June 20, 2024 • 43 mins
This week, on How I Met Your Data Anjali and I have a special edition for you. We're going to debrief on the Snowflake Data Cloud Summit that took place just a couple of weeks ago.
Joining us is Michael Cochrane, Cervelo's Global Data and Analytics Consulting Practice Leader, and a longtime colleague of mine.
We dive deep into the evolution of Snowflake, its impact on clients, and how it has transformed from a cloud-optimized data solution to a comprehensive platform supporting multi-cloud and real-time data needs. Mike also discusses the significance of governance, privacy, and cost management in today's data landscape.
We explore the latest advancements and releases showcased at the summit, including Snowflake's partnership with NVIDIA and the push towards native apps and containerized services. The conversation also touches on the increasing complexity of data architectures and the importance of maintaining transparency and optionality in technology investments.
Join us as we reflect on the exciting developments from the Snowflake Summit and discuss the future of data engineering, AI, and the ever-evolving role of technology in driving business value.
Mark as Played
Transcript
Episode Transcript
Available transcripts are automatically generated. Complete accuracy is not guaranteed.
(00:00):
Hi, Sandy here. Welcome to another episode of How I Met Your Data.
This week, Anjali and I have a special edition for you. We're going to debrief
on the Snowflake Data Cloud Summit that took place just a couple of weeks ago.
And in order to do that effectively, we actually asked one of my colleagues,
Michael Cochran, to join us.
Mike is currently Cervelo's Global Data and Analytics Consulting Practice Leader.
(00:22):
He is someone I have worked on and off with for over 22 years of my career,
and someone I deeply admire.
So I met Mike in 2002 at a company called Painted Word, where we both started
off there as front-end developers.
Our roads absolutely diverged. Mike went from front-end development to wanting
(00:43):
to learn a little bit about the back end.
So he dove into data warehousing and data warehousing concepts.
I remember just seeing his nose deep deep into books at his cubicle for days
on end, if not months on end, as he got all his certifications back in the day.
He also became a software developer.
He used to build most of our custom solutions, our web custom solutions as well
(01:08):
during the later part of our career there.
He moved on and became the CIO for the Palladium Group, where he learned all
the challenges CIOs his face by living in that role day to day.
He did that for a number of years.
He then joined Cervelo as the first employee and non-founding member of the
firm, where he was quickly asked to think through the cloud strategy and look
(01:32):
at all the technologies that were coming to fruition about 14 years ago when we were founded.
So he did that and quickly culminated into him becoming the Global Data and
Analytics Consulting Practice Lead,
which he co-leads today with another colleague of ours, Glenn Heatley.
So now that you have that background on Mike, we can go ahead and dive into the conversation.
So with that said, let's dive in.
(01:52):
Music.
Hey, Mike. Hey, Mike. Hi. Welcome. How are you? Wait, this is recording already? already.
Yes, we're recording already. You got to give me a heads up.
(02:18):
There's no video. Just walking right into the show.
We like to keep it spontaneous.
Welcome, Mike. Thanks for joining us. I know you and Sandy just recently got
back and have recovered from your time at the Snowflake Summit.
Cervelo has been a Snowflake partner since the very beginning, I think.
(02:39):
2016, I think. They went GA in 2014.
We were about a year into that. So it was probably late sometime in 2015.
And then really pushed it in 2016. Great. Cervelo has been a partner for a number of years.
And I think there's been a lot of changes that we've experienced in the platform
(03:00):
in the way that our clients are really deriving value.
So I'd love to hear from you. What have you seen as that snowflake evolution
that's occurred and how has it been impacting our clients?
Yeah, it's a great question. I think it's been a pretty interesting journey to date.
When we were first introduced to Snowflake, I think my impression was that here
(03:20):
was a company that finally addressed how to build a data and analytics solution
in the cloud that is optimized for the cloud.
I think up until that point, which Cervelo was founded in 2009,
we really dove into our cloud strategy in 2010.
And what you saw was a number of different companies that were taking legacy
(03:42):
technology, and then just kind of putting it in the cloud and then calling it
cloud software or cloud platform.
And it really wasn't optimized for that. You ran into all kinds of performance
issues, you ran into scalability, you ran into portability challenges,
it's just the list sort of went on, right?
So you always kind of ran into a brick wall with a lot of these different companies.
And when we first saw the demo of Snowflake, we were blown away.
(04:05):
It was all right, this is going to address this problem, this is going to help
this client, the number of things that we saw value in that could really contribute
to our client success was pretty amazing.
And, you know, I remember I was being in a QBR in New York with them early on in our relationship.
And somebody asked me the question, hey, what are your thoughts?
(04:25):
What should we be thinking about?
I said, listen, you know, your platform and your technology is great,
but you've got a limited window where people are going to start to catch up.
And therefore, it's like, what's the next thing? And what's the next thing?
So as I think about where they've really come over over the last however many
years, think they've sort of gone down that strategy path, right?
They started with, all right, here's a technology that's built and born in the
(04:50):
cloud, and then really realized that, okay, our clients have other needs and
use cases that we need to close the gap on.
And so, you know, seeing them add things like multi-cloud support.
So getting beyond just AWS, opening up for Azure and Google, And now more recently,
being able to combine all of those capabilities in more of a hybrid structure
(05:13):
is pretty impactful for many enterprise clients that have a multi-cloud or hybrid strategy.
And so it's been great to see that evolve over time.
The idea of clients having the demands of we need data in real time, right?
You always had to dig into that need or that requirement because it wasn't always
(05:34):
real time. It was, all right, I need it just in time or at the right time.
But now the way that technology is advanced, real time is a reality and it's
really table stakes now.
Like if you're not delivering things in real time or near real time,
you're missing out on many opportunities.
So their ability to start to close that gap as well has been very interesting.
(05:58):
And, you know, I think more recently coming out of the conference.
I always tie it back to what's happening in the market and where people demanding
probably falls in a few different areas.
But I'd say the top ones are customers or consumers having more control over their data.
So privacy is pretty key. And so when you think about what they're doing to
(06:19):
build in that compliance, to build in that governance, data clean rooms is an example.
And being able to share data without having to move all kinds of private consumer
or customer data is pretty important.
And so seeing them put emphasis behind that when we know it's a big issue for
our clients and something that they're struggling with and they need help with
(06:42):
is extremely important.
Cost management, cost optimization. So FinOps, seeing them make strides there
as well, been pretty interesting to see.
And then just the complexities of now there's a big focus in AI and you have
all these different technologies that you have to make interoperable,
them going beyond just data and bringing compute in the NVIDIA relationship
(07:06):
and bringing that to the data versus the data going to the compute and having
to move all of that around and then building capabilities around apps and.
So when I talk about complexity, all of these things together add complexity
to our clients' data architectures.
And they've really been able to bring all of that together with applications,
(07:31):
with the governance, with data, with high performing compute that AI applications will need.
It's just been a pretty remarkable journey seeing all of that come together.
And then, Sandy, I know you and I were kind of joking about this,
but the very first conference that we were at, the entire room was probably
(07:51):
smaller than the partner keynote at the most recent conference.
So, I mean, that just speaks volumes to how big they've gotten in the ecosystem
and adoption that's come along with that.
Yeah, it's funny because when we first, I still remember telling you,
hey, we need to look at this.
And if it's as good as what they told me, you need to be all over this.
(08:12):
Because I had, I remember talking to the rep and they were telling me what it could do.
And I had a flashback of like the last nine years of my career and then heavy
nights and weekends talking to architects, DBAs, trying to get them to do something that was basically.
Impossible and impossible feats of performance that we were asking of them over
the last nine years prior to Snowflake.
(08:33):
And then seeing that, I mean, I still remember you and I were like, what?
Like, what is going on here? But if you fast forward from that point forward,
I feel that they did have a large
window of being the only one in the marketplace that really could do.
And they actually said it in the conference. They said, hey,
look, we don't just take storage from compute.
We're separating compute from compute. Nobody has done that.
(08:55):
I don't think anybody still hasn't done that holistically the way they have.
So they did have a big time in there in the marketplace where they were the only one.
But now things are changing with data science and AI and ML and all these other
capabilities that people are trying to build.
So they had to accelerate. And I feel that.
The change we see in them today only happened over the last three years.
(09:16):
They've been around for eight, nine, right?
Five years of that, it was just growing a data capability, very traditional
in the cloud, making it work, making it very different than what people had done in the past.
But now I think they're really accelerating that and thinking beyond that because that has been solved.
And I think that that's what's super exciting about this is being able to see
(09:38):
them take this to a completely different place and truly make it a platform
where anything you have to do on data, you can do on there.
Yeah. Yeah. I think it's like a really exciting time to see what is coming out from Snowflake.
And so that kind of brings me to my next question is you've been back for a
week now from having been at the summit, you've seen and reflected on some of
(10:01):
the advancements and releases that were showcased at the summit.
So what were those things that they were sharing with us? And what were your
takeaways from those those demos?
Yeah, the you know, just like any big conference, right? There's there's always the the big reveal.
So I'd say some of the things I'm most excited about probably falls in a couple camps.
