Northern Trust – Breaking the Mold to Become a Digital, Data-Driven Leader  (Cloud Next ’19)
Articles,  Blog

Northern Trust – Breaking the Mold to Become a Digital, Data-Driven Leader (Cloud Next ’19)


[MUSIC PLAYING] BRAD FOSTER: My
name’s Brad Foster. I’m with Maven Wave. I’m here with Mary Beth
LoVerdi from Northern Trust. I’m going to be spending a few
minutes talking to you guys about how Northern Trust
is breaking the mold, as a 100-plus-year-old
financial services company, to really become more
of a data-driven leader in the financial services space. So just a quick backdrop
of myself, and then I’ll let Mary Beth introduce herself. I’m a partner at
Maven Wave Partners. I’ve been IT consulting
for 20-plus years and working with
Google Cloud Platform since really it was invented. And it was App Engine beta
back seven or eight years ago. So I’ve had a fun journey seeing
what Google Cloud Platform has become. I feel like five years ago,
this was the entire audience for a GCP or a Google
Cloud Conference. And now it’s 35,000 people. So it’s pretty cool to
see that, and pretty cool to partner with companies
like Northern Trust to build some of these
types of solutions. So Mary Beth, you want
to introduce yourself? MARY BETH LOVERDI: Yeah. As Brad said, I’m
Mary Beth LoVerdi. I’ve been in digital delivery
for the past 20-plus years– served at, worked
at startups, user experience consultancy firm. And I’ve had over 10-plus
years at Northern Trust. BRAD FOSTER: All right. So agenda, what we’ll talk
to you guys about today, Mary Beth will give a
little bit of an overview of Northern Trust. Some of you may not be
familiar with what they do, so we want to make sure
you’re grounded in that. She’ll talk a little bit
about the current employee experience, their sales
teams, relationship managers, and what they experience today. And we’ll talk to
you about how we’re trying to help change the way
they work with their clients and use data as a
central part of that. We’ll talk to you
about our approach, which, obviously, Google Cloud
Platform was a key component, but there’s a lot
of other things that go into making
a project successful. And then finally,
we’ll talk a little bit about where we’re
heading, what’s going to be happening
in next 12-plus months. MARY BETH LOVERDI:
So Northern Trust, we are a financial services firm
based in Chicago with 20,000 employees around the globe. We have three sides of
the business– the wealth management side, which we serve
high-net-worth individuals and families; and the asset
management and asset servicing side, which we serve
institutions and organizations. We have over $1 trillion
in asset under management. Core to the Northern
Trust culture is its roots in the
client relationship. True to the founding family’s
DNA was, keep the client at the center of
everything we do. When we sit down
with our clients, we talk about their
goals, and then we devise complex
financial solutions to meet those expectations. And as we talk about where
Northern Trust is going and some of the complex
challenges that we’re facing, I want to ground ourself in
a personal story about what happened to a friend of mine. His mother received a
phone call from a lawyer. And the lawyer
said, your grandson asked us to give you a call. He was in an
incident last night. He’s been held in custody. He’s retained our
services, and he’s asked us to give you a call– Pete, he’s at [INAUDIBLE]– for the $10,000
as a retainer fee. Don’t worry about Pete. He’s in good hands. We’re going to solve this issue. Not a problem. So my friend’s mother
gets in the car. She goes to the bank, and
she withdraws $10,000, FedExes it over to the
law firm in Brooklyn. The next day, the law
firm calls her back. They reassure her that her
grandson is in good hands. They’re doing
everything they can, and they’re confident they
can resolve the situation. But unfortunately, the
incident involved a woman. She was pregnant. She lost the baby. So there are some
additional charges that have been placed
against her grandson. Don’t worry, but we do
need an additional $10,000, and we’ll resolve this issue. She goes to her bank. She takes out the $10,000,
FedExes it to the law firm in Brooklyn. The third day, the law
firm calls her back. And now she knows
something’s up. And she says, can
I call you back? She calls her daughter,
who’s two doors down. And she realizes she’s
been a victim of a fraud, and she’s devastated. She calls her son,
who’s my friend. They drive down. And that’s just an
example of the level of service and expertise
that financial services are required to provide. And as we connect these
disparate data sources and run analytics on top
of them, it’d be pretty straightforward
to understand this client never
withdraws $10,000 in cash, sure enough wouldn’t do
it two days in a row, and then provide the
high-level service to safeguard against these type of measures. Now, imagine that type
of situation for an asset management pension
fund portfolio manager that’s making significant
transactions on a daily basis with– sometimes, pension
funds have a billion dollars of investable assets. BRAD FOSTER: Yeah, I think a
very great story we can all relate to, and
we’ve all probably had friends and family members
or personally experienced people trying to
