Analytics requires a real involved effort across the entire organization. So what does it take to build team that not only understands your organization's data but the analytic and business requirements as well?
Join hundreds of practitioners and leaders like you with episode insights straight in your inbox.
Checkout our brands or sponsors page to see if you are a match. We publish conversations with industry leaders to help data practitioners maximise the impact of their work.
Data and analytics is a team effort that involves everyone across the entire organization. It's not just about finding someone smart who is knee-deep in the data and wants to step up. Not just the data engineers, database managers, or data scientists. The business also needs to get involved.
So what does it take to build a data team that not only understands your organization's data but the analytic and business requirements as well?
Today I learn from John Thompson, a best selling author, analytics innovator, data thought leader. John has written two books, Analytics: How to Win with Intelligence, and Building Analytics Teams: Harnessing analytics and artificial intelligence for business improvement.
John has also worked consistently and diligently to increase access and understanding of data and analytics at the high school and college levels across the US educational system.
Join the Discovering Data community!
Do you want to turn data into business outcomes and get promoted? Discovering Data just launched a new Discord server to connect you with people like you. Discover new ideas, frameworks, jobs and strategies to maximise the impact of your work. Data can be a lonely and challenging career, don’t do it alone!
Request access now: https://bit.ly/discovering-data-discord
For Brands
Do you want to showcase your thought leadership with great content and build trust with a global audience of data leaders? We publish conversations with industry leaders to help practitioners create more business outcomes. Explore all the ways to tell your data story here https://www.discoveringdata.com/brands.
For sponsors
Want to help educate the next generation of data leaders? As a sponsor, you get to hang out with the very in the industry. Want to see if you are a match? Apply now: https://www.discoveringdata.com/sponsors
For Guests
Do you enjoy educating an audience? Do you want to help data leaders build indispensable data products? That's awesome! Great episodes start with a clear transformation. Pitch your idea at https://www.discoveringdata.com/guest.
💬 Feedback, ideas, and reviews
Want to help me stir the direction of this show? Want to see this show grow? Get in touch privately or leave me a review with one of the forms at discoveringdata.com/review.
Your ideas help us create useful and relevant content. Send a private message or rate the show on Apple Podcast or Spotify!
Loris Marini: How do we build effective data teams?
Today I speak with John Thompson, bestselling author, analytics innovator, and data thought leader. John is the author of analytics, How To Win With intelligence and Building Analytics Teams, Harnessing Analytics and Artificial Intelligence for Business Improvement. Both links to those books are gonna be in the show notes as usual.
Just a bit of context. John has worked consistently and diligently to increase access and understanding of data analytics at high school and college levels across the US educational system. He regularly is featured on the Top podcast, the Top Data podcast that I personally follow, and I've learned a ton from him over the years, even though this is the first time, actually the second time that we meet live.
really the privilege of having a conversation with you, John. So thank you very much for making the time and welcome to the podcast.
John Thompson: Thanks Loris. I'm so excited to be here. You get rave reviews from people who not only listen to your show, but are on your show. And a number of people had brought it up to me. They said, Oh, if you haven't been on his show, you must and when you reached out, I was really excited. So happy to be here.
Thank you.
Loris Marini: That's good. That's good. The word is getting out there. That's exciting. Cool John, so help me understand I'm trying to really picture the person we are talking to today the primary audience for this conversation as I was going through the, that exercise, I thought of myself when I was leading a team or a stepped up to lead a data team cuz.
Typically happens, you're not born leader. You'd become the person. I go, Okay, let's build a data team. And one of the biggest challenges when you are in that state is that you don't really know best practices. You don't know who to wire first. You follow your instinct and. Bit of gut feeling, a bit of experience
you work things out along the way, which is great, but also sometimes came backfire. And you you might end up losing a little bit of time or hiring the wrong people in the wrong water.
So for. People that are in that position that are not, veterans, they're not, they haven't been 20 years leading data teams, but they find themselves building one. The business started to start believing in data. They've got some funding, they have the mandate to do it. They wanna do it well. Where should they start?
John Thompson: It's a great.
And it's something
that I've thought about a lot over the last 5, 6, 7, 10
years. was working
at Dell and I was traveling all over the world, and I hit about the three, three year mark
and I thought, I'm talking
to all these C level executives,
CEOs, CFOs, COOs, and
every time I talk to them about starting an analytics team or a
data team, you could
see the body language tighten up, you could see them get tense, and I thought, Wow, there's something
really here.
I see this
over and over again. So that's why I wrote the first book, Analytics, How To Win With Intelligence. And that book is really a primer for business executives to understand who to hire, when to hire them, how
much to pay what to expect,
what not to expect. And how long it's gonna take.
