00:00welcome to the a 16z podcasts I'm
00:02Michael Copeland big data is evolving
00:05it's moving from the sole domain of the
00:07high priests of data science to
00:09something that practically every
00:11organization big and small and every
00:13group within that organization can get
00:15its hands on so what happens now the
00:18implications of the democratization of
00:19Big Data are bigger than just big says
00:22Pat Mogae CEO and founder of cos ena
00:25implication of Big Data cloud making it
00:29really democratic is access to everyone
00:32flattening of the organization
00:34collaborative culture and ultimately
00:38faster decision making and it's not just
00:41the corporate world that will benefit he
00:43adds having access to big data tools
00:45will change how all kinds of
00:47organizations including government
00:49agencies and other social services
00:51operate and solve their particular
00:53problems you know like in New York City
00:56I think this was New Year's Eve and the
00:59collector statistics on random gunfire
01:02and they use that data to predict where
01:06they should deploy police so that they
01:09could ensure safety right and what they
01:12realize is that they drastically cut
01:14down on random gunfire but they needed
01:17fewer police actually to be deployed so
01:20you can actually cut costs increase
01:23safety all because you're smart about
01:25you know collecting data
01:28joining Pratt on the pod to describe
01:30this changing world of big data in the
01:32cloud is also a 16 ZZZ Peter Levine if
01:35every company and every organization is
01:38becoming a did a driven company he says
01:39it makes sense to start to put that data
01:42to work this not only democratizes the
01:47it democratizes the organizations that
01:50use big data such that it's not limited
01:52to the only the Fortune 2000 who have
01:55the capabilities to set up these large
01:57data centers Frattin Mogae welcome
02:00all the way from the Greater Boston area
02:03that's right and Peter Levine welcome as
02:05always thank you good to be know Pat's
02:07we're gonna talk about why big data for
02:10starters hasn't made it to the cloud and
02:12I know cos inna Pratt you're the CEO and
02:15Founder is all about moving big data to
02:18the cloud and offering it as a service
02:19but let's back up why has it been so
02:22hard for big data to move to the cloud
02:26it's a great question I think if you
02:29look at a typical enterprise and in the
02:32last 15 years if you look at all the
02:34other application stacks that have moved
02:37to the cloud CRM ERP there are all
02:42siloed applications if you look at big
02:45data big data is not a siloed
02:47application it gets infused through the
02:49organization so it's usually part of
02:52some operational business process and
02:55it's really hard to lift and shift it
02:57into the cloud it's got tentacles all
03:00over got to figure out how to move the
03:02data you got to figure out security a
03:04lot of that data could be customer data
03:08employee data and all that patient data
03:10so security complexity real challenges
03:14that's holding big data back and right
03:17now it's it's on-premise I mean Peter
03:19when you look at the big data landscape
03:21how do you sort of see what's happening
03:24with cozinha and and others for that
03:26matter versus how big data has kind of
03:27been rolled out over the last couple of
03:29years I sort of characterize Big Data
03:33one detto as being an on-prem offering
03:36it's been great but again that requires
03:38a deep knowledge of infrastructure
03:41buying out computing all of the you know
03:43it's basically an on-prem data center to
03:47run big data applications and now what
03:50we're seeing is just like we saw
03:52applications move from on-prem to a SAS
03:54offering we are thankfully seeing big
03:58data enters a big a big data 2.0 era
04:02which is moving big data from on-prem
04:05into the cloud and there's just
04:06tremendous benefit by having
04:10Big Data 2.0 in the cloud as a post on
04:13from the same advantages that we see
04:16with SAS applications are the same
04:19similar advantages that we now have with
04:21with Big Data 2.0 so I want to get into
04:24that because I think for a lot of people
04:26just by itself as a bit of a vague term
04:28but what are those like why do it why is
04:31it so important why do I as an
04:33enterprise or you know an organ taejun C
04:37for that matter why do I care about big
04:39data so deeply now yeah so just to add
04:43to Peter's comments like when he talks
04:45about 2.0 vs. 1.0 I just want to make
04:48clear that the 1.0 which is where most
04:50of the world is it's a 10 billion dollar
04:52market largely going to a few incumbents
04:56right the old stack largely relational
04:59stack the dupe is a decimal point sparks
05:02just coming along and if you look at it
05:05and if you go to a marketing guy in one
05:07of these companies if you go to the guy
05:09who runs risk in some of these companies
05:11if you go to somebody who's running
05:14intelligence like security and you say
05:16do you understand what's going on in
05:19your company like do you understand what
05:22your customers are doing what the