take advantage of us. And the stakes are huge
for Northern Trust, as Mary Beth mentioned. The money that they
manage is enormous, so it’s key to get that
right and be secure and be forward-thinking
for their clients. MARY BETH LOVERDI: So just
to step back and give you a little bit of an overview
of the transformation that’s happening at
Northern Trust, we’re well on our way on our journey. We’ve moved to the cloud. We’re in a multi-cloud
environment. We have broken down
the monolithic stack into microservices. We’re using test-driven
development, continuous integration,
continuous delivery. Data analytics is
the next frontier that we want to tackle. So now I want to ground
you all in the use case that we tackled together. On the asset management
side of the business, their typical client
would be a pension fund, something like the
California teachers’ union. So we’re thinking not only
about the pension fund manager that’s overseeing– could be up to a billion
dollars of investable assets. And one new business
opportunity might be somewhere between $80 to
$100 million for, let’s say, a fixed-income product. So we’re thinking
about optimizing the performance, the investment
performance for that pension fund manager. And we’re also thinking
about the retired teachers who are counting on the benefits
check month after month. So our relationship teams,
they work in groups, and they oversee a
portfolio of business. And the collaboration
amongst the team is critical. So to walk you a day in
the life of a relationship manager in the asset management
side of the business, they start their day
early, like most of us. But in a regulated
industry, they’re using secured tools,
making it a little bit more difficult to access
information and to collaborate. They’re not slacking
information to colleagues. So they start their
day reading the news, and then traverse across
their day, checking emails, pinging the fellow
relationship teams. And they read multiple
sources of information, from subscription-based
sites that they authenticate into,
and then read, looking for activities
that’s happening within their portfolio
of work, to Excel spreadsheets or 50-page
Word documents that have new business
opportunities in them. And then they spend the majority
of their day in meetings. 80% of their time, if
they’re spending it as they would wish to do,
is spent in client meetings. And then they’re
back in the office, catching up on the activity that
happened throughout the day. And then they head on their
way home at the end of the day. So today, when we worked
with our senior leadership in the asset management
side of the business, they’re saying, hey, we
have all these relationship managers that are spending
all this time combing through information,
identifying individual ways to find new opportunities,
collaborate across a team. It’s very cumbersome. And we need to give them more
time back into their day. How do we take this
cumbersome process and simplify it so that they
can get the right information at the right time
throughout their day? So as I was talking
about in the example, some of the common questions
they ask on a day-to-day basis is, is there a new
mandate search? A new mandate search is a
new business opportunity, like I was saying, for the
California teachers’ union. And let’s say they have a
new investment consultant, and they rate Northern
Trust portfolio services in high regard. They’ve outperformed
their current provider. So today, they have
to go through all these different ways to
get this information. What we built is the ability
to blend this together in a very insightful
and timely manner. BRAD FOSTER: And I think talk
a little bit about the data sources themselves. So we started small. And I think one of
the things I took away from this first project is– I’m sure you guys
are in many seminars, and there’s sexy
machine learning things. And this is
revolutionizing the world. This project is something that’s
very simple, but yet extremely impactful. And I think in all
of your organizations out there, that is 90%
of the work to be done– is just listen to somebody
in your organization whose job is inefficient or who
could benefit from better data to be making better decisions,
and just go tackle that, get into some of the
hand-to-hand combat of getting the data pulled
together for them, giving them some simple, useful
tools like Tableau and Data Studio. And you’ll find that they’re
very sharp, intelligent people. They just need a little bit
better tools to do their job, and it really makes
a huge impact. So as we go through this, we
started with four third-party data sources– Fundmap, eVestment,
S&P Global, FactSet. So you guys could go buy
those sources yourself and subscribe to the
same type of information. You can see it’s a large
amount information– 8,000 mandates, 20,000
products, tons of news articles. And as you look at a couple
of the first-party data sets, you start out with
Salesforce, which, obviously, that’s where the
client information is, and then Marketo and how you’re
interacting with clients. It’s obscure and time-consuming
when a relationship manager is essentially– think of
them pulling those four third-party things
into a spreadsheet. And then I’m saying, OK,
here are my 10 clients. Let me go pull out all