So that was the
first book and the second
book was for all the things that you just said. I've done this a number of times. I've built analytics teams 3, 4, 5 times, and I've made mistakes. Every time. We all make mistakes. That's the way it is. Unless someone is there mentoring us and teaching us, and this is really
still an art.
It's, this
is not something that's been done for 40, 50 years. This is been done for maybe 6, 7, 10 years in big organizations. We felt our way around for the last 30 years, but nobody's really systematically done it. So that's why I wrote the second. Building analytics teams is because people don't do it well and they make, I'll started some mistakes and
I thought I've made
all the mistakes, so let me write it all down so people don't have to do it by trial and error.
Loris Marini: Yeah.
Fantastic.
we need, that, we need more of
that. There's definitely,
I know why. why can't you just absorb the
knowledge and insights of someone
that already figured out
how to
do it or, how not to
do it?
John Thompson: Yeah, more than likely.
Loris Marini: More and more
likely. Tell me a
little bit
more about the context.
If you imagine hopping on a time machine and going back to the day that you resolved to write the first book and
how you felt
what prompted you, perhaps
emotionally or what were you thinking
when you said, Okay, that's it.
I'm gonna dedicate two hours every day to write this bloody book,
John Thompson: Yeah, I I remember
the day
vividly. I was having
a conversation at
Dell with my management
team, my leaders John Swainson and Michael Dell, and they were telling me some very confidential information at the time. They said we're gonna, Sell your division and
many others to get the
cash to fund an acquisition of emc.
And I said, Okay what do you
want me to do? And they said, Keep running the division. Keep running the business unit. Keep doing what
you're doing
Until it gets sold and then you'll be very happy.
And I'm like okay. I guess it's a leap of faith.
But I
thought I don't have to work, as many hours
a day as I had been trying to build the business cuz we're just gonna run it to sell it, which is probably gonna happen in just a few months.
Anyway. So the next day I sat down
at my desk and my wife
had been urging me for months at that point to write a book. And I
thought, what am I
gonna do with this extra time? And I.
I'm gonna
write a book. So I started on it and I didn't really tell anybody, I didn't say anything cuz I thought, I probably won't be able to do this.
It'll probably fall apart. I'll fail, I'll stop, I'll lose
interest. Something
like
that. And my wife asked
me a couple weeks
later, she said how, what do you
think about this idea of writing a book?
And I said I've written
two chapters already. And she just threw her hands up and started laughing.
She goes, Only you would just start and not say anything
about it.
That's how I did it. So
I thought I've got
a lot to say and I got free time, so I might as well.
Loris Marini: Yeah. On behalf
of the data community, thank you for taking the
time because I'm going
through a sort of similar process now where I'm collecting my ideas. I would love to write a book, but the more I think about the structure, the more I realize how big of an effort that is and running out on top of the podcast can be a bit
challenging.
But,
Thank you for writing the book and that's what I wanted to
say because we need this
sort of stuff more. Honestly had read the
full book to date.
It's on my reading list, but
I did skim it through and
I wanted to dive more into the structure, the second
book, the building
teams.
So let's imagine.
we work in an organization
that
understands the data is
a strategic imperative that has to be managed well, that there is a ton of business benefits and value that they can extract from it. They don't really have a proof of that, so
they, but they believe
in that enough to say, okay.
Let's try, let's give it a good
go. Let's, dedicate some
funding and build a team. And the first thing that they need to do is hire somebody or maybe get someone internally to step up.
Smaller companies
have seen a lot of the engineers and the software developers being the, typically the ones that step
up and go Okay, yeah, we've been
dealing with databases for 20, 30, 40 years.
Surely we can figure out how to do data.
assuming there's a lot of many more
scenarios, but what would be your advice to the business looking to set up that function?
John Thompson: Yeah, it's it's a, it
is a plausible scenario and
we see it over and
over again in Fortune 500 companies and smaller
companies companies
of
all sizes. And it is
something that's very pertinent and relevant to talk about today. And we do see a lot
of that. Hey, can we
find someone who
is smart and.
steeped
in our data who wants to step up and do more or different. And usually what happens is that doesn't work out too
well.
There's a lot to building a data team. There's even more to building an analytics
function. And it's not
just understanding
your data it's about understanding.
Analytic requirements
about understanding business requirements. It's about understanding the people that are gonna be involved in it. And it's
not just, the data engineers
or
the database managers
or the data
scientists. It's really the business
too.
The business needs to
be involved the subject
matter experts line of business.