05:24terrorists are doing all that you sort
05:26of realized that many of these guys will
05:28basically say I don't have access to
05:30data if somebody else has that data
05:32they're going to generate a report get
05:35it to me six months later and that's how
05:37I'm going to make that decision so I
05:39think the the biggest issue I hear from
05:42many of these companies and it's it's a
05:44horizontal problem it's not just a
05:45vertical problem and and size doesn't
05:48large or small mm-hmm people just don't
05:51have access to data they can't make
05:52decisions fast enough usually reacting
05:55to something after it has happened and
05:58this idea that they can rack and stack
06:00gear and make these projects happen you
06:04know the cycle time is way too long six
06:07to nine months to deploy millions of
06:10dollars of spend every year and the data
06:13is growing 200% every year so that was
06:15my that was my question like big or
06:17small you say it doesn't matter but we
06:18know that Google for example is a data
06:20driven company but how many Google's are
06:23there out there one but but is every
06:26company now a data-driven company and
06:28I'm just wondering why do I care about
06:30big data if I'm you know a consumer
06:32packaged goods company or if I'm I don't
06:35know selling shoes why do I care you
06:38care because I think you know it's all
06:41it's the world of hyper targeting the
06:42the days of you know mass direct
06:45marketing are over where you could do a
06:47big yard for everyone now it's all about
06:50hey Peter I want to get after Peter who
06:53is Peter what's that demographic what's
06:56that profile and that's different from
06:58I'm guessing and look at our yeah
07:04anything but yeah and so it's on now I
07:10think social mobile it's all about you
07:12know who's my customer who's my prospect
07:14what are they like what are they like
07:16what are they saying about my product
07:18like my CPG product and how do I get to
07:22them I think all of that starts with
07:23getting the data fast that data is being
07:26generated in the cloud and so you want
07:28to be close to that and then you want to
07:30be making fast decisions so you can't
07:32wait for six to nine months right to
07:34understand your customer every
07:35organization is becoming a data-driven
07:38organization and you know the whole
07:42advent and and use of mobile devices
07:45we're all holding supercomputers in our
07:47hands and you know kind of targeting
07:50information and using data imagine if
07:53Big Data were available on every device
07:55that we hold access to and I think
07:59that's a very compelling a very
08:02compelling use case that doesn't
08:04necessarily exist if I have silos of
08:07on-premise kind of data centers the
08:10whole beauty of cloud computing and SAS
08:13kind of applications is that the
08:18information and the the the data is
08:20available to a wide variety of users so
08:24if I make a if I am in a company and I
08:28have my supercomputer in my hand I now
08:31can go ask questions to my data set that
08:35exists in the cloud and get answers back
08:37in the same way that salesforce.com
08:40allows the sales organization to go get
08:42information about the sales organization
08:44and we just haven't seen that
08:46transformation yet on the big data and
08:48analytics side because it has been a
08:52and so you know what we're seeing this
08:542.0 kind of step takes all of the
08:58benefit of on-prem with the beauty of
09:01SAS and cloud and combines those things
09:05together and I think that will create an
09:07agile use of data that we just don't see
09:11right now so it sounds to me like Big
09:13Data is in some sense getting
09:14democratized and if if that kind of
09:17superpower was only available to large
09:20companies that were dated written like
09:21Google and Facebook can you expand the
09:24sort of thinking and the possibility I
09:26mean what happens when like you say
09:29everyone has access to their data set
09:30and that sounds somewhat sterile but
09:32that could be anything right that could
09:34be customers it could be people it could
09:37be outcomes yes also what happens it's
09:40very interesting I think like Peter
09:42mentioned its agility first that happens
09:44immediately like you don't wait to make
09:47decisions yes and and by the way even if
09:52you got it wrong you fail fast right and
09:54then you do iterate right that's the
09:56first immediate impact the second impact
09:58is cost there's a definitely impact on
10:00cost because you're no longer building
10:02you know big iron which is not amortized
10:06across any one it's just for you
10:08the third long-term impact I think is
10:11culture I recently visited an e tailor
10:15that's competing with Amazon right and
10:18and so the contrast in this e tailor
10:21compared to like a typical retailer in a
10:24typical retailer you would have the
10:25supply chain guys the merchandising guys
10:27all these guys are siloed they don't
10:28talk to each other here this e tailor I
10:31walk into the CEOs office
10:33he's got Tablo on his desktop hooked
10:36into the whole whole company and he all
10:40that the thousand people in this company