the news articles, mandate searches that match them
manually, and then look at it. So that’s time-consuming. So A, we said, we can
match that together very quickly and easily,
and just take that off the table for them. And then from the
obscure perspective is once we have
that matched, then they’re filtering and
doing pivot tables and looking for what they want. So, well, let’s surface
what they want to see right at their fingertips. Let’s start giving them alerts
about that information as opposed to them having
to go in and find what they need to find. So remove the obscurity,
make it less time-consuming. And as Mary Beth said, these
people have 25% of their day where they’re not in meetings. That 25% is huge. The more they can understand
about those clients, about what’s happening in the
market and key trigger points, is really going to make a
huge difference for them. And so that’s really
what we were focused on. And while, I think,
the data was important, the platform’s important, there
are other components, I think, that made this project
very successful. And you guys are all familiar
with these paradigms of, add data, and they will come. Go build a data warehouse. And IT goes and does
it, and then you turn it over to the users. And they’re like, this doesn’t
even have half the data I want. Half the data in here,
I don’t even care about. So we really said,
align with the business, and solve a problem
one step at a time. What’s your problem? Six data sources, I can go solve
that problem in a major way. Then let’s come back afterwards
and ask you the next question. What’s the next
business problem? Can I answer it with the
data we already have, or do I need to add more data? Moving away from on-prem, legacy
platforms to cloud platforms. So Google Cloud
Platform’s obviously a key part of the team getting
down to working with the data and working with
the business and not worrying about infrastructure
and standing up a server or Hadoop
cluster, or dealing with all of those things. Just was right to
getting into the work. Moving away from waterfall
delivery to agile delivery is huge. They have a great agile team. When you walk into Northern
Trust’s office now versus five years ago, five
years ago, it was kind of the oak walls and the
big conference room chairs, the leather chairs. And you walk in now. You think you’re here
in Silicon Valley. It’s glass conference rooms;
sticky notes everywhere; big, open floor spaces; people
standing, eight people together in a real-time scrum
stand-up fashion. So it’s a very cool environment. They really embraced
that, which has been a key in these
types of projects. And then moving away from
that central IT analytics team to a cross-functional team. So if you look at that
eight-person scrum team, there’s somebody from
the asset management. The VP that runs the
business is sitting there at the table with
the data engineer, with the business
analyst, who understands how to make this work,
with possibly sometimes the actual relationship manager,
who’s there, asking them, can you do this? Show me this and this dashboard. So again, this cross-functional
team of Northern Trust business people, Northern Trust
technology people, and then partners like Maven
Wave filling in the gaps where they need help. To talk about the
solution, as I mentioned, is really, how do
you, on Google Cloud, provide more descriptive
analytics in real time in one place? And that one place is
ultimately Salesforce, where that’s the
place where people are logging in every day. And we wanted to get those
analytics pushed to that. And I’ll talk a little bit about
that solution here in a second, but maybe Mary Beth can
give you a little bit of an example of what
a dashboard looks like for one of these
users that’s looking at this information today. MARY BETH LOVERDI:
And just to step back, when we looked at the solution,
as you can imagine working at a bank, we had to be very
thoughtful about the data that we used to create the joins
and to provide the insights. We wanted to put the least
amount of sensitive or highly sensitive data into the cloud. Part of that scrum team was
a member of the infosec. And each step along the
way, we would revisit it with that engineer, asking
them to audit our process. And he would say
to us, OK, imagine if this data is compromised, and
all of it appears on Wikileaks. How do you feel about it? So we put one small
amount of sensitive data to create the joins. We tokenized that. And that gave the
scrum team, as well as the business, confidence to
move forward into the cloud. So as Brad was saying,
here is an example of one of the insights. It shows the product sentiment. So before, our relationship
management team used to log into
Excel spreadsheets, go through Word
documents try to assess, what’s the most recent
quarter product sentiment? Do they have a favorable
position on the Northern Trust product? And what was it the last several
quarters, and how does that benchmark against
the competition? We did this in Data Studio. And as Brad did
mention, our goal is to get it into Salesforce. BRAD FOSTER: So just
looking at the architecture of this from a
technical perspective, again, fairly straightforward. And something you
guys will see in just general reference
architecture is we have the data sources on the left. We have Cloud Composer