It's a team effort
and analytics
is not
something that is a solo kind
of
journey. It is a real involved
effort across the entire
organization. So my, council
to anybody who's doing this is go out and find a leader. Go out and find someone that you're gonna
hire at, a vice president
level executive director, level senior vice president, whatever makes sense in your organization.
Hire a leader. And have them design it in conjunction with the other, their peers that they're at the VP level design this function with the other vice
presidents because the success
of the analytics and the data functions are tied to the commitment of the vice president of supply chain, the vice president of sales, the vice president of
manufacturing.
If those people
don't partner with this new leader and this new team. They are not going to have the impact that is expected.
Loris Marini: I love that you're targeting VPs in particular, right? Is it because heads and leads are too, perhaps technical, the not. After the business and executives are too close to the
business and so VPs sit in
that sweet
Yeah I use those titles very purposely as you noted, Loris, you need to
John Thompson: be at a
level where those people
have impact and control of the operations of the business.
Some people would say, Hey, you can hire a chief data scientist or a chief data officer, and they'd be at
the C level and I would
still be advocating for the same kind of activities.
But what
you need is a partnership between. Data analytics
and the functional areas of the business. Because if this is really gonna make a difference, if you're really going to see value realization, they're gonna have to change how they do business. They're gonna have to change the processes that run the supply chain.
They're gonna have to change manufacturing processes. And if they're not willing to do that, and you as an analytics professional, do not. have The responsibility, authority, or ability to change those processes if you're not doing it in conjunction with those business operators, value realization will not happen.
Loris Marini: Yeah,
absolutely.
Let's imagine that we are, flies on the wall
a room where the conversation
is happening.
The business is committed
and there's a bunch
of VPs and and a few
C level folks. They're all together
trying to run a discovery session. maybe
there's a candidate to lead a function.
What do you think, in your experience is the number one or maybe the top
three sort of sources of
resistance?
What kind of What kind
of hurdles can we expect when we say, Hey, you gotta change how you do stuff. And we gonna, do we gonna be data informed? Are we gonna build teams? And you'll have to build relationships.
Is that something that is well received by most people and or do
we have to think about
creating an incentive structure before we even start having that
John Thompson: Yeah, usually that's not the setting that I.
the VP of analytics
or the VP of data, having a conversation with multiple other people. It's usually a one on one conversation. And usually the way it goes is that you sit down with someone and you
say, Hey, you've been.
Talking
about this, supply chain
efficiency or manufacturing yield or pricing or something like that. You've had a strategic imperative and the CEO's been pressing
you on it. I think I
can help you
with that. Oh really?
How can you help me
with that? And we have
a conversation and start talking about data
informed analytics
and process change and those kind of things.
And then people are
like, Okay I'm getting
interested. I'm a little bit more intrigued in what you have to. And then there's usually a conversation about a project or a program,
and I do this
very
thoughtfully too, is Okay, we can do
this project and that's where we'll get
some data, we'll have
some of your subject matter experts
work with us and we'll, bring some
things together.
We'll do like a proof
of concept or a pilot, and we'll come back in four to six weeks and we'll have a conversation about what we think can be done with the. And we do that, and we do an exploratory data analysis, and we come back and we show them some insights and ideas and statistics about the reality of the business, how the business is really running, as opposed to what they think the business is doing now.
They usually walk away from that project. Really excited. But it's a project and you haven't done anything to change the business and you haven't asked them to do anything other than let you have access to some systems and maybe a person or two. Very low level,
low risk. They're not
putting anything on the table
really for any for the
most part.
But I have a conversation all the way through that project about a program. And a program in my mind is where now we're
really partners.
You're going to give us access to systems and people, and we're gonna do analytics and we're gonna come back and we're gonna show you how you may have a five step process that really could be a two step process.
Or maybe these two systems need to be integrated or maybe how you're doing personalization on the website needs to be changed. Something that involves a pretty substantial process change or a digital transformation. Usually you leave that project, you have that theme of conversation about program. And then six months later, that person comes back
and says, those insights
were really useful.
We really
found out about some customers or
operations or, production of waste or
whatever it was. And we're really interested in possibly changing how we do business. How would we do that program? So then you have a conversation about a true partnership, an investment, and a
real program to deliver
value realization.
That's usually how that conversation goes, and it usually happens over from the first conversation until you have the conversation about the program is usually about three months, and then the program delivery is probably six to eight months later, and
then the fun starts, word
of mouth starts kicking in, and then people come back and go.
You worked
with those guys in supply chain and really made a difference for them. Why don't you come over here in pricing and work
with us,
manufacturing's
got some real problems over in Germany. Can you give us a hand? So that's usually how those conversations go.