10:41had the same tableau of thing so guess
10:43what they all look at the same data yeah
10:45they all can figure out what's my issue
10:47how's that tied to somebody else sitting
10:49next door that's what you want so that I
10:52think the cultural distance between the
10:54supply chain specialists in that company
10:56and the CEO is almost not there mm-hmm
10:58right I think that's the Google I or the
11:00Facebook like culture you want to see
11:03permeated across all
11:05mm enterprises and it's not there to
11:07this I think the long-term implication
11:10of big data cloud making it really
11:13democratic is access to everyone
11:16flattening of the organization
11:18collaborative culture and ultimately
11:22faster decision making so Peter how how
11:25else do you see that then changing sort
11:27of the nature of companies as an
11:29organization and even work how we get it
11:31done once Big Data moves to the cloud it
11:38unlocks a number of possibilities
11:40including the ability to iterate over
11:43datasets that may not be accessible in
11:46silos internally think about Big Data to
11:51Daddo as the sort of an out the analogy
11:54is when Amazon did for compute big data
11:57to do is doing to the databases that
12:00exist on Prem right and so you think
12:03about what an Amazon Cloud is unlocked
12:05in terms of compute one I don't need to
12:08build out any data center I pay as I go
12:11and I can do as many big operations
12:14little operations as I need to with the
12:18Amazon infrastructure so think about now
12:20big data moving into the cloud it's got
12:23a lot of the same characteristics the
12:25agility the benefit of iterating over
12:27data the ability to collect data on a
12:31fungible basis so as data grows I have
12:34more capacity if I need less capacity it
12:37sort of you know grows and shrinks with
12:39with that environment and as a result I
12:43think decisions get way better made by
12:46having this iterative collaborative sort
12:49of process on data that exists not in
12:51silos but in a in a centralized place so
12:55that to me is is the value of having a
12:58man it's the same as what happened with
13:00SAS applications moving through the
13:03cloud as people got a lot more
13:04productive because we all had access to
13:07lots of applications were you know when
13:11things were on Prem a few people had
13:13access to the applications and everyone
13:15else was downstream of that just
13:18you know kind of twiddling their thumbs
13:19waiting for the output to come to them
13:22so the the risk it sounds to me like I'm
13:24not doing this this is your you're still
13:26one of those folks twiddling your thumbs
13:28and everyone else is resting ahead or
13:31sure asking one question yet
13:34you're not staying relevant yeah I you
13:38the cloud model works very well for
13:41exactly this use case right you have a
13:44large amount of information centrally
13:47located with lots of users accessing
13:49that information it is the perfect use
13:52case for cloud computing and so with
13:56this I do believe that organizations
13:58will be much more effective in analyzing
14:01their information through this movement
14:04from Big Data 102 to doe and Pratt you
14:07and I have discussed in the past a
14:08little bit it's not it's not just
14:11Fortune 1000 companies it's kind of any
14:14organization right I mean like like you
14:16were saying Peter most organizations if
14:18not all organizations now throw off just
14:21tons and tons boatloads of data right
14:23and do you see this then kind of being
14:26applied outside the realm of you know
14:28the Fortune 1000 yes other kinds of
14:31great I think it's pervasive and this
14:34could apply to governments this is going
14:38to apply to you know all kinds of social
14:41services you know I was reading recently
14:43about predictive policing so the idea is
14:47not new but just this idea that you know
14:52like in New York City I think this was
14:53New Year's Eve and the collected
14:56statistics on random gunfire and they
14:59use that data to predict where they
15:03should deploy police so that they could
15:06ensure safety right and what they
15:08realized is that they drastically cut
15:10down on random gunfire but they needed
15:13fewer police actually to be deployed so
15:16you can actually cut costs increase
15:19safety all because you're smart about
15:22you know collecting data so I think you
15:25know I think this idea of analytics is
15:28pervasive and I think bringing it to the
15:31the way it helps you is now instead of
15:34having New York as a district think of
15:37itself as a silo you can start to
15:39collect this data and pull this data
15:40across lots of data and the more data
15:43you collect the more accurate you are
15:44what saying is New York different from
15:46Chicago really is that different from
15:48Los Angeles you know and so I think the
15:51power of big data in the cloud is that
15:54you can start to see signals in the
15:56noise a lot faster you can collaborate
15:59and you can spot trends before before
16:03they become incidents right big
16:05incidents let me also add that large
16:09organizations who have the expertise and