(10:23):
One is probably the less exciting stuff, you know, to me, protecting our clients
data to making sure that they're governing it properly.
If you think about the ability to productionalize AI, which which was a big, big topic.
Companies really focusing on experimentation and trying to figure out the use cases last year.
(10:44):
And then this is the year of productionalization.
I agree and disagree with that. I think many companies are trying to productionalize
and get the real value out of AI.
But at the end of the day, AI is only as good as the data that you feed it.
So it's this whole thing of garbage in, garbage out that we've been talking about for decades.
It's the same thing with AI. AI doesn't change the fundamentals and the block
(11:07):
and tackling that needs to happen with your data.
So I'm happy that they're putting emphasis around the governance and the transparency
and the compliance aspects of it, because that's going to be very critical as we move forward.
And as customers and consumers get more control over their data,
as I had mentioned previously, like that is key.
And you have to be in a position to be able to react to those realities.
(11:31):
On the flip side of that, you know, more of the innovation in the AI side,
the fact that they're pushing on the concept of native apps and containerized services,
and then bringing a lot of the machine learning and LLM capabilities to be able
to build these chatbots and other customer or client facing solutions.
(11:51):
I think it's of high value. I think it's still early days in terms of where
this will go, but I definitely see them on a great track and a great trajectory.
Our clients, when we speak to them.
It's all about speed to market, right? It's speed to market to differentiate.
It's speed to market to stay relevant. It's speed to market to be able to continue
(12:13):
to be productive and efficient in the work that you do.
And so that requires acceleration. That requires applications.
It doesn't just mean, hey, I need data and then I can do anything with data.
It's like, well, I need data. I need data for supply chain. I need data for
finance. I need data for a number of different use cases.
(12:34):
And so to see that you can now,
all of this kind of coalescing and the lines being blurred of applications versus
data and compute versus storage and those things are very important and very
relevant in the modern world.
So I'd say those are some of the big takeaways for me.
One thing I'll be critical of is I was looking for maybe a little bit more of
(12:56):
a bigger reveal on some of the things. So the things that they've had in private
preview, I was hoping would become publicly available.
A couple of them are pretty close, but not quite there yet. And I know those
have real applicability to some of the things that we're doing with our clients today.
So I'm really looking forward to when those do become publicly available and
we can productionalize those for our clients.
(13:19):
I totally agree with you, Mike. I think one trend, though, I've seen with our
larger clients is creating kind of a bit of a commodity around data engineering.
Larger organizations start to look at it as a commodity and less of a thing
that they have to have really skilled workers in, which is a bit of a challenge.
So the thing I was excited about with Snowflake was that they're doing things
(13:42):
like there's a co-pilot, but now that co-pilot is in kind of the SQL window itself.
So you don't have to go somewhere else to get, is this right?
You can actually finish your code right in there.
So the AI is helping you. So a lot of those developer tools,
they announced a ton of DevOps stuff like monitoring, Git integration.
Little things like that, that they've decided to shovel into the platform.
(14:03):
Because the bare bones capability of moving data from one place to another,
making sure it's transformed effectively, ensuring that it's available to the
business, that work still needs to be done.
And the easier Snowflake can make it for your engineer to do the right thing,
the better everybody's going to be. A lot of their competitors aren't spending
time there trying to make it easy for the developer.
So they really are trying to make it easy. And that, to me, left an impression
(14:26):
because I'm not a very technical person.
So if you can explain something to me and show it to me in a way that actually
makes sense and I'm not completely lost in where you are in your windows.
That's actually pretty enticing because that means you can take a junior developer
and throw it at them and they'll be able to navigate that. No,
I tend to agree with you, Sandy.
And it's interesting because this isn't just isolated to Snowflake.
(14:48):
I think what you're seeing is that the things that are maybe you would consider
commodity, right, within that role.
So maybe connecting to different data publishers, right?
So like an ERP system or CRM, you know, anywhere where you need to get data,
essentially, those things have evolved over time with different connectors.
(15:09):
So you don't have to build those, not that you don't have to build them anymore,
but there's a whole ecosystem around that.
The way that you transform and work with data is now evolving.
Some of the things that you had to build in, in terms of validation controls
and those types of things, those are becoming a lot more easier.
So in a lot of ways, the engineering role is becoming more like a data orchestrator
(15:32):
in a way that just an engineer.
So it's like somebody who really needs to understand, well, where does my data originate from?
And then what does it need to look like on the other side?
And you still need to make sure that all that high quality governance and all
those things are built in.
And so I think what we'll see is an evolution of that role.
But companies like Snowflake are really investing in taking away the inefficient
(15:58):
work that has to happen, creating the assemblies, creating connections in the databases.
Repeating the same data quality routines over and over again and reinventing the wheel.
Things that are kind of table stakes, these companies are taking care of.
And therefore, data engineers are going to have to think about,
all right, what does my role start to look like and how can I elevate to that
(16:21):
role, knowing that some of the commodity things will go away over time.
And another part of that is the native app framework that they've created,
because anybody can create an app, if you will, that functions with data.
And then you can leverage it in flight and snowflake. We have a couple of life
science clients and everybody's talking about the standard of fire.
(16:41):
I saw there's a native app that allows you to do that.
And I was just like, wow, you just put like 20 consulting firms out of business
with a native app. Well, and that's right.
I think that and just and then just market awareness. Right.
So now that they have the marketplace and being able to allow partners,
allow customers, allow the entire ecosystem to build these applications that
(17:05):
can address fundamental issues and make those readily available.
Similar to what we saw with digital natives back in the early 2000s and beyond,
open source, a lot of the innovation they had, it's sort of akin to that, right?
Like somebody solves a discrete problem, make it available.
Some of that's going to be paid. Some of it's going to be free.
But at the end of the day, it creates this awareness and this transparency to
(17:28):
allow everyone to evolve and to innovate.
That's exciting to see because at the end of the day, the more eyeballs you
can get on a problem and have diversity of thought and diversity of how to solve
that problem just makes it all the better for all of us. Yeah.
Yeah. It leaves us room to tackle the harder stuff, right? Not the stuff that
we've already solved time and time again.
(17:50):
Exactly. Because at the end of the day, you're going to be solving the specific
thing for the client, not all the stuff that sits behind it that business people
don't typically see, right?
You know, it's that whole iceberg image. I only see the stuff at the top of
the water, but not all the mess that kind of sits below the surface. That hasn't changed.
And so it's like, how quickly can I get to that tip of the iceberg and show
(18:12):
that value to a customer, to a client? That's the important thing.
And so all this other stuff, automate it as best you can.
You can't automate everything, but that's really where the world is going, right?
It's automating a lot of human behaviors that are repeatable.
And I think what we're seeing is your data engineering point,
Sandy, doesn't just stop there.
(18:33):
It goes into analytics and a number of other things.
There's a lot more automation of that human behavior that we're seeing.
And so everyone needs to figure out how do I adapt to that? What's my strategy
to do that? What are the people I'm going to need? How do skill sets evolve?
How do we train people? How do we deal with the change management aspects of it?
(18:53):
There's a lot that goes behind that. So as exciting as it is and the impact
that it's having, you know, you really got to keep your eye on the prize that
it does take a lot of hard work to do it right and do it successfully.
It's just the efforts are being shifted to more high value work.
Than maybe some of the commoditized things that Sandy mentioned.
(19:14):
Yeah, for sure. And do you think our clients are going to understand and realize that benefit sooner?
Well, I think there's certainly a recognition that I don't think anyone is sitting
there saying, oh, this stuff is going to be super easy.
But at the same time, it's pretty clear that, look, all the stuff that you've
had to do for decades around managing data and making sure data is of high quality,
(19:38):
creating transparency, transparency observability all that all that great stuff
that takes work and at the end of the day it's it's
about process it's about diligence it's about
accountability and ownership it's all these things that
we talk about that just doesn't go away and so it is making sure that that's
crystallized and that if you're talking about a ai effort right or a transformation
(20:01):
for that matter all of the stuff you got to do to get your data right doesn't
change you still got to do it now Now, how much of that can be automated?
How much of that can be fast-tracked? How much can that be addressed time to
market? That will certainly improve, but it certainly doesn't go away.
And the effort is still pretty big. Yeah, no, it sure is.
Mike, one of the other things that you had mentioned was the NVIDIA relationship.
(20:25):
So that was one that we heard as an announcement.
Would love to hear your reactions to that, as well as any other partnerships
that had been announced during the summit.
Yeah, I think the NVIDIA one is pretty exciting. They're obviously a juggernaut.
Ever since OpenAI released ChatGPT and the world just ballooned into the AI revolution,
(20:46):
the amount that NVIDIA has been able to do in this time period,
the emphasis on their chipsets to be able to do massive amounts of computing is pretty impressive.
And bringing that to the Snowflake data cloud or AI data cloud,
it's incredibly important because these use cases are only going to continue
(21:09):
to evolve and the amount of horsepower you have to put behind that is only going to evolve.
So I think it's pretty exciting that there's this strategic relationship to
try to make this as efficient.