Loris Marini: So proof of value before you dive in with a massive architectural design and the long list of cloud providers that you should partner with. Absolutely. I think that's the critical step. So you said two or three months? What is the best way in your experience to round
that, proof of concept
phase in
terms of do you need
to have a already a bunch of
analysts and data engineers,
or should you
really treat it as
viable product?
So it doesn't matter if it's stitched together, it doesn't matter if it's an excel as long as you prove value, because that's not the end goal anyway.
John Thompson: Yeah. I'm a, I'm
making some assumptions in this conversation. Of
course, Loris. I'm assuming that, you've already
hired your team. You've got 'em sitting there, you've got your tools ready to go. Your data
scientists are onboard
and
operating and those that's the
environment
you're in.
Generally
we don't do anything in. Excel We We assume
that if we do something once, we're going to do it a thousand times. So we don't build things on one off bases. We don't fully automate 'em
either. But, we do everything
with databases. We do
everything with enterprise grade tools and those
kind of things.
So we don't really ever
think of a minimum viable product
in a way that we're gonna throw it. away We're always building things that we assume we're gonna use
again. We've done
a bunch of work in the
last year on donor segmentations.
We're in a business where we
have donors, giving us
plasma donations.
We've also done a bunch of work on customer and center segmentations, and all those things have been built over the last four years, and we never built anything that we threw away. We always built little building blocks that were then used later on.
There's a related
thing that people talk about, and it's true for the most part, for many shops, is that data scientists and analysts spend most of their time acquiring and integrating and managing data.
Yeah. Okay. People talk
about 80% data integration and data acquisition and 20% analytics.
I think if
you run your shop in a certain way, that is true. If you run your shop in a different way, that is not. because what we've been doing over the last four years is automating away all our data integration work and building it into a unified repository.
So we have data objects for pricing. We have data objects for donors, we have data objects for customers. We have data objects for factories. So you know,
when we are, getting into
different projects now, It's hardly ever that anybody's
going back and acquiring
data.
Of course, we bring in new data all the time, internal and external, but most of the data we use has been acquired.
It's been cleaned, it's been integrated, and it is brought in on a daily basis and
integrated into our data repository.
I don't want to use data lake or data warehouse. Those are loaded terms. it's in the back end somewhere. my data scientists are probably on any one project are spending 15, 20% of their time on data acquisition and the rest of the time they're doing analytics.
Loris Marini: Yeah. Yeah. So it's interesting
you say that. So you're assuming also at certain level
of maturity in the organizational
data
maturity.
being too scrappy. It's okay to prove
value and treat things as
projects to demonstrate that there's business value and then turn into a program.
But you wanna start with the right set of foundations. You don't wanna. Do work for the sake of just showing some numbers to
somebody and convince them
that
there's value.
You wanna actually build
the thing a list in its ins basic form.
of course that conversation would change
quite a lot in in startups
or scale up.
So any sort of organization that is trying to figure out the part of the market fit that has, or the
other way like a big
enterprise that has been operating for a long time and they. Run on systems that have been in cobble have been written 40 years ago. databases don't talk to one another. So you could go to the VP as a VP to another VP and say, Hey, I know your struggle.
I can
help you. Let me show
it to you. But it won't take you three months to get to something that actually works. It'll take you three years if
you're lucky. So in that situation it's tricky
to balance,
scrappy do
you have to be so that the other person doesn't lose their patience and you get to actually
demonstrate this has
value and
then he rate on
the second time.
John Thompson: those exploratory
data analysis really change
people's minds, because you'd notice,
or at least I've noticed over the last four years in my
current job, most people
have been in these jobs for a
long time and they have. Perceptions
of how the business is.
And when you show them how it's really
running, either better
or
worse from a data perspective,
it really gets their
attention. They will
call you outside of work hours and say, Hey, I finally got to that email that you sent me, and I'm a little blown
away by it. Can we.
Loris Marini: Yeah. Interesting. John, let's dive into that process. Slowly. Going backwards from that email is, I think, a critical point. It's a pivot point where you got someone attention. And attention. This days is really hard to get. We're
bombarded with
What seems to be an infinite stream of
priority one attention
requests.
Only a few of them are actually P ones as we know, and it's very hard to tell '
em apart. So someone that goes Hey, your
email was really insightful. That is a massive win. Now how do we get to that?
kind
of skills do we need to put in place? What kind of people to get at the bottom of the actual business
need that, that department,
that function has
right now?