16:12budgets to go create on-prem data
16:15centers have a huge competitive
16:18advantage over the hundreds of thousands
16:22of small to mid-sized companies that are
16:24probably using Excel spreadsheets as
16:26their data analytics tool right now so
16:29what this is going to do by moving in
16:31the same way that you know I keep going
16:33back to the SAS application being moved
16:35to the cloud then everyone has access to
16:38these great tools and services so what
16:41happens here is small I believe that big
16:45data to do and the cloud is going to
16:47empower hundreds of thousands of smaller
16:50organizations who don't have any access
16:52to big data analytics because they don't
16:56have the expertise or budgets to go
16:58create a on-prem data center and so this
17:02not only democratizes the use of big
17:05it democratizes the organizations that
17:08use big data such that it's not limited
17:10to the only the Fortune 2000 who have
17:13the capabilities to set up these large
17:15data centers and I think that that is
17:17it's going to be very empowering for all
17:19organizations that's a great point
17:21something to add to that one other
17:24challenge that we've seen keep companies
17:28from jumping to the claw cloud is that
17:31the clouds are really expensive real
17:33estate to go to if you have to keep
17:36hopping back and forth what I mean is
17:40because you you know it's expensive to
17:42move data into the into the cloud if you
17:45compute and we've got to move data back
17:47there's a lot of friction so many large
17:52enterprises look at the cloud and say
17:54they're really excited about it they'll
17:56usually get into AWS they'll start
17:58playing around with redshift they'll
17:59start playing or with loop spark and
18:01then they say wait a minute
18:03for me to take all my you know hundred
18:06million dollar infrastructure and yeah
18:12and to move it I can't do that if it's
18:14just a one-off right I got to be able to
18:17go there land there it's like landing on
18:19the moon right a land on the moon is it
18:21just a one trip how many trips have been
18:24made to the moon five six but if you
18:28really want to go stake it out it's got
18:30to be like a civilization right and for
18:32me to go land there I gotta be able to
18:34do all kinds of processing there and and
18:36so the other challenge in with landing
18:39on let's call it landing on the cloud is
18:42to take the other challenge which is the
18:45stack challenge in big data so the stack
18:47has evolved a lot over the last five to
18:49ten years it was all sequel ten years
18:51back now Hadoop came along five years
18:54back and and has a lot of strength and
18:57sparks now the latest addition to the
19:00stack very promising and so the
19:02enterprises are really struggling to
19:04figure out which technology to use yeah
19:06do I do you guys see winners emerging or
19:08is that even the right question I think
19:11it's the wrong question because I think
19:13it's ultimately about can you get your
19:15work done at a certain price points are
19:19from certain price performance and and
19:21the way we look at it this is every
19:23technology is well-suited for a specific
19:25problem and it's not about one
19:27technology being a silver bullet for
19:29everything the point is though why it
19:32becomes really interesting when it comes
19:33to the cloud is for an enterprise to
19:36you got to solve both those problems in
19:38one shot you got to figure out how to
19:40make the cloud be really connected and
19:42easy secure you also got to solve the
19:45second problem which is you got to solve
19:47that platform problem for them where you
19:49say it's all about the work and
19:50somebody's gonna figure out how to map
19:52that work to the right technology if not
19:55somebody's going to be forced to just
19:57move bi stuff to the cloud
19:59and they got too hot back again or they
20:01gonna do spark jobs and then move back
20:03and you can't do that right and and so
20:05that's the baby see the this big data as
20:07a service as a concept with casinos
20:10hatching and then launching is all about
20:12make cloud your next platform so you can
20:17do all your work there so you're not
20:20necessarily you know signing on for one
20:23particular flavor of the stack you're
20:25just getting work done you could but you
20:27know once you see that succeed
20:29you don't want you're not you shouldn't
20:30be forced to hug back right you can do
20:32more and more you can do a data Lake you
20:35can do a data Mart you can do a sandbox
20:36you can connect Hadoop to spark to
20:38sequel and that is not a technology
20:41question it's a workload question right
20:42so just wanted to add that as another
20:45barrier right if Big Data
20:48so Peter your description of you know
20:50all these tens of thousands hundreds of
20:53thousands of small and medium-sized
20:54businesses if they're all you know Big
20:57Data armed up and you know ready to go
20:59what where do we go next like where's
21:02the next competitive advantage then if
21:04if if we've all got this capability is
21:08there something else that you see out
21:09there that that's just kind of emerging