And cost effective as possible. And so that's sort of what I see that this relationship
being is how do we bring the best solutions to customers and clients while minimizing
(21:35):
a lot of the complexity that goes in into it?
Because, you know, we're seeing our clients go down this path.
And when they start thinking about the cost aspects and the sophistication of
technology, there's still a big learning curve there, right?
So I think as much as that can be sort of packaged, commingled and simplified,
(21:57):
the better. And I think this is just a step in that direction.
I think there's going to be a point of view fight between Databricks and Snowflake
on what's the right approach with managing costs, because Databricks announced
the ability to just forget it.
Like, don't worry about it, we got it for you. And I thought that was pretty interesting.
It's just like, no, you don't have it for me. I want to be able to manage this myself, right?
(22:22):
I have to trust you now that you're using the least compute possible for my workload.
And what if I want it to be faster than what you're providing to me?
It's just so odd. They announced that and I cringed.
I immediately cringed. It reminds me of when we were trying to sell Snowflake
because all the DBAs were holding on to the fact that they could tweak the levers.
(22:43):
And I was like, don't worry about it. It's got it for you. You just have to
worry about the size of workload you need, et cetera.
So it's the same feeling. I had that feeling when Databricks announced it.
And I was like, that's not going to go well. And I'm not even the one building this stuff.
So I don't think people will have a reaction to that.
I do like the idea of not having to worry about it, but there's a trust factor there, right?
(23:03):
You're now, you're asking your customers to trust that you have their best interests
in mind, which is not always the case.
Yeah. And it's very different, right? When you talk about feature and functionality
versus, well, am I getting the value or the ROI in what I'm actually spending on, right?
So when I think about feature and functionality, it's okay if it's very simplified.
(23:26):
There can be that layer of abstraction of an understanding that,
okay, I just need to know it works. I don't necessarily need to know everything.
Science or the physics behind how it works, unless you're nerds like us,
and you really want to dig into the detail.
Well, how does that actually work in reality? But when it comes to cost,
you need a high level of transparency.
Think about Google BigQuery, challenges that our clients had with that in the
(23:50):
early days of, okay, well, we're spending money, but we don't quite know what
we're spending money on.
And so I think that transparency and being able to see which departments,
which divisions, which business units, you know, what are they spending? What are they using?
I think it also having that transparency helps to drive innovation, right?
(24:11):
So if you think about, well, all right, it costs X amount to do this,
and we feel that's too high, make that better, right? On the feature functionality side of things.
Don't try to demystify what we're spending money on because clients and CFOs
and COOs, they want to understand that.
CIOs want to understand that I'm getting the best value for my bucks,
(24:35):
so to speak. And that's important.
On the feature functionality side, I just feel that, hey, if it works and it
works great, you're not going to get asked a lot of questions.
But if you're given a number that has a currency associated with it,
people want to know the details and they should have that level of transparency.
Well, it doesn't just drive innovation. It drives decision-making on the business
(24:58):
as well, because the business could be asking for something that costs a lot of money, right?
And it could be as simple as I want that data load to happen on an hourly basis
or every five minutes, all right?
And that could be taxing because if I need to get it done every five minutes,
maybe I need a larger compute engine to make that happen.
But I'm looking at that going, well, and this has happened to us where we're
(25:19):
getting into these conversations with our clients and saying,
well, actually that decision is worth $10,000 a month.
Is it worth to you $10,000 a month to have it seven times a day?
And the answer was actually, oh God, no, it's not worth it. Right?
So you can have those really pointed conversations about, I want to deliver
X, it's going to cost me Y.
Is it worth the, not just the effort to make that happen, but actually the ongoing
(25:43):
costs to enable it? Yeah, exactly.
Because otherwise it becomes one size fits all.
So if you need to determine, is this delivering on the business value?
If you have a one size fits all model, you may not be able to quantify that answer, right?
But if you can break it down to your point, if you just took a very basic example
(26:03):
and say, well, okay, if I don't have to run it as many times during the day,
or I can scale back in terms of how long it runs,
that's the optionality that people want, because they may be able to say,
well, we need to bring the cost down in order to deliver on the ROI.
Well, great, I have the levers to be able to do that.
But if I don't, then it makes a much harder choice.
(26:26):
So in some ways, you feel like you're simplifying things for people, but at the same token,
you're creating challenges. Because at the end of the day, when you talk about
investments and investments in technology, they're always going to be scrutinized.
And they always have to have detailed backing behind them.
So to me, it's almost like you want to create more transparency, not less transparency.
(26:49):
For sure. For sure. And I'm really glad you brought up Databricks.
Because if I think back to 2016, 2017, we would have these singular conversations about Snowflake.
It was just Snowflake. And then, you know, you fast forward a couple of years
and now you're either having a conversation about Snowflake or about Databricks,
but never really a joint conversation.
(27:11):
But I think we've seen a lot of clients starting to move to a model where they
have both Snowflake and Databricks in their environment.
So, you know, any thoughts on Snowflake and Databricks better together?
Yeah, I think what's and this was one of the things that that was covered in
the in the Snowflake conference, and I think is, is very important.
(27:31):
And I'd say it's a very big takeaway, I think, for for a lot of companies is
that you're always going to want to have optionality.
And there's no one technology that's going to be able to solve every problem for you. Right.
And so bring those two things together.
And what do you have you, you need in an architecture that is not honed in,
(27:53):
on any one technology, right?
And so when you think about the Databricks or Snowflake, to me,
it's being agnostic and taking a step back and saying, well,
what are my objectives? What's my strategy?
What business outcomes am I trying to drive towards? And then what does my technology
ecosystem and architecture need to look like to support.
(28:15):
And, you know, one of the things that I thought was interesting,
and one of the partner sessions, Sandy, you and I sat in on is that Snowflake,
in many ways, is beating that drum of making sure you have an agnostic architecture.
And so that's not only inclusive of technology, but also the patterns that you
deploy as part of that, right?
Whether it's data fabric, data mesh, data lake house, data, data lake,
(28:38):
I mean, the list goes on, right? And there's different reasons why you may choose
one approach over another.
There's different reasons why you may choose one technology over another.
But at the end of the day, it comes down to, it's probably going to be a little
bit of a few things that you're going to need to make sure you have modularity
built into your overall architecture.
(28:59):
And that's going to require multiple technologies at the end of the day.
And so really what you have to look at is what is the business strategy of these
companies And where is their secret sauce and what are they well known for?
Right. And when they start getting out on the fringes of stuff that you sit
there and say, OK, this is this might be a stretch or this might be an area
of growth that maybe doesn't make a lot of sense.
(29:22):
Well, then you may want to stick with best of breed in that case.
So, you know, I think it's important of balancing those things,
of continue to be agnostic and knowing that there needs to be an overall ecosystem
of technologies that have to, you know, have interplay with each other.
And to me, it's not just going to be Snowflake or Databricks, one or the other, both.
(29:44):
You know, I think it really depends on the situation and the client and their
investments and their strategy, all the things that I mentioned.
And I think we'll see that evolve over time. It was going to be acquisitions
in the market and a lot of these AI companies are going to become part of bigger companies.
So I'd say in the next five years, it'll be interesting to see how it all plays out.
(30:05):
At the end of the day, it's what is your product strategy? Where are you taking this?
And I do think Snowflake's on the right track. They're not trying to break out
into areas that just are far removed from what they do.
They're creating that ecosystem around them to be able to do that,
and they're staying true to the core.
And that's important, I think, because once you start deviating from your core
(30:27):
strategy, that's when things get messy.
And when you start seeing lots Lots of acquisitions and then stuff starts to fall apart.
It wasn't as great as it used to be. And that's just, it's sort of the evolution of growth in many ways.
But it'll be interesting to see over the next five years. Absolutely.
I think what's been interesting is the whole true decoupling of data from the compute, right?
(30:48):
Because now with Iceberg, because Databricks bought Tabular,
Tabular allows them to use anything, including Iceberg and other formats. mats.
And Snowflake is making this huge iceberg push to be able to work on data sitting on iceberg.
So I think it's just going to be a matter of a rush of who's actually able to
work on data and compute with data faster,
(31:11):
and who has the most tools and the tool belt to enable you to do that.
Is it going to be Databricks? Is it going to be Snowflake?
It could be both. Your data could be in one place and both these platforms can now work on it.
So that's like a completely different thing that I can literally put it in this like other thing.
And then both these platforms can actually do workloads on that data set.
(31:33):
And not move it. That changes the game. And even Salesforce got in on it, right?
So Salesforce now can load data in and out of Iceberg as well,
or read data in and out of Iceberg as well.
So this whole interoperability of where your data could be sitting in this one
thing and all these other applications and platforms can work on it, touch it, read it.
I think that's a whole new world.
(31:54):
That's a great point. Totally agree. So with all that being said,
were there any Any even better if announcements that you were looking for?
Maybe a little bit more information, a little bit more clarity?
Yeah, I kind of come back to my earlier point. I think some of the things that
wanting to see come out of private preview be generally available would probably
want harder timeframes of when that's going to happen. So I'd say that was one of the takeaways.