Does that
happen in a water cooler? Is that a conversation that we can start there, or do we
have to, intentionally
schedule
John Thompson: You have to be very intentional about
those things, and it usually
starts one, one step back. People ask me this all the time. It's like,
you're working in 12 billion
public
company, how do you
decide what to do first?
It's fairly
easy to decide what to do first because the CEO and all the C level executives are talking about the challenges that the
corporation has. Make more money, get more product out of the material
you have. Serve more
patients, make the donors, more comfortable
and have a
better experience.
Those are
the things you work on.
You don't have to create new things to work on. There's plenty of challenges in every corporation, so you go to the people that own those strategic problems and you sit down with them and have that project program conversation and they will tell you in the first 30 seconds that you're sitting with them the first thing outta their mouth is, I'm really struggling with this, and that's what you work
people are
always asking me, How do you prioritize?
It's what are the
problems that the business has? And that's what those business executives are charged with fixing, and that's what they want you to help them fix. So that email comes from, The conversation you had the commitment, you
got to do the project.
And that email is the summary of the project, and that's usually when you get
people's attention.
Loris Marini: of course,
like I'm imagining there are two scenarios. One, you do get that email and so there's an interest and you demonstrated that you know you can help. The other scenario obviously is that email doesn't come in, whether it's because you know you didn't have the right.
raw ingredients,
the right data to do what you wanted to do.
So in terms of risk management, for those that want
to, Hey I want to
make this work, I want to help you. So after that conversation, now I know what you're struggling with.
I wanna make sure that first project is successful. Because if I I don't get
there, nothing will ever happen. You will lose interest in.
You're like, Why do I have to work
with you? You clearly
can't. So I need to
manage my risks how, the
way that I would do it,
first off is once you
understand the business need, then try to brainstorm what kind of data would you need To solve the problem, of course, there's no way to tell a hundred percent, but you can always make hypothesis and then test it out and rule out, just focus on
that 20%.
Then you need to get access to the data. And so there is a partnership that you have to have with IT or with
engineering which might
not be that easy in large companies. And then you need to put together the
team starts analyzing
this data and meshing it all together to try and answer that question.
you would also need a validation step where somebody with a domain expertise looks at what you've done
and goes Oh, this is
interesting, or, No, this doesn't make any sense at all. Or perhaps you wanna do that earlier.
Not sure. I'm just brainstorming here. How would
you manage risk in that critical
John Thompson: Yeah,
I've been doing this for nearly four
decades I have a finely tuned ear to, how people
describe problems that they have and finding elements of those problems. Can deliver success.
Now that's just me and other people can other people have that skill and you can develop that skill. So you know when someone comes to you and says something along the lines of, Can you predict the mindset of every donor that's gonna come in? And we have a million donors coming in
every year and give
me insight into how I can make each of those donors happy.
The answer to that.
I don't have insight into people's brains. I don't know what thinking. I can't do that project. That's magic. I do data science, so if there's data. Generally there's a good chance we can make it
work. You lead them
in a direction
from this, magical idea
that they've put forward into something that is data driven that you know, you can get data for, that you can do some work on that you can deliver result that is going to drive their business
forward.
Often there's
something related there that they're not saying that is an
underlying driving force.
Now in this scenario, in this
situation the way it
was
phrased was, neural implants we don't
have
those but we do
know how people respond to our marketing. We know how they respond to our pricing.
We know how they respond to poor experiences that they have, and all that data is at
our fingertips. So we can
collect all that up and do a project that comes back
and says, the driving
forces in donor experience in our business are price, experience,
and,
customer comfort or donor satisfaction, or something along those lines, or center
efficiency or employee
training or something of that nature.
And we actually did that and we came back
and said, if you want
to improve x. You need to work on Y, Z and W
and they that, I got the
response to the email like,
holy cow. We've been
hypothesizing and supposing and theorizing about this forever, but you've actually quantified it, so let's talk.
Loris Marini: Yeah.
Interesting. That you just
mentioned is the magical formula to sell anything. If you want x do y.
It's incredible how well it works. It works with kids too. I don't
John Thompson: Yeah.
Loris Marini: know , if anyone
has
kids in the audience. But I definitely
try
to mine and
surprise works like a jam.
If you want x you need to do y. Any other thing, any other attempt to try and explain it, Rationalize it doesn't work. You tap into the want and then you tell 'em what they need and the doors, the neural pathways are open
John Thompson: That's right. That's right. Yeah. Our kids are 22 and 25 and it worked
pretty well and continues
to work well as this day.
Loris Marini: Cool.
let me
recap. So we talked
about the process of starting a project to demonstrate business value. So creating new
initiatives, new new use cases
for data and
analytics, and get
to that pivotal
email that demonstrates
that there is an interest in that business function to become a partner.