21:11where do we go from here my belief is
21:16that once we solve the big data to dado
21:18problem we'll have the Big Data three
21:20days no problem but but seriously there
21:22is always the next that are always the
21:24next generation I think that Big Data
21:26evolves too once we have the platform in
21:29the cloud which is sort of the 2.0
21:32aspect that practice brought about we
21:35now can build on top of that
21:37whether it's machine learning and
21:39machine intelligence on top of the base
21:42that's being created as one aspect and
21:44or applications that start to get built
21:48for this big data pool that because the
21:53application itself becomes big data
21:56aware so that way we are building
21:59applications that are inherently tied to
22:01the big Big Data the Big Data
22:04information system such that we're not
22:06going through an analytics process
22:09separate separate from the applet
22:12I believe that application-aware Big
22:15Data is sort of the next driver in this
22:18space and we'll get applet will we will
22:22get an application layer that will
22:23become very intelligent as a result of
22:26this under as in this underlying Big
22:29Data but the first movement has to be
22:32that we enable a very powerful real-time
22:35system that is offered in the clouds
22:38such that we can move to the next you
22:42know kind of the machine learning and
22:43application-aware Big Data as some of
22:46the next pieces so that's that's sort of
22:48my belief on where things where things
22:51had I mean remember we're only in one
22:53dot okay so well yeah I think that that
22:57this this movement of big data into the
22:59into the cloud with the right platform
23:02and the right tools is you know the next
23:05several years of sort of this big data
23:08transformation and then we start
23:10layering on these and then there's this
23:12foundational layer that have a run to
23:15interesting okay so you've convinced me
23:19that I'm headed towards the big data
23:21world but let's say I am walking into my
23:25board meeting tomorrow we've done this
23:27move to the cloud and now I'm going to
23:29convince them that our next move is Big
23:33what's my opening argument what's my
23:35opening gambit my opening gambit if I
23:39were walking into that boardroom I mean
23:42I'd be talking to the CEO and basically
23:43be saying look the you know these new
23:46initiatives we are rolling out you know
23:49they are one year two years three years
23:51late name your time period you know our
23:54competition is kicking our ass and if
23:58you really want to change the game or
23:59you want to catch up cloud is the way to
24:03go have you've moved our CRM stock there
24:07we're moving our ERP stack now we got to
24:10put big data on the truck now and you
24:14know and if you do that we can roll
24:16things out much faster and by the way I
24:18can do it for one-fifth the cost I go
24:21into the CFO next who's sitting there
24:23and you can see him nodding right
24:26and then the question is can you get the
24:28security guys you know optimal
24:31comfortable with yeah can you get the
24:33CIO there who's concerned about running
24:35the trains and making sure nothing gets
24:37disrupted and if you do that then you
24:39come out of the board meeting with
24:40actions and a project
24:41Peter anything to add I think that's
24:45spot-on you know it's about I think it's
24:49again it's about agility and an
24:52empowerment that really is the big sell
24:55the cost element of course is very
24:56important I don't have to invest in the
24:59same movement that you know basically
25:01it's the same argument that has already
25:04happened with SAS applications is the
25:05same thing you don't have to build your
25:08whole server data center infrastructure
25:10arguably this is even more expensive
25:12than those sort of on-prem applications
25:14because there's huge storage
25:16requirements and networking and all
25:18kinds of stuff very complex environment
25:19and so to the extent that I don't have
25:22to go build that anymore
25:23I can offload you know kind of the
25:26expense and and this whole notion of
25:28building out a data center I think is a
25:30very compelling argument and provided
25:33that we can of course get over the
25:36security and performance issues and like
25:38Pratt said before the performance issue
25:40is only an issue when we're going back
25:45and forth from the cloud to on-prem so
25:47there is a leap of faith that what we're
25:50going to go and move everything to the
25:53cloud not unlike what we've done with
25:55Salesforce like we don't arbitrate back
25:57and forth you do all your stuff on
25:59Salesforce and you get data back it's
26:01not like the information is is forced to
26:04have to come back to be pre processed on
26:06one side and the other side so those
26:09elements obviously have to be solved and
26:12have to be taken care of but once we're
26:14there there's a lot of very good reason
26:17to go and do this well Big Data 1.1 dot
26:22o is now moving to Big Data 2.0 and
26:24someday we'll get to 3.0 very well lead
26:28us to the promised land
26:29about me but Pratt so Peter thank you
26:32guys so much yeah thanks great yeah