(32:18):
I think the other thing is you always want to see customer cases and customer stories, right? Right.
For me, going to these conferences, it's less about understanding the nuts and
bolts of the technology, but more what's the applicability of that technology
to a business use case. Right.
How is this actually going to help move the needle for for one of my clients?
(32:39):
But definitely want to see where people are being successful,
but then also being transparent and saying, hey, we're not being so successful here. and here's why.
And these are some of the things that we've tried. Because again,
more information and more eyeballs and more voices in this just makes it all
the better for everybody else.
Were there any use cases that maybe are different than what we've seen with
(33:03):
our clients and maybe something that we should be considering?
I'd say just the contrary.
I mean, it was good to validate that a lot of the things that our clients are
struggling with and the solutions that we're bringing to bear for them.
You were seeing other customers of Snowflake in the same camp and trying to
address maybe in different ways or nuanced ways.
(33:24):
But at the end of the day, I think the same client problems and themes are consistent.
And so I think in many ways, it was just a validation of a lot of the great
work that our teams are doing to support our clients.
Good. So our clients' challenges are not each special snowflakes.
That was such a terrible pun.
I mean, I didn't see anything crazy and outlandish. I actually agree with Mike.
(33:48):
I think what I saw, and most of it was from Snowflake, like the art of the possible
with these challenges that we've seen clients have time and time again.
Art of the possible with all the new capabilities that they're bringing to there.
That was what I was noticing was, okay, there's a different way to solve this,
that they're pitching to their partners and customers, because obviously no
one's doing it quite yet. They haven't released a lot of this stuff.
(34:10):
So it was very much pie in the sky conversations about where it was headed.
But I felt there was a shift in terms of, I don't have to ask the question anymore
of, can I do something? Like, I think you can do a lot now.
I think the question now is what should I be doing? And that's something that
we do very uniquely with our clients.
I think a lot of people wait to hear from their clients in terms of what problems
(34:33):
they're trying to solve.
And we kind of are front footed in terms of helping our clients figure out what
are the right problems to solve And then how to solve them becomes secondary.
So that, to me, was exciting because it allows us to have more freedom in terms
of the advice that we're giving and the propositions that we're discussing with
them. Yeah, I think that's a great point. I mean, I see a couple things.
(34:55):
So one is, I think the art of the possible, I think there's a recognition on,
yes, a lot of these things are real and quite possible.
I do agree that there's a bit of what are the use cases we should be going after
and where can we get the most pull through.
I think there's going to be continued on the art of the possible side of it,
(35:16):
of just understanding and learning of the applicability of AI and gen AI.
I mean, it's not like these things are new, right? They're just,
they've evolved very quickly.
And that's what's opened up the aperture, I think, for many companies.
But this has been around for decades, right? Right.
So it's just now it's just supercharged and it's it's a lot more realistic for
(35:38):
companies to be able to do this at scale.
But I think there's a big learning curve that comes with that in a number of
different areas. Right. We talk about responsible AI.
We talk about high quality data. We talk about hallucination.
We talk about all these different things. Well, people have to learn what does that all mean?
And how do I make sure that I'm addressing all of those things on equal footing, right?
(35:58):
That I'm not letting one thing to chance that could really have,
you know, bad implications on your company.
And so there's, there's going to be this continued learning that that goes along
with that being true to understanding where you are in that journey, right?
And what you're capable of doing, you know, short term and long term,
right? And how you build up to that.
(36:18):
And what are the use cases we should go after? or will they create the value that they're meant to?
You know, you sort of have to get past this building proof of concepts and prototypes.
And so I think you have to be realistic about what companies can do and what
they can evolve to over time.
What I'm seeing is that a lot of clients, because a lot of these capabilities
are now sort of real and evident and right in front of them,
(36:42):
you know, they're off building a lot of different things.
And so it's almost like you got to take a step back and say,
all right, you need to see the forest or the trees.
Are these investments you're making and some of these short-term things,
because you can do it, are they the right investments to be making right now?
Because what you may build today, where there may not be a solution,
just a lot of capability to do it, well, those things could just become available
(37:04):
in the next year or two, right?
So you have to balance the investments that companies are making and building
things bespoke versus things that are going to start being built into different
software, different platforms.
Different maybe open source capabilities, you really have to look at that and
have a view of the future.
Because case in point, a client came up to us recently, and they said,
(37:27):
hey, we want to take this beyond POC, we want to build this.
And I said, well, maybe one of the things we should first do is is just do a
quick scan of well, what's happened in the last six months since you started
on this journey, because there is there's a lot more advancement that's happened.
And as a result of that, there's sort of a recognition, okay,
it probably doesn't make sense for us to build this at this time.
(37:50):
We can build a portion of it, but we know that there's some other things coming soon.
And so let's treat it like an evolution, right? Let's build the things that
are going to differentiate, but let's not build things that aren't going to
give us the true value proposition that we're looking for.
I think of two things based on what you just said. One is tech debt velocity
(38:12):
is at an all-time high around this like nobody else ever in the past because
not only are we building things that within a few months become obsolete.
That is at a higher level than we've ever seen before, like ever.
And this is not just in AI specifically, I would say, Gen AI specifically.
And then number two is really making that call of what do I wait?
(38:37):
What is the right time to wait on this? Because it is moving so quickly.
You really have to think about that and have that optionality that we always used to push for.
You have to have that optionality as you build and model these things out because
we're moving so quickly. So minimize the tech debt, create it in a modular fashion,
think about the optionality around this.
Otherwise, you're going to be paying for it three times over. A hundred percent.
(39:01):
And seeing that, because I look at client roadmaps, right?
And you talk about the sequencing of certain things and you talk about which
ones that they're going off and building.
And it's like, did you know that there's this technology is coming out in the next six months?
You may want to think about reprioritization of your roadmap.
And so I think this continual checkup that you do around your data and your
(39:24):
AI strategy is going to be incredibly important.
It's not, hey, let's build a three-year strategy and roadmap.
Set it and forget it and stay on that journey for the next three years,
and then we'll revisit it.
It's like, no, you should probably be revisiting this quarterly.
Because every quarter, you may be in a situation where you might want to reprioritize
certain things based on what's happened in the last 90 days,
(39:47):
because it is moving that fast. Yeah.
I mean, who would have thought we would be talking about revisiting roadmaps every 90 days?
We create them so we can follow them for the period that we created them for,
which is typically two to three years.
So to say, let's revisit in 90 days, I don't know, it's a little different than
(40:07):
what we're used to. Well, I don't think it's as much about revisiting the business objectives.
I mean, you're still moving to that North Star, right?
It's more of just, are we on the right track? Are we incurring debt?
I mean, if you talk about Agile and the whole point of why Agile sort of took
(40:28):
over Waterfall, and when you talk about technical delivery, it's because you
can get ahead of the issues much quicker, right?
Right. You don't want to wait till you're in UAT to realize we completely missed
the mark for the business.
And so the last year we've spent on this project, we're now behind the eight
ball. Nobody wants to get to that point. Right.
(40:48):
And so to me, it's no different. Right. It's almost like Agile where it's like,
hey, we want to do retrospectives.
We want to understand what's working, what's not working and reassess.
You know, so it's not about like constant replanning. It's just more.
Are we on the right track?
Are we being sensible? And so it's just having that checkup and making sure
that, yep, we feel good. You know, nope, we don't feel good.
(41:09):
Here's why. Okay, let's reassess.
But the business objectives, you're still moving towards that.
As I said, data work is hard. Data work is going to continue to be hard, but in different ways.
Yeah. Yeah. I mean, there's no easy solution to saying, I want my data perfect, right?
Yeah, exactly. I want it 100% accurate. I want it perfect.
The technology advancements are there, but it's a discipline.
(41:33):
I mean, we all know that. That's why we say, do the hard work on governance.
It's because it's not easy, right? And if you really want those things,
that requires human beings to be accountable.
And that's what it takes. And that hasn't changed in however many years.
AI will help it. It's not going to change it. And in some ways,
it will create more challenges, which we've already seen. But it's exciting times.
(41:57):
Snowflake aside, just that tagline of being able to see clients evolve with
data, with the right tools, with the right capabilities, with the right strategies.
It's going to be an interesting few years ahead of us. Yeah, it's exciting.
Yeah, totally. I'm personally very, very excited about this.
Mike, thank you so much for taking the time to chat about Snowflake Summit with us.
(42:21):
My pleasure. It's very enlightening. I learned a lot. I hope everybody else
that's listening did as well.
I love that I just get to get listen to Mike geek out for a little bit.
It's always fun. Yeah, it's nice to take a step back and sort of reassess all
the information overload that you get from talking to partners, talking to clients,
(42:41):
listening to the keynotes, listening to all the different sessions to take a
step back and say, okay, what does this mean in the grand scheme?
I think is important and love the podcast.
Appreciate you having me on and look forward to when I get to come on again
and whatever that topic's going to be.
We'll figure that one out soon. All right. Well, thank you. Thank you, Paul. Appreciate it.
(43:06):
You.
Hi, Sandy here. Welcome to another episode of How I Met Your Data.