With the data analytics function and explore more and do more, and actually share resources, create a shared budget or whatever needs to happen to
make sure that collaboration
keeps on going. Is that it?
John Thompson: No, there,
there's, there's always more that's how
you get their attention.
And then after
that, and then you get into the programs and you start building things that actually
show them, how value
can be
realized. We're working
through a project right now.
Where we did forecasting, at a very detailed level and everybody
came to us
and said, hey, we only need this, at the day
level for all the centers for the next eight months. We actually did it at the hour level for every center for the next two years. And
they said we
overkill.
We'll never use that. And we said, Yeah, we know you won't, but we are going to. And they're like, What do you mean you're going to, So we did a project then where we
looked at the arrival
patterns of all the donors in our centers. And then we went back and we redid all the labor scheduling gonna
Loris Marini: Oh, smart
John Thompson: save tens of millions of
dollars a year.
And we
tested it in a small set
of centers. We tested it in a bigger set of centers and now we're working to roll it out across all those centers and it's gonna save lots and lots of
money.
Loris Marini: And And you probably don't need deep learning for that, right?
John Thompson: Statistical
analysis, correlations. Now that kind of stuff works just as
well. No one cares
what it's built on.
They don't care that the forecast is a neural network. They don't care
that the, that the labor scheduling is K means
they don't care about
that stuff. What they
care about is that it's gonna be tens of millions of dollars in their budget that they get to keep.
One project
that we did recently that is nothing to do with analytics was that we had an outside vendor that was gonna
charge us, a lot of money,
hundreds of thousands of dollars a year to hand over media performance data.
And we said that's crazy.
We want that
data anyway.
Media
data on, search engine
optimization
display, all this different
kind of stuff, different
channels of marketing, execution.
We said we want that anyway for our analytics. So point us to the source systems, we'll get it, we'll bring it in, we'll clean it up, we'll load it and we'll use it for all sorts of purposes.
And we
saved them,
hundreds of thousands of dollars. Our marketing team, we saved them that money, they're happy as clams. And we didn't even write, we didn't do one algorithm. All we did was load
some data,
Loris Marini: yeah. And
you
prepared the system
so that you can reuse it
for whatever other
John Thompson: feeds.
It feeds every system that reports and analyzes marketing performance.
Loris Marini: So what's a business value of
that?
John Thompson: Huge.
Loris Marini: It is huge. It's not, it is not a hundred thousand dollars. It's way
John Thompson: No. Yeah it's in that value
calculation is into the
millions. And that was
just, simple. was just listening to a vendor saying what they were
gonna do with our data.
We're like, Eh, I think we can do better than that.
Loris Marini: You know what I find? I find that personally I got into
this data, podcasting
and data management and the strong passion for trying to understand how to do it well, because I saw at
some point I had an insight.
I understood that. If, and this was before I met Doug Laney and read his book before I talked to Bill
Schmarzo about, the economics
of data ai.
But I had this really, this gut feeling that there's something special about information and
knowledge that any
other asset doesn't
have. And now I know that is, it's the fact
that they're intangibles. They can be reused over and over, they don't deplete.
And you can
get really principle exponential ROI from them if you manage it properly.
And so that, that really got my
interest. The idea of
not doing things once and forget, but. Putting some work, spend some energy, some resources and time to lower the entropy the system. So the chaos data is all over the place now. We bring it
in and we structure,
we clean it, we use it, and every time we use it, we improve its quality because people get to see there are inconsistencies or we find other ways, things that are not quite working with the data.
And so it, it becomes this.
This loop,
this closed loop that you go through. Not one, not two, not three times, but hundreds of times. And every time you go through that loop, in principle, it should be very easy, should be low
friction and all it's should
be required from you as a data team. Is data professional, is that.
I for the 1% improvement because it's, if you combine, if you compound 1% improvements, one on top of the other, you get to incredible results while keeping your resources in the day to day limited, because it's only 1%
improvement. Talking about
migrating to a new system, you're just talking about, Hey, did you notice an inconsistency?
Great.
Flag it, create a ticket
for it. Spend that extra 10
minutes.
John Thompson: it.
Loris Marini: To fix it because tomorrow you'll find it, you'll find
it fixed. If you continue
to do that over and over, eventually
the number, the number
of inconsistencies gets below a threshold, a critical magic threshold with that, which I don't know how to calculate, but I know how it feels and it feels that people on the team now start thinking.
Heck, I can, I've got data
for this. The first
thought is not, I don't know how to do it. And you talk to people. No, I know
the data. I know it's
there. We can just use And that is that is a
massive acceleration compared to, okay, let's sit down and feel, map all the
databases. It takes forever
to do it.
