accounting and finance world. Our
featured guests for this episode is Ivo Sokolov,, a data office
division co-leader at one of the most profitable and best
capitalized banks in Austria where he specializes in data
engineering, regulatory tech and project governance Ivo joined
Mitch to talk about how data engineering fits into today's
accounting and finance function and it's emphasize the importance
of proper controls over any data project. So what exactly is data
engineering and why should accountants care? Let's jump into the
conversation and have Ivo explain. 


 


Ivo: (00:48)


Engineering is a discipline that is, consists of parts of data
center, acknowledged meaning databases. Then as an architecture,
data structures extract, transform load processes and tooling on
the one hand programming skills. On the other hand, self care
architecture and and a bit of all new generation technical
infrastructure knowledge. For example, cloud infrastructure,
containerization, dev-ops and all of that combined, is sort of
data-engineering. 


 


Mitch: (01:25)


And how exactly does data engineering fit into accounting and
finance?


 


Ivo: (01:32)


The purpose of data engineering is to enable the organization to
prepare data pipelines for, whatever analytical usage is may be
required in the difference business functions. And finance is one
of those functions providing the proper, the underlying data
architecture that enables the finance function to quickly make
sense of the data within the organization too quickly build and
maintain data, pipelines for analysis and, and ETL processes and
data science, including more, let's say more standard tasks like
data cleaning and amortization, setting up batches, feeding them
data into proper business intelligence tools, and preparing
dashboards. 


 


Mitch: (02:22)


So now with all of this, how does data flow through an
organization and for the finance function particularly impact
something like forecasting or analyzing financials? 


 


Ivo: (02:33)


Using newer technologies in newer platforms, for example, Python
and R, are tools that are usually employed nowadays, one can do
predictions and forecasts on financial figures. For example,
income statement, balance sheets, cashflow statements. And that
can be done within the finance function, within the business
division without requiring a specialized technique of the teams.
Given that, you know, data engineering has prepared the data
properly. The proper data pipelines ensures that finance has,
depending on the news, the needs of the company, access to near
time data or micro badge data, meaning finance does not work with
data from the previous month or from the previous quarter. Having
these, the underlying the data flows properly in the organization
enables the forecasting and that is say for the financials to be
much more timely. 


 


Mitch: (03:30)


So I know you mentioned a couple of the tools, but which tools
and particular skills really would be most useful for finance
professionals to kind of borrow from data engineering and assist
with these analyses? 


 


Ivo: (03:43)


The tools are essentially around using modern of scripting
languages such as Python or R. That also includes a lot of
libraries with functions that are useful within finance for
forecast for analysis. So a popular tool would be a set of source
would be like the Jupiter, the Jupiter hub environment, Jupiter
notebooks, finance professionals can simply log onto a browser
from anything client and build their analysis in a, in a way
quite similar to what, with Jupiter and Python or R. Similar to
the manner in which you're a software developer, would you use
the same tools to write software for other purposes. Other skills
that are useful to borrow from data engineering are having
whatever code one writes to do their analysis or their forecast
or their models be put into, into version control. That way it
can be, and used within the department that way people were
structuring or going about solving a finance desk, the way that
in software development, one would stop libraries with certain
functions. For example, getting the proper customer segments or
getting certain field there is that are used throughout. You
know, one doesn't have to ride the same code three times and this
is definitely, we see more and more of that in business
departments and in finance 


 


Mitch: (05:23)


As these tools and skills are starting to become shared across an
organization, essentially. How is the data that is the end result
ultimately viewed differently across these functions? For
example, finance versus it or even something like
marketing. 


 


Ivo: (05:40)


Now the problem with a siloed data and every function having
their own data warehouse or data to do their analysis. Looking at
pretty much I'd say customer data in a finance looks at the
customer and account from a different perspective that marketing
was, but let's say 50%, 60% of the underlying selection of the
data would be the same. And now if you've moved into a proper
data architecture, you'd expect certain basic fields or basic
definitions to be shared, to be put into version control. And
that is different than the case there was before using the BI
tools would imply publishing some of these dashboards on a server
so that they can be shared throughout the organization so that
they're not my Excel file sitting on a drive in my division. But
also could be shared with marketing. It could be shared with IT
if they have to add something or do something with it. So this
goes into having a central aligned data architecture. 


 


Mitch: (06:49)


So with this data architecture and the version control that you
referenced, I know data is pretty free flowing across
organizations now, so who's responsibility is it then to make
sure there's proper governance in place and governance projects
that are set up with internal controls to monitor how this data
is used and seen? 


 


Ivo: (07:13)


This really depends on the type of organization for certain
organizations such as banks, there's a regulatory mandate, to do
proper data governance and data aggregation capabilities across
risk and finance. And data governance would imply that every
individual owner of data within the organization is defined and
then they would know when that data structure changes and
continue to maintain it such that overall if you have a figure on
your balance sheet and if you want to understand how that figure
comes about, there's a very clear data lineage and that you know
how to which steps and which data transformation steps or they
engineering steps took place in order for the figures to be as
they are, who is responsible for implementing that would differ.
But we definitely see a lot of master data management or data
governance initiatives and sometimes, depending on the state of
the legacy systems of how new or how old the underlying data
architectures their organization might need to rethink and
initiate a strategic project in order to create the necessary
there. The architecture for combining data from usually a really
tens or or even hundreds of systems, operational systems. Where
that piece of data source data originally resides. And usually
...