This week, Anjali and I have a special edition for you. We're going to debrief
on the Snowflake Data Cloud Summit that took place just a couple of weeks ago.
And in order to do that effectively, we actually asked one of my colleagues,
Michael Cochran, to join us.
Mike is currently Cervelo's Global Data and Analytics Consulting Practice Leader.
(00:22):
He is someone I have worked on and off with for over 22 years of my career,
and someone I deeply admire.
So I met Mike in 2002 at a company called Painted Word, where we both started
off there as front-end developers.
Our roads absolutely diverged. Mike went from front-end development to wanting
(00:43):
to learn a little bit about the back end.
So he dove into data warehousing and data warehousing concepts.
I remember just seeing his nose deep deep into books at his cubicle for days
on end, if not months on end, as he got all his certifications back in the day.
He also became a software developer.
He used to build most of our custom solutions, our web custom solutions as well
(01:08):
during the later part of our career there.
He moved on and became the CIO for the Palladium Group, where he learned all
the challenges CIOs his face by living in that role day to day.
He did that for a number of years.
He then joined Cervelo as the first employee and non-founding member of the
firm, where he was quickly asked to think through the cloud strategy and look
(01:32):
at all the technologies that were coming to fruition about 14 years ago when we were founded.
So he did that and quickly culminated into him becoming the Global Data and
Analytics Consulting Practice Lead,
which he co-leads today with another colleague of ours, Glenn Heatley.
So now that you have that background on Mike, we can go ahead and dive into the conversation.
So with that said, let's dive in.
(01:52):
Music.
Hey, Mike. Hey, Mike. Hi. Welcome. How are you? Wait, this is recording already? already.
Yes, we're recording already. You got to give me a heads up.
(02:18):
There's no video. Just walking right into the show.
We like to keep it spontaneous.
Welcome, Mike. Thanks for joining us. I know you and Sandy just recently got
back and have recovered from your time at the Snowflake Summit.
Cervelo has been a Snowflake partner since the very beginning, I think.
(02:39):
2016, I think. They went GA in 2014.
We were about a year into that. So it was probably late sometime in 2015.
And then really pushed it in 2016. Great. Cervelo has been a partner for a number of years.
And I think there's been a lot of changes that we've experienced in the platform
(03:00):
in the way that our clients are really deriving value.
So I'd love to hear from you. What have you seen as that snowflake evolution
that's occurred and how has it been impacting our clients?
Yeah, it's a great question. I think it's been a pretty interesting journey to date.
When we were first introduced to Snowflake, I think my impression was that here
(03:20):
was a company that finally addressed how to build a data and analytics solution
in the cloud that is optimized for the cloud.
I think up until that point, which Cervelo was founded in 2009,
we really dove into our cloud strategy in 2010.
And what you saw was a number of different companies that were taking legacy
(03:42):
technology, and then just kind of putting it in the cloud and then calling it
cloud software or cloud platform.
And it really wasn't optimized for that. You ran into all kinds of performance
issues, you ran into scalability, you ran into portability challenges,
it's just the list sort of went on, right?
So you always kind of ran into a brick wall with a lot of these different companies.
And when we first saw the demo of Snowflake, we were blown away.
(04:05):
It was all right, this is going to address this problem, this is going to help
this client, the number of things that we saw value in that could really contribute
to our client success was pretty amazing.
And, you know, I remember I was being in a QBR in New York with them early on in our relationship.
And somebody asked me the question, hey, what are your thoughts?
(04:25):
What should we be thinking about?
I said, listen, you know, your platform and your technology is great,
but you've got a limited window where people are going to start to catch up.
And therefore, it's like, what's the next thing? And what's the next thing?
So as I think about where they've really come over over the last however many
years, think they've sort of gone down that strategy path, right?
They started with, all right, here's a technology that's built and born in the
(04:50):
cloud, and then really realized that, okay, our clients have other needs and
use cases that we need to close the gap on.
And so, you know, seeing them add things like multi-cloud support.
So getting beyond just AWS, opening up for Azure and Google, And now more recently,
being able to combine all of those capabilities in more of a hybrid structure
(05:13):
is pretty impactful for many enterprise clients that have a multi-cloud or hybrid strategy.
And so it's been great to see that evolve over time.
The idea of clients having the demands of we need data in real time, right?
You always had to dig into that need or that requirement because it wasn't always
(05:34):
real time. It was, all right, I need it just in time or at the right time.
But now the way that technology is advanced, real time is a reality and it's
really table stakes now.
Like if you're not delivering things in real time or near real time,
you're missing out on many opportunities.
So their ability to start to close that gap as well has been very interesting.
(05:58):
And, you know, I think more recently coming out of the conference.
I always tie it back to what's happening in the market and where people demanding
probably falls in a few different areas.
But I'd say the top ones are customers or consumers having more control over their data.
So privacy is pretty key. And so when you think about what they're doing to
(06:19):
build in that compliance, to build in that governance, data clean rooms is an example.
And being able to share data without having to move all kinds of private consumer
or customer data is pretty important.
And so seeing them put emphasis behind that when we know it's a big issue for
our clients and something that they're struggling with and they need help with
(06:42):
is extremely important.
Cost management, cost optimization. So FinOps, seeing them make strides there
as well, been pretty interesting to see.
And then just the complexities of now there's a big focus in AI and you have
all these different technologies that you have to make interoperable,
them going beyond just data and bringing compute in the NVIDIA relationship
(07:06):
and bringing that to the data versus the data going to the compute and having
to move all of that around and then building capabilities around apps and.
So when I talk about complexity, all of these things together add complexity
to our clients' data architectures.
And they've really been able to bring all of that together with applications,
(07:31):
with the governance, with data, with high performing compute that AI applications will need.
It's just been a pretty remarkable journey seeing all of that come together.
And then, Sandy, I know you and I were kind of joking about this,
but the very first conference that we were at, the entire room was probably
(07:51):
smaller than the partner keynote at the most recent conference.
So, I mean, that just speaks volumes to how big they've gotten in the ecosystem
and adoption that's come along with that.
Yeah, it's funny because when we first, I still remember telling you,
hey, we need to look at this.
And if it's as good as what they told me, you need to be all over this.
(08:12):
Because I had, I remember talking to the rep and they were telling me what it could do.
And I had a flashback of like the last nine years of my career and then heavy
nights and weekends talking to architects, DBAs, trying to get them to do something that was basically.
Impossible and impossible feats of performance that we were asking of them over
the last nine years prior to Snowflake.
(08:33):
And then seeing that, I mean, I still remember you and I were like, what?
Like, what is going on here? But if you fast forward from that point forward,
I feel that they did have a large
window of being the only one in the marketplace that really could do.
And they actually said it in the conference. They said, hey,
look, we don't just take storage from compute.
We're separating compute from compute. Nobody has done that.
(08:55):
I don't think anybody still hasn't done that holistically the way they have.
So they did have a big time in there in the marketplace where they were the only one.
But now things are changing with data science and AI and ML and all these other
capabilities that people are trying to build.
So they had to accelerate. And I feel that.
The change we see in them today only happened over the last three years.
(09:16):
They've been around for eight, nine, right?
Five years of that, it was just growing a data capability, very traditional
in the cloud, making it work, making it very different than what people had done in the past.
But now I think they're really accelerating that and thinking beyond that because that has been solved.
And I think that that's what's super exciting about this is being able to see
(09:38):
them take this to a completely different place and truly make it a platform
where anything you have to do on data, you can do on there.
Yeah. Yeah. I think it's like a really exciting time to see what is coming out from Snowflake.
And so that kind of brings me to my next question is you've been back for a
week now from having been at the summit, you've seen and reflected on some of
(10:01):
the advancements and releases that were showcased at the summit.
So what were those things that they were sharing with us? And what were your
takeaways from those those demos?
Yeah, the you know, just like any big conference, right? There's there's always the the big reveal.
So I'd say some of the things I'm most excited about probably falls in a couple camps.
(10:23):
One is probably the less exciting stuff, you know, to me, protecting our clients
data to making sure that they're governing it properly.
If you think about the ability to productionalize AI, which which was a big, big topic.
Companies really focusing on experimentation and trying to figure out the use cases last year.
(10:44):
And then this is the year of productionalization.
I agree and disagree with that. I think many companies are trying to productionalize
and get the real value out of AI.
But at the end of the day, AI is only as good as the data that you feed it.
So it's this whole thing of garbage in, garbage out that we've been talking about for decades.
It's the same thing with AI. AI doesn't change the fundamentals and the block
(11:07):
and tackling that needs to happen with your data.
So I'm happy that they're putting emphasis around the governance and the transparency
and the compliance aspects of it, because that's going to be very critical as we move forward.
And as customers and consumers get more control over their data,
as I had mentioned previously, like that is key.
And you have to be in a position to be able to react to those realities.
(11:31):
On the flip side of that, you know, more of the innovation in the AI side,
the fact that they're pushing on the concept of native apps and containerized services,
and then bringing a lot of the machine learning and LLM capabilities to be able
to build these chatbots and other customer or client facing solutions.
