Can DuckDB replace your data stack?

0:08[Music]

0:27You never know about that.

0:35About [Music]

0:44back

0:55to me back to me.

1:02[Music] Hello and welcome to another episode of >> the Super Data Brothers Super Show.

1:09>> That's right. I am Ryan >> and I am Eric >> and we are two real life brothers who work in the data and in data and analytics industry and have we got a good one for you today folks.

1:19>> Boy, do we. And you know, I'm so excited for today's show, Ryan, because people have been hearing about Duck DB. People have been hearing about Mother Duck and how it's gonna it's going to speed up everything. Well, we've got just the person to help talk about that. We have Ryan Boyd, co-founder of Mother Duck joining us today.

1:35>> Absolutely. And he's a super cool dude, by the way. Um, one of the coolest dudes uh I know in data. So, we're going to bring Ryan on in just a second. But first, >> we have to thank today's sponsor, Good Data. That's right. Uh, Good Data is sponsoring season 4 of the Super Data Brothers. And today we want to cordially

1:54invite you to attend an upcoming good data webinar at 9:00 Pacific Standard Time, 6:00 Central European time all about agentic AI in the financial industry. So this is an area where good data has a particular strength and that is helping you come up with kind of AIdriven data applications or or agentic data applications for finance. a really

2:18strong uh focus in that and we're gonna be talking all about it on November 12th. So, I will actually be hosting this webinar. It's another reason for you to attend to make me look good in front of my boss when we get a lot of people in the uh in the uh asking questions and that sort of stuff. So,

2:33check it out. We're going to be talking about use cases, doing a real demo, all sorts of cool stuff. So, uh, join us again November 12th, uh, 9 Pacific, 12 Eastern, 6 Central European time for Aentic AI in the financial industry. I hope to see you there. All right. Um, so, as you know, we shout people out

2:53when they let us know uh, where they're watching from. So, we got Johnny Winter saying, "What the duck?" Johnny's in uh, in the UK. Um, good to This is one of those dudes every year. I see him once a year, Big Day to London. And, uh, it's always good to see him. So, uh, thanks for tuning in today, Johnny. We got

3:08Aaron Wilkerson. I only I'm only here to see Eric. Ryan didn't attend our event last night, and I'm still not over it.

3:14Um, >> well, I didn't I didn't attend, but I I play D and D on Wednesday night, so there's an order of priority here.

3:21>> Yeah. Yeah. Yeah. And then, of course, we've got um Oh, what do you know? It's Ryan Bod saying, "Duck." Yeah, Johnny.

3:27So, why don't we why don't we use that as an opportunity to bring Ryan on to the show. Ryan, good to have you.

3:32>> Hey, it's great to be here. So, uh, why don't you tell everybody a little bit about yourself?

3:39>> Oh, about myself? Wow, that's a big question. Um, so, uh, I, uh, have been working in the data

3:50industry for many years. Uh, we're currently leading, uh, developer relations and marketing at Motherduck as a co-founder. Uh, one of my co-founders, Jordan Tagani, is who pulled me in. Uh, and Jordan and I met when we first worked on BigQuery back in 2012.

4:09Um, so been in data, BigQuery, then um

4:14BigQuery, then Neoforj, then data bricks and and now here. So all sorts of different spots in both OLTP as well as the analytical side of databases.

4:25uh previous to that was a you know developer building applications with Oracle and and things like that. So um yeah been in the data industry a long time and super excited uh to to just make things easier in data but my life isn't all data. I also have a nine-year-old daughter and a wonderful wife and uh I love sailing and plenty of

4:47other things too. So >> awesome. So, we're going to go deep today with Ryan on DuckDB, Mother Duck, how they help for analytics and AI workloads, how they're different from a Snowflake or a Big Query. But first, we have to do our hero of data. So, this week's hero of data is Florence Nightingale. Uh, many people know the

5:10basic story of Florence Nightingale and they know, oh, she was like a nurse or something, right? But actually it goes way deeper than that and there's a big connection to what we do uh in data. So it's imagine it's 1854, Britain's at war in Crimea and thousands of soldiers are dying, but they're not dying from bullets. They're dying from filth. So

5:31into this chaos walks Florence Nightingale. She's a wealthy English woman who defies social class to become a nurse. Very unusual for the time. She arrives at a British Army hospital in Constantinople expecting to treat people with battle wounds, but instead she's treating people with preventable communicable diseases. And she finds sewage running throughout the halls.

5:50There's rotting food. There's just disease everywhere. Initially, she, you know, she really believed that sanitation was the cause of most of the casualties they were having in the war rather than the actual war. People dismissed her at first. The generals ignored her. But she had one weapon that the military men didn't have, and that was data. So she began

6:11tracking everything. Admissions, causes of death, time of year, the conditions of the facilities. And the numbers began to reveal that far more soldiers were dying from pre preventable disease than combat. And then of course she goes to

6:26advocate for this, and no one wants to hear it. So Florence created what is called the Coxcom diagram. You may have seen these. You might not have known this is what they're called. Um, but if you look in your BI tool and and you see one of these kind of exploded pie slice diagrams, this was actually created by

6:43Florence Nightingale to prove that the majority of the casualties in the Crimean War were from uh were not from

6:52battle related injuries. Um