that’s orchestrating, getting that data into– Cloud Storage is the data lake. And that orchestration
can happen– some of that’s
available via API. Some of it’s a SQL pull. Some of it’s flat
files are being dumped. So it helps orchestrate whatever
data source format is there into Cloud Storage. In this scenario,
we’re actually using an ELT. Those familiar
with data warehousing, there’s extract,
transform, load. And then there’s extract,
load, and transform. And why we’re using ELT is
that the joins of this data are fairly straightforward. And so we’re using
the power of BigQuery to put the data there, stage
it, query it, join it together. And the resultant data set
goes back into BigQuery as the landing
curated data mart. So two forms of access. There’s Data Studio can
connect directly to BigQuery, and that’s where you can
do ad hoc analysis built in those dashboards. Once they have
dashboards that they want to see on a
continual basis, actually, we have Salesforce developers
that are putting those directly in Salesforce for them to view. They can obviously always
go straight to Data Studio if they want as well. But we also have custom
alerts and other things that we want to be developing. So we’re building an abstraction
layer with Cloud Functions to make that data available so
that a Salesforce developer can develop against that. And as we grow the database,
change the data schema, we’re not going to break
the Salesforce application. They log in one day, and
the dashboard’s not there. The alert’s not working. So the key abstraction
layer, a very common architectural pattern that most
of you are probably aware with and how people are
building things. And I think, as Mary Beth said,
security was front and center. I mean, the very first
meetings that were had were not about that
architecture or anything else. It was all about the data. What do we need to
put in the cloud? How do we secure
it, and how do you keep building mature
security practices? And the very first
thing becomes, don’t put any sensitive
data in the cloud, or anywhere, if
you don’t need to. And so we really
looked at, what’s the minimal amount
of data required to link third-party
data to the clients? And just use the very minimal
amount that you need, A. B, just make sure
people are accessing Salesforce via your VPN. Make sure you have modern
IAM policies of who can access this information. So a lot of meetings with
infosec and continual meetings there, just to continue
to mature and always make sure security is first in
everything that’s being built. So very important thing there. We developed a whitepaper
with Google on 11 steps to take to make sure, when
you’re building these things, they’re secure and you’re
following some best practices. So feel free to access that. And maybe, Mary Beth,
you can give a little bit of what your infosec
person told you guys in terms of how
to think about what you’re doing with the data. I thought it was
pretty interesting. MARY BETH LOVERDI: Yeah. Well, he said to us, imagine
if all this data is put out on Wikileaks, how is that
going to make you feel? BRAD FOSTER: Yeah. And I think, as
heads of business, that’s how you have to
think about this stuff. And I think those of you
have sat in Google’s security discussions, I mean, they have
a very mature perspective. And one of the things
that I took away was they view
everything as a threat– their own employees,
for example. Assume you have an employee
that wants to steal data. How would you
protect against that? So everything’s a threat–
everyone, every device, internal devices, external
devices, internal networks, external networks. And when you start maturing
your mindset like that, you really can put some
amazing security in place for these types of solutions. So talk a little bit
about where we’re headed– I think, as we talked about,
it is about, through the day, giving people alerts, not
making it the train ride and then the one hour before
I start my meetings at 9:00 or 10:00 in the morning and
sifting through information. It’s more about I
see stuff that’s relevant throughout the day. It’s easily accessible. And I’m really consuming that
information as I need it, not just building
blocks of time. I don’t know if you
want to add anything. MARY BETH LOVERDI: Yeah. And in the larger context at
Northern Trust, as I said, is we’re moving to the cloud. Our next frontier
is data analytics. If we can connect data
across the siloed experience, we can better
service our clients, identify anomalies or
patterns in behavior, and find opportunities
for delight. BRAD FOSTER: So
ideas for where this goes– so you get a
simple foundation in place with this data. You can start to add
additional data sources. So as those requests come,
as the relationship managers and other salespeople, as
their wealth management business and those
users begin to ask for more descriptive,
predictive analytics, we’ll add the
appropriate data sources. We do think Google
Analytics 360 is going to be a key
component of that because of the understanding
of the behavior and interactions of clients
with Northern Trust. So as they’re
going on the portal and reading articles or
consuming information or doing a banking transaction,
how do you serve them better? How do you make