(11:51):
I think it's of high value. I think it's still early days in terms of where
this will go, but I definitely see them on a great track and a great trajectory.
Our clients, when we speak to them.
It's all about speed to market, right? It's speed to market to differentiate.
It's speed to market to stay relevant. It's speed to market to be able to continue
(12:13):
to be productive and efficient in the work that you do.
And so that requires acceleration. That requires applications.
It doesn't just mean, hey, I need data and then I can do anything with data.
It's like, well, I need data. I need data for supply chain. I need data for
finance. I need data for a number of different use cases.
(12:34):
And so to see that you can now,
all of this kind of coalescing and the lines being blurred of applications versus
data and compute versus storage and those things are very important and very
relevant in the modern world.
So I'd say those are some of the big takeaways for me.
One thing I'll be critical of is I was looking for maybe a little bit more of
(12:56):
a bigger reveal on some of the things. So the things that they've had in private
preview, I was hoping would become publicly available.
A couple of them are pretty close, but not quite there yet. And I know those
have real applicability to some of the things that we're doing with our clients today.
So I'm really looking forward to when those do become publicly available and
we can productionalize those for our clients.
(13:19):
I totally agree with you, Mike. I think one trend, though, I've seen with our
larger clients is creating kind of a bit of a commodity around data engineering.
Larger organizations start to look at it as a commodity and less of a thing
that they have to have really skilled workers in, which is a bit of a challenge.
So the thing I was excited about with Snowflake was that they're doing things
(13:42):
like there's a co-pilot, but now that co-pilot is in kind of the SQL window itself.
So you don't have to go somewhere else to get, is this right?
You can actually finish your code right in there.
So the AI is helping you. So a lot of those developer tools,
they announced a ton of DevOps stuff like monitoring, Git integration.
Little things like that, that they've decided to shovel into the platform.
(14:03):
Because the bare bones capability of moving data from one place to another,
making sure it's transformed effectively, ensuring that it's available to the
business, that work still needs to be done.
And the easier Snowflake can make it for your engineer to do the right thing,
the better everybody's going to be. A lot of their competitors aren't spending
time there trying to make it easy for the developer.
So they really are trying to make it easy. And that, to me, left an impression
(14:26):
because I'm not a very technical person.
So if you can explain something to me and show it to me in a way that actually
makes sense and I'm not completely lost in where you are in your windows.
That's actually pretty enticing because that means you can take a junior developer
and throw it at them and they'll be able to navigate that. No,
I tend to agree with you, Sandy.
And it's interesting because this isn't just isolated to Snowflake.
(14:48):
I think what you're seeing is that the things that are maybe you would consider
commodity, right, within that role.
So maybe connecting to different data publishers, right?
So like an ERP system or CRM, you know, anywhere where you need to get data,
essentially, those things have evolved over time with different connectors.
(15:09):
So you don't have to build those, not that you don't have to build them anymore,
but there's a whole ecosystem around that.
The way that you transform and work with data is now evolving.
Some of the things that you had to build in, in terms of validation controls
and those types of things, those are becoming a lot more easier.
So in a lot of ways, the engineering role is becoming more like a data orchestrator
(15:32):
in a way that just an engineer.
So it's like somebody who really needs to understand, well, where does my data originate from?
And then what does it need to look like on the other side?
And you still need to make sure that all that high quality governance and all
those things are built in.
And so I think what we'll see is an evolution of that role.
But companies like Snowflake are really investing in taking away the inefficient
(15:58):
work that has to happen, creating the assemblies, creating connections in the databases.
Repeating the same data quality routines over and over again and reinventing the wheel.
Things that are kind of table stakes, these companies are taking care of.
And therefore, data engineers are going to have to think about,
all right, what does my role start to look like and how can I elevate to that
(16:21):
role, knowing that some of the commodity things will go away over time.
And another part of that is the native app framework that they've created,
because anybody can create an app, if you will, that functions with data.
And then you can leverage it in flight and snowflake. We have a couple of life
science clients and everybody's talking about the standard of fire.
(16:41):
I saw there's a native app that allows you to do that.
And I was just like, wow, you just put like 20 consulting firms out of business
with a native app. Well, and that's right.
I think that and just and then just market awareness. Right.
So now that they have the marketplace and being able to allow partners,
allow customers, allow the entire ecosystem to build these applications that
(17:05):
can address fundamental issues and make those readily available.
Similar to what we saw with digital natives back in the early 2000s and beyond,
open source, a lot of the innovation they had, it's sort of akin to that, right?
Like somebody solves a discrete problem, make it available.
Some of that's going to be paid. Some of it's going to be free.
But at the end of the day, it creates this awareness and this transparency to
(17:28):
allow everyone to evolve and to innovate.
That's exciting to see because at the end of the day, the more eyeballs you
can get on a problem and have diversity of thought and diversity of how to solve
that problem just makes it all the better for all of us. Yeah.
Yeah. It leaves us room to tackle the harder stuff, right? Not the stuff that
we've already solved time and time again.
(17:50):
Exactly. Because at the end of the day, you're going to be solving the specific
thing for the client, not all the stuff that sits behind it that business people
don't typically see, right?
You know, it's that whole iceberg image. I only see the stuff at the top of
the water, but not all the mess that kind of sits below the surface. That hasn't changed.
And so it's like, how quickly can I get to that tip of the iceberg and show
(18:12):
that value to a customer, to a client? That's the important thing.
And so all this other stuff, automate it as best you can.
You can't automate everything, but that's really where the world is going, right?
It's automating a lot of human behaviors that are repeatable.
And I think what we're seeing is your data engineering point,
Sandy, doesn't just stop there.
(18:33):
It goes into analytics and a number of other things.
There's a lot more automation of that human behavior that we're seeing.
And so everyone needs to figure out how do I adapt to that? What's my strategy
to do that? What are the people I'm going to need? How do skill sets evolve?
How do we train people? How do we deal with the change management aspects of it?
(18:53):
There's a lot that goes behind that. So as exciting as it is and the impact
that it's having, you know, you really got to keep your eye on the prize that
it does take a lot of hard work to do it right and do it successfully.
It's just the efforts are being shifted to more high value work.
Than maybe some of the commoditized things that Sandy mentioned.
(19:14):
Yeah, for sure. And do you think our clients are going to understand and realize that benefit sooner?
Well, I think there's certainly a recognition that I don't think anyone is sitting
there saying, oh, this stuff is going to be super easy.
But at the same time, it's pretty clear that, look, all the stuff that you've
had to do for decades around managing data and making sure data is of high quality,
(19:38):
creating transparency, transparency observability all that all that great stuff
that takes work and at the end of the day it's it's
about process it's about diligence it's about
accountability and ownership it's all these things that
we talk about that just doesn't go away and so it is making sure that that's
crystallized and that if you're talking about a ai effort right or a transformation
(20:01):
for that matter all of the stuff you got to do to get your data right doesn't
change you still got to do it now Now, how much of that can be automated?
How much of that can be fast-tracked? How much can that be addressed time to
market? That will certainly improve, but it certainly doesn't go away.
And the effort is still pretty big. Yeah, no, it sure is.
Mike, one of the other things that you had mentioned was the NVIDIA relationship.
(20:25):
So that was one that we heard as an announcement.
Would love to hear your reactions to that, as well as any other partnerships
that had been announced during the summit.
Yeah, I think the NVIDIA one is pretty exciting. They're obviously a juggernaut.
Ever since OpenAI released ChatGPT and the world just ballooned into the AI revolution,
(20:46):
the amount that NVIDIA has been able to do in this time period,
the emphasis on their chipsets to be able to do massive amounts of computing is pretty impressive.
And bringing that to the Snowflake data cloud or AI data cloud,
it's incredibly important because these use cases are only going to continue
(21:09):
to evolve and the amount of horsepower you have to put behind that is only going to evolve.
So I think it's pretty exciting that there's this strategic relationship to
try to make this as efficient.
And cost effective as possible. And so that's sort of what I see that this relationship
being is how do we bring the best solutions to customers and clients while minimizing
(21:35):
a lot of the complexity that goes in into it?
Because, you know, we're seeing our clients go down this path.
And when they start thinking about the cost aspects and the sophistication of
technology, there's still a big learning curve there, right?
So I think as much as that can be sort of packaged, commingled and simplified,
(21:57):
the better. And I think this is just a step in that direction.
I think there's going to be a point of view fight between Databricks and Snowflake
on what's the right approach with managing costs, because Databricks announced
the ability to just forget it.
Like, don't worry about it, we got it for you. And I thought that was pretty interesting.
It's just like, no, you don't have it for me. I want to be able to manage this myself, right?
(22:22):
I have to trust you now that you're using the least compute possible for my workload.
And what if I want it to be faster than what you're providing to me?
It's just so odd. They announced that and I cringed.
I immediately cringed. It reminds me of when we were trying to sell Snowflake
because all the DBAs were holding on to the fact that they could tweak the levers.
(22:43):
And I was like, don't worry about it. It's got it for you. You just have to
worry about the size of workload you need, et cetera.
So it's the same feeling. I had that feeling when Databricks announced it.