John Thompson: It does take forever.
And Loris, you're
actually talking about
something that, we talked
about
earlier. We talked
about the 80 20 on data acquisition
and and the projects.
What you're saying in different terms is that if you do this right, You flip that script, just like we talked about earlier.
You have the data. You are refining the data, you are integrating the data, and you are spending less time
doing that on a ongoing
basis. You're spending more time on the value added
analytical work. Now,
another thing
that, that I always talk about and I want people
to be thinking about is.
it's not
about how many instances of yellow buses.
You have a
billion, images to
train a neural network. It's about the breadth of the sources of data
you have. We built one
application in my current role. It's got 12 different data sources in it, in seven different models underneath it, and we're always looking for ways to add more.
So my
view is that a data ensemble
of multiple sources of data are much more valuable and much more insightful
than any,
one single source of data. Even if that single source of data has a hundred times, a thousand times more
records. We as if you
think about how we make decisions
as humans, we get up
in the morning and we.
Oh, I'm hungry. Or I want a coffee, or I'm looking outside and it's raining and I need to wear a
raincoat and I gotta go
to
the store and Timmy's
gotta get
to school, how many sources
of data have you just used in that scenario? 10, 15, 20. That's how we make decisions. That's how our analytical systems need to make decisions.
So as we
integrate more and more
data into those systems,
they become better and
more robust, and more reliable and more useful for us.
Loris Marini: And there lies the
challenge John? Because
every data
integration. Work you do takes time and resources.
And one of the hard things often is to,
when we talk
about selling into the business, the challenge really is you've gotta
show value. You have to
build Some sort of solid, ideally business case for whether
anything you do when you spend
money, right?
It has to be some sort of clear roi, even though if you can put a, you can put an exact number on it, but it should be clear what problems you're
solving. Now, I see data
integration problems become cheaper to solve
because of, technology getting better
and better.
But there are
other problems with that.
What I
often see in new teams is yes,
now you've got all
your data in one place,
but there is absolutely no best practice or process to
transform it. So you get
things like, and analytics engineers writing sequel to essentially recreate what a financial tool is supposed to have or build within the data warehouse, because now all the data is there, there's obviously a ton of ways you can do it
wrong.
But I always find
tricky the process of. It's not just, It's not just connecting to a new data source. It is connecting to a data source and. Making that data compatible with all the other
data you have so you can use,
and that, that process, I think it's overlooked.
Often there's
no, and it's hard
to sell, it's hard
to explain why that matters other than, hey, 80% of our data engineers time is spent fixing bugs or understanding why
things break, the data pipelines.
And without those, there's no analytics we can do because that stuff comes before.
And so yeah. , I understand the frustration of a lot of data professionals that I'm trying to do that
John Thompson: and that's
where I think the analytics team can help the
data team, because the
analytics team should have a vision of where they're going and what they're building and what they're gonna build next. Just like I
talked about, this forecast
that we did at a, an exceedingly detailed level that everybody said that is
overkill.
We knew we
were going to the next level, so we also know that we're going to the next level after that too. So you can go and talk to the data team and say, Hey, in six months, I'm gonna need this US overlay of crime data and I'm going to need it at the census track level, and I'm gonna need it across the entire United States.
And they come back and they say, Yeah, it's gonna take us a year to integrate that. And I'm like, Great, that's good. I'm gonna need it in a year. So start working on that
now. You need to
be
looking out, you can't
be
reactive, you, the analytics
professionals need to have a vision of where they're going.
Because if you're sitting around waiting for the subject matter experts to come and say, Hey, I had this brilliant idea about supply chain efficiency, they're not gonna do it. They're not incented to do that. They're not compensated to do that. You need to go to them and say, Hey, I got this brilliant idea about supply chain effectiveness.
Let me talk to you about it. Generally, they're gonna look at you and say, That's crazy. And then they're gonna come home, go home, come back a week
later and go, this doesnt
seem so crazy anymore. What did, Let's
Loris Marini: Let me get you a coffee. Yeah,
so is that why you have a specific chapter in your book called Ensuring Engagement with Business Professionals?
John Thompson: true. You can have,
we can have all sorts of ideas. We don't own the
business, and I say
this all the time, Look, I do projects, I do programs, and my job is to make these other people look better. Now, hopefully they'll stand up at some point and say,
John's team, did all this
great stuff.
Now if they get promoted and all that other kind of stuff, that's great.
That's what I'm here to do, is to do things that make the business better, more efficient, more money, make the donors happy, make the patient's lives better, all that kind of stuff. I'm not there to take their glory. I'm there to make their lives easier and better, and they're not compensated to do that, so they're not gonna think about it.