that experience rich and serve the content that
they want to see, and just make it a better client experience? We would expect, as we go,
possibly using Dataflow as we have more complex data
transformations that are needed and we can’t quite do the
ELT strategy of letting BigQuery do all the joins. We could add Dataflow
into the equation. And then I think you hear ML
and reinforcement learning. As we get more advanced, we
can start asking relationship managers, was this helpful? Just those questions. I give them an alert
about something that’s happened in the marketplace. Just get the simple,
was this useful or not? And if they say
yes, no, you start to learn, even internally,
what content they want to see, what content they
want to consume. Another thing we
could put in place, we talked about the enhanced
customer interaction. So Private Passport’s an example
of the wealth management portal that customers can log into and
do a lot of their interactions. So understanding that
journey and enhancing the portal for the customers. And then along the
machine learning lines, several use cases– and this
isn’t an exhaustive list, but just some examples of
things we could look at– is behavioral segmentation. So as opposed to
segmenting customers based on their net worth or the number
amount of assets and those, just, general kind of
demographical type data, how are they
interacting with you? This group of people likes
real estate investing, and this group likes
other types of investing. And maybe they have completely
different net worths, but you could
really segment them based on some of that behavior. Next best content
and touchpoint– so I think Next Best
Action is a very big topic. And the reason we’re
calling this next best content and touchpoint
is Next Best Action has really, I think, a
real sales connotation to it. So how am I closing the deal
and getting the sale done? And a lot of Northern
Trust touchpoints are not about a sale. There is no impending
sale to be had. It is this is a
longstanding relationship, multi-years with people,
multi-generations with people. And so the next best
touchpoint might be, remember to send a birthday
card to them because they’re an important client. Or Brexit has
happened, and we have people that are in a
position where they probably want some education
on what that means. And let’s just be
proactive about helping educate them and trust
that, when those times come to work with Northern Trust,
that that relationship is there. So it’s that content,
that touchpoint which doesn’t always mean you’re
trying to sell them a product. And the risk and
propensity scoring, those are, what’s the
risk for their clients? What’s the risk on
investment portfolios? What’s risk for churn if
somebody’s sentiment unhappy, a product’s not performing well. How would they be more
proactive on engaging with their customers
on those things? And then some of those
additional data sources. We fully expect that
the unstructured data is going to continue to grow. And how do you comb
through tons of news and other things that are
happening in the market, apply some classification
models to group it into buckets of information
that you can then easily correlate to the clients and the
relationship managers that care about that type of content? So those are some of
the things that we’ll be looking at as we move
forward and just making this a richer experience,
and obviously using some of the common Google
machine learning platforms and tools to help do that. So I think just to
wrap up the discussion and recap on the journey– and this is really
an iterative thing. This is something
we’ve gone through. And then you’ll just come
back to the beginning and go through this again. And so it started with
the problem definition. It was that
relationship manager, and their day was inefficient. And how do you provide them
with more seamless analytics in an easier fashion? And then how do you use data? What data do you need? How do you put it
in front of them? What’s the platform
that’s needed to do that? So Google Cloud
Platform, choosing that as a great
platform for providing some of these analytics
and storing the data and working with it. I mean, once you know that,
then, as we talked about, security is first and foremost. So not only your core security
around VPCs and VPNs and IAM, but then when you actually
look at the data itself, what’s sensitive? What’s not? Do I have PII? Do I need a tokenized data set? So two steps of looking at
making sure everything’s secure before you start
deploying things and working with it. And then proving the technology
and expanding the footprint. So we went through a fast prove
it out, get some data there, get some manual dashboards
in front of the business. Then make sure it’s
what they want, and then come back and
productionalize data feeds and put your incremental
nightly feeds in place and scale it and put
all those things. And so once we’ve really
finished that cycle, it just comes back to
the beginning again. And it’s, hey, NTM
business leaders or users, what’s the next problem
you have for us? And then what data do we need? And you just keep
going through that in that agile,
iterative fashion. [MUSIC PLAYING]

Leave a Reply

Your email address will not be published. Required fields are marked *