And I was like, that's not going to go well. And I'm not even the one building this stuff.
So I don't think people will have a reaction to that.
I do like the idea of not having to worry about it, but there's a trust factor there, right?
(23:03):
You're now, you're asking your customers to trust that you have their best interests
in mind, which is not always the case.
Yeah. And it's very different, right? When you talk about feature and functionality
versus, well, am I getting the value or the ROI in what I'm actually spending on, right?
So when I think about feature and functionality, it's okay if it's very simplified.
(23:26):
There can be that layer of abstraction of an understanding that,
okay, I just need to know it works. I don't necessarily need to know everything.
Science or the physics behind how it works, unless you're nerds like us,
and you really want to dig into the detail.
Well, how does that actually work in reality? But when it comes to cost,
you need a high level of transparency.
Think about Google BigQuery, challenges that our clients had with that in the
(23:50):
early days of, okay, well, we're spending money, but we don't quite know what
we're spending money on.
And so I think that transparency and being able to see which departments,
which divisions, which business units, you know, what are they spending? What are they using?
I think it also having that transparency helps to drive innovation, right?
(24:11):
So if you think about, well, all right, it costs X amount to do this,
and we feel that's too high, make that better, right? On the feature functionality side of things.
Don't try to demystify what we're spending money on because clients and CFOs
and COOs, they want to understand that.
CIOs want to understand that I'm getting the best value for my bucks,
(24:35):
so to speak. And that's important.
On the feature functionality side, I just feel that, hey, if it works and it
works great, you're not going to get asked a lot of questions.
But if you're given a number that has a currency associated with it,
people want to know the details and they should have that level of transparency.
Well, it doesn't just drive innovation. It drives decision-making on the business
(24:58):
as well, because the business could be asking for something that costs a lot of money, right?
And it could be as simple as I want that data load to happen on an hourly basis
or every five minutes, all right?
And that could be taxing because if I need to get it done every five minutes,
maybe I need a larger compute engine to make that happen.
But I'm looking at that going, well, and this has happened to us where we're
(25:19):
getting into these conversations with our clients and saying,
well, actually that decision is worth $10,000 a month.
Is it worth to you $10,000 a month to have it seven times a day?
And the answer was actually, oh God, no, it's not worth it. Right?
So you can have those really pointed conversations about, I want to deliver
X, it's going to cost me Y.
Is it worth the, not just the effort to make that happen, but actually the ongoing
(25:43):
costs to enable it? Yeah, exactly.
Because otherwise it becomes one size fits all.
So if you need to determine, is this delivering on the business value?
If you have a one size fits all model, you may not be able to quantify that answer, right?
But if you can break it down to your point, if you just took a very basic example
(26:03):
and say, well, okay, if I don't have to run it as many times during the day,
or I can scale back in terms of how long it runs,
that's the optionality that people want, because they may be able to say,
well, we need to bring the cost down in order to deliver on the ROI.
Well, great, I have the levers to be able to do that.
But if I don't, then it makes a much harder choice.
(26:26):
So in some ways, you feel like you're simplifying things for people, but at the same token,
you're creating challenges. Because at the end of the day, when you talk about
investments and investments in technology, they're always going to be scrutinized.
And they always have to have detailed backing behind them.
So to me, it's almost like you want to create more transparency, not less transparency.
(26:49):
For sure. For sure. And I'm really glad you brought up Databricks.
Because if I think back to 2016, 2017, we would have these singular conversations about Snowflake.
It was just Snowflake. And then, you know, you fast forward a couple of years
and now you're either having a conversation about Snowflake or about Databricks,
but never really a joint conversation.
(27:11):
But I think we've seen a lot of clients starting to move to a model where they
have both Snowflake and Databricks in their environment.
So, you know, any thoughts on Snowflake and Databricks better together?
Yeah, I think what's and this was one of the things that that was covered in
the in the Snowflake conference, and I think is, is very important.
(27:31):
And I'd say it's a very big takeaway, I think, for for a lot of companies is
that you're always going to want to have optionality.
And there's no one technology that's going to be able to solve every problem for you. Right.
And so bring those two things together.
And what do you have you, you need in an architecture that is not honed in,
(27:53):
on any one technology, right?
And so when you think about the Databricks or Snowflake, to me,
it's being agnostic and taking a step back and saying, well,
what are my objectives? What's my strategy?
What business outcomes am I trying to drive towards? And then what does my technology
ecosystem and architecture need to look like to support.
(28:15):
And, you know, one of the things that I thought was interesting,
and one of the partner sessions, Sandy, you and I sat in on is that Snowflake,
in many ways, is beating that drum of making sure you have an agnostic architecture.
And so that's not only inclusive of technology, but also the patterns that you
deploy as part of that, right?
Whether it's data fabric, data mesh, data lake house, data, data lake,
(28:38):
I mean, the list goes on, right? And there's different reasons why you may choose
one approach over another.
There's different reasons why you may choose one technology over another.
But at the end of the day, it comes down to, it's probably going to be a little
bit of a few things that you're going to need to make sure you have modularity
built into your overall architecture.
(28:59):
And that's going to require multiple technologies at the end of the day.
And so really what you have to look at is what is the business strategy of these
companies And where is their secret sauce and what are they well known for?
Right. And when they start getting out on the fringes of stuff that you sit
there and say, OK, this is this might be a stretch or this might be an area
of growth that maybe doesn't make a lot of sense.
(29:22):
Well, then you may want to stick with best of breed in that case.
So, you know, I think it's important of balancing those things,
of continue to be agnostic and knowing that there needs to be an overall ecosystem
of technologies that have to, you know, have interplay with each other.
And to me, it's not just going to be Snowflake or Databricks, one or the other, both.
(29:44):
You know, I think it really depends on the situation and the client and their
investments and their strategy, all the things that I mentioned.
And I think we'll see that evolve over time. It was going to be acquisitions
in the market and a lot of these AI companies are going to become part of bigger companies.
So I'd say in the next five years, it'll be interesting to see how it all plays out.
(30:05):
At the end of the day, it's what is your product strategy? Where are you taking this?
And I do think Snowflake's on the right track. They're not trying to break out
into areas that just are far removed from what they do.
They're creating that ecosystem around them to be able to do that,
and they're staying true to the core.
And that's important, I think, because once you start deviating from your core
(30:27):
strategy, that's when things get messy.
And when you start seeing lots Lots of acquisitions and then stuff starts to fall apart.
It wasn't as great as it used to be. And that's just, it's sort of the evolution of growth in many ways.
But it'll be interesting to see over the next five years. Absolutely.
I think what's been interesting is the whole true decoupling of data from the compute, right?
(30:48):
Because now with Iceberg, because Databricks bought Tabular,
Tabular allows them to use anything, including Iceberg and other formats. mats.
And Snowflake is making this huge iceberg push to be able to work on data sitting on iceberg.
So I think it's just going to be a matter of a rush of who's actually able to
work on data and compute with data faster,
(31:11):
and who has the most tools and the tool belt to enable you to do that.
Is it going to be Databricks? Is it going to be Snowflake?
It could be both. Your data could be in one place and both these platforms can now work on it.
So that's like a completely different thing that I can literally put it in this like other thing.
And then both these platforms can actually do workloads on that data set.
(31:33):
And not move it. That changes the game. And even Salesforce got in on it, right?
So Salesforce now can load data in and out of Iceberg as well,
or read data in and out of Iceberg as well.
So this whole interoperability of where your data could be sitting in this one
thing and all these other applications and platforms can work on it, touch it, read it.
I think that's a whole new world.
(31:54):
That's a great point. Totally agree. So with all that being said,
were there any Any even better if announcements that you were looking for?
Maybe a little bit more information, a little bit more clarity?
Yeah, I kind of come back to my earlier point. I think some of the things that
wanting to see come out of private preview be generally available would probably
want harder timeframes of when that's going to happen. So I'd say that was one of the takeaways.
(32:18):
I think the other thing is you always want to see customer cases and customer stories, right? Right.
For me, going to these conferences, it's less about understanding the nuts and
bolts of the technology, but more what's the applicability of that technology
to a business use case. Right.
How is this actually going to help move the needle for for one of my clients?
(32:39):
But definitely want to see where people are being successful,
but then also being transparent and saying, hey, we're not being so successful here. and here's why.
And these are some of the things that we've tried. Because again,
more information and more eyeballs and more voices in this just makes it all
the better for everybody else.
Were there any use cases that maybe are different than what we've seen with
(33:03):
our clients and maybe something that we should be considering?
I'd say just the contrary.
I mean, it was good to validate that a lot of the things that our clients are
struggling with and the solutions that we're bringing to bear for them.
You were seeing other customers of Snowflake in the same camp and trying to
address maybe in different ways or nuanced ways.
(33:24):
But at the end of the day, I think the same client problems and themes are consistent.
And so I think in many ways, it was just a validation of a lot of the great
work that our teams are doing to support our clients.
Good. So our clients' challenges are not each special snowflakes.
That was such a terrible pun.
I mean, I didn't see anything crazy and outlandish. I actually agree with Mike.
(33:48):
I think what I saw, and most of it was from Snowflake, like the art of the possible
with these challenges that we've seen clients have time and time again.
Art of the possible with all the new capabilities that they're bringing to there.
That was what I was noticing was, okay, there's a different way to solve this,
that they're pitching to their partners and customers, because obviously no
one's doing it quite yet. They haven't released a lot of this stuff.
(34:10):
So it was very much pie in the sky conversations about where it was headed.
But I felt there was a shift in terms of, I don't have to ask the question anymore
of, can I do something? Like, I think you can do a lot now.
I think the question now is what should I be doing? And that's something that
we do very uniquely with our clients.
I think a lot of people wait to hear from their clients in terms of what problems
(34:33):
they're trying to solve.
And we kind of are front footed in terms of helping our clients figure out what
are the right problems to solve And then how to solve them becomes secondary.
So that, to me, was exciting because it allows us to have more freedom in terms
of the advice that we're giving and the propositions that we're discussing with
them. Yeah, I think that's a great point. I mean, I see a couple things.
(34:55):
So one is, I think the art of the possible, I think there's a recognition on,
yes, a lot of these things are real and quite possible.
I do agree that there's a bit of what are the use cases we should be going after
and where can we get the most pull through.
I think there's going to be continued on the art of the possible side of it,
(35:16):
of just understanding and learning of the applicability of AI and gen AI.
I mean, it's not like these things are new, right? They're just,
they've evolved very quickly.
And that's what's opened up the aperture, I think, for many companies.
But this has been around for decades, right? Right.
So it's just now it's just supercharged and it's it's a lot more realistic for
(35:38):
companies to be able to do this at scale.
But I think there's a big learning curve that comes with that in a number of
different areas. Right. We talk about responsible AI.
We talk about high quality data. We talk about hallucination.
We talk about all these different things. Well, people have to learn what does that all mean?
And how do I make sure that I'm addressing all of those things on equal footing, right?
(35:58):
That I'm not letting one thing to chance that could really have,
you know, bad implications on your company.
And so there's, there's going to be this continued learning that that goes along
with that being true to understanding where you are in that journey, right?
And what you're capable of doing, you know, short term and long term,
right? And how you build up to that.
(36:18):
And what are the use cases we should go after? or will they create the value that they're meant to?
You know, you sort of have to get past this building proof of concepts and prototypes.
And so I think you have to be realistic about what companies can do and what
they can evolve to over time.
What I'm seeing is that a lot of clients, because a lot of these capabilities
are now sort of real and evident and right in front of them,
(36:42):
you know, they're off building a lot of different things.
And so it's almost like you got to take a step back and say,
all right, you need to see the forest or the trees.
Are these investments you're making and some of these short-term things,
because you can do it, are they the right investments to be making right now?
Because what you may build today, where there may not be a solution,
just a lot of capability to do it, well, those things could just become available
(37:04):
in the next year or two, right?
So you have to balance the investments that companies are making and building
things bespoke versus things that are going to start being built into different
software, different platforms.
Different maybe open source capabilities, you really have to look at that and
have a view of the future.
Because case in point, a client came up to us recently, and they said,
(37:27):
hey, we want to take this beyond POC, we want to build this.
And I said, well, maybe one of the things we should first do is is just do a
quick scan of well, what's happened in the last six months since you started
on this journey, because there is there's a lot more advancement that's happened.
And as a result of that, there's sort of a recognition, okay,
it probably doesn't make sense for us to build this at this time.
(37:50):
We can build a portion of it, but we know that there's some other things coming soon.
And so let's treat it like an evolution, right? Let's build the things that
are going to differentiate, but let's not build things that aren't going to
give us the true value proposition that we're looking for.
I think of two things based on what you just said. One is tech debt velocity
(38:12):
is at an all-time high around this like nobody else ever in the past because
not only are we building things that within a few months become obsolete.
That is at a higher level than we've ever seen before, like ever.
And this is not just in AI specifically, I would say, Gen AI specifically.
And then number two is really making that call of what do I wait?
(38:37):
What is the right time to wait on this? Because it is moving so quickly.
You really have to think about that and have that optionality that we always used to push for.
You have to have that optionality as you build and model these things out because
we're moving so quickly. So minimize the tech debt, create it in a modular fashion,
think about the optionality around this.
Otherwise, you're going to be paying for it three times over. A hundred percent.
(39:01):
And seeing that, because I look at client roadmaps, right?
And you talk about the sequencing of certain things and you talk about which
ones that they're going off and building.
And it's like, did you know that there's this technology is coming out in the next six months?
You may want to think about reprioritization of your roadmap.
And so I think this continual checkup that you do around your data and your
(39:24):
AI strategy is going to be incredibly important.
It's not, hey, let's build a three-year strategy and roadmap.
Set it and forget it and stay on that journey for the next three years,
and then we'll revisit it.
It's like, no, you should probably be revisiting this quarterly.
Because every quarter, you may be in a situation where you might want to reprioritize
certain things based on what's happened in the last 90 days,
(39:47):
because it is moving that fast. Yeah.
I mean, who would have thought we would be talking about revisiting roadmaps every 90 days?
We create them so we can follow them for the period that we created them for,
which is typically two to three years.
So to say, let's revisit in 90 days, I don't know, it's a little different than
(40:07):
what we're used to. Well, I don't think it's as much about revisiting the business objectives.
I mean, you're still moving to that North Star, right?
It's more of just, are we on the right track? Are we incurring debt?
I mean, if you talk about Agile and the whole point of why Agile sort of took
(40:28):
over Waterfall, and when you talk about technical delivery, it's because you
can get ahead of the issues much quicker, right?
Right. You don't want to wait till you're in UAT to realize we completely missed
the mark for the business.
And so the last year we've spent on this project, we're now behind the eight
ball. Nobody wants to get to that point. Right.
(40:48):
And so to me, it's no different. Right. It's almost like Agile where it's like,
hey, we want to do retrospectives.
We want to understand what's working, what's not working and reassess.
You know, so it's not about like constant replanning. It's just more.
Are we on the right track?
Are we being sensible? And so it's just having that checkup and making sure
that, yep, we feel good. You know, nope, we don't feel good.
(41:09):
Here's why. Okay, let's reassess.
But the business objectives, you're still moving towards that.
As I said, data work is hard. Data work is going to continue to be hard, but in different ways.
Yeah. Yeah. I mean, there's no easy solution to saying, I want my data perfect, right?
Yeah, exactly. I want it 100% accurate. I want it perfect.
The technology advancements are there, but it's a discipline.
(41:33):
I mean, we all know that. That's why we say, do the hard work on governance.
It's because it's not easy, right? And if you really want those things,
that requires human beings to be accountable.
And that's what it takes. And that hasn't changed in however many years.
AI will help it. It's not going to change it. And in some ways,
it will create more challenges, which we've already seen. But it's exciting times.
(41:57):
Snowflake aside, just that tagline of being able to see clients evolve with
data, with the right tools, with the right capabilities, with the right strategies.
It's going to be an interesting few years ahead of us. Yeah, it's exciting.
Yeah, totally. I'm personally very, very excited about this.
Mike, thank you so much for taking the time to chat about Snowflake Summit with us.
(42:21):
My pleasure. It's very enlightening. I learned a lot. I hope everybody else
that's listening did as well.
I love that I just get to get listen to Mike geek out for a little bit.
It's always fun. Yeah, it's nice to take a step back and sort of reassess all
the information overload that you get from talking to partners, talking to clients,
(42:41):
listening to the keynotes, listening to all the different sessions to take a
step back and say, okay, what does this mean in the grand scheme?
I think is important and love the podcast.
Appreciate you having me on and look forward to when I get to come on again
and whatever that topic's going to be.
We'll figure that one out soon. All right. Well, thank you. Thank you, Paul. Appreciate it.
(43:06):
You.
Popular Podcasts
On Purpose with Jay Shetty
I’m Jay Shetty host of On Purpose the worlds #1 Mental Health podcast and I’m so grateful you found us. I started this podcast 5 years ago to invite you into conversations and workshops that are designed to help make you happier, healthier and more healed. I believe that when you (yes you) feel seen, heard and understood you’re able to deal with relationship struggles, work challenges and life’s ups and downs with more ease and grace. I interview experts, celebrities, thought leaders and athletes so that we can grow our mindset, build better habits and uncover a side of them we’ve never seen before. New episodes every Monday and Friday. Your support means the world to me and I don’t take it for granted — click the follow button and leave a review to help us spread the love with On Purpose. I can’t wait for you to listen to your first or 500th episode!
Stuff You Should Know
If you've ever wanted to know about champagne, satanism, the Stonewall Uprising, chaos theory, LSD, El Nino, true crime and Rosa Parks, then look no further. Josh and Chuck have you covered.
Dateline NBC
Current and classic episodes, featuring compelling true-crime mysteries, powerful documentaries and in-depth investigations. Follow now to get the latest episodes of Dateline NBC completely free, or subscribe to Dateline Premium for ad-free listening and exclusive bonus content: DatelinePremium.com