They're just gonna think about, Hey, what are my quarterly objectives? What are my annual objectives? How do I do this? How do I do?
I try to come
in and be a support structure that has a vision for the future that will take them to a place that they probably hadn't thought of before.
Loris Marini: Boom. When I extract that as a quote and print it on my I'm going to be a support structure to the team. So it's about enabling. It's not about owning, it's not, it's about coaching for development and performance. It's not about necessarily
advising and saying this
is how you're supposed to
do things.
John Thompson: Because you
don't own it. You can't tell them how to do it. You can show them a better
way. It's, the old thing
about leading a horse to water, but you can't make '
em drink. And it needs to
be of the scale that it's interesting enough
to them, this project
that we're doing, that we're trying to get
implemented in labor scheduling,
it's tens of millions of dollars.
Now, when I told people that, they said, Oh, that can't.
We're better
than that.
And I'm like I don't know.
Let's test it. So we did the first test
then maybe that's a fluke, Then we did
a bigger test and
they're like, I don't think
it is a fluke. It actually looks like it
works.
A lot of times
people, as we, when we first started this
conversation, people always
wanna be in
the know they're the
experts in these area. They don't want to have someone come in from the outside for the most part and say, Hey, you really
could be, saving $35
million here.
That's a big number.
And if they're
not secure in who
they are and feel good
about where they are in the organization, they might easily reject that.
But when you
come around and say, Hey, that 35 million is going right in your
back pocket, for you to
use some other.
Then they're like all right,
let's work on this.
Loris Marini: Yeah. The incentives the incentive
is strong enough
to
John Thompson: Yeah. I don't, my team doesn't
have a dog in
that fight. We're just
there to help.
Loris Marini: Yep. So that's how we select winning
projects is talk to business lines develop that
we establish a genuine relationship first,
and have some sort of, confidence
that we can solve the problem with the data
we have. And then so we don't
make empty
promises and then demonstrate
value with the.
define timeframe.
So a project, not a program. Get to show
value and then the rest will come. We, when once you convince them, they will partner with and it will share resources
and you will keep collaborating.
John Thompson: right.
Loris Marini: Awesome.
Good. So John, help me recap
this. I've been trying to do this 60-second summary.
It doesn't have to be
60 seconds, but you tell me when when you're
ready and we'll try to
summarize the conversation
in 60 seconds for someone that just tuned in and I
don't have 45 minutes to listen
John Thompson: Let's do it.
The conversation Loris and I have been having today is around the high level concepts of how to be successful in an analytics team, partnering with subject matter experts, partnering with executives, focusing on value realization and picking projects that. Are already aligned with what the business users have and the struggles that they are trying to solve.
So we've just barely scratched
the surface on building
and managing analytics teams, and we really tried to talk about how the data and analytics teams can collaborate together, how the subject matter experts and executives can collaborate with the analytics teams and how it can come together in a way that makes a difference for the.
Loris Marini: Boom. Fantastic. Thank you very much, John. That was a fantastic
summary and I think we
covered all the main talking points for us. Do you think there's
anything left to to say or anything that you wanted
to cover
before we we wrap it up?
John Thompson: Yeah. I think that
we didn't get to architecting an analytics
team. We could probably
do another entire podcast on
that. That's a whole
topic that we didn't
touch on. We could talk
about that at a different time. There's just, at a high level,
there's the, there's the.
Loris Marini: you made
John Thompson: There, there's the factory team where you have individuals that are focused on individual jobs like feature engineering and model
building. There's the
other side where you have artisan team members that handle everything about a project. And then there's the hybrid where you have a factory team that supports the
artisan team.
And that's
a really interesting topic and I've spent hours talking
about it. And in my
current job I started out with an Artisan team, and then I supported it with a factory team. And for the last year
I've had an artisanal
or a hybrid
team. So it's it's intriguing.
So if
anybody who really wants
to delve into the nitty gritty of managing an
analytics team, we could pick
that up another time and talk about it.
Loris Marini: That sounds like another podcast, , that's for sure. . Awesome.
Awesome. John Thompson bestselling author the two books
we mentioned today. One is Analytics, How to Win with Intelligence. The other one is building analytics teams, harnessing analytics and AI for business improvement, both links and the
show notes.
As usual,
John, thank you
very much for your time. I truly enjoy
this conversation. I hope you have two and I'll see you on LinkedIn.
John Thompson: Thanks, Loris. I really did enjoy it. I look forward
to our next conversation.
Loris Marini: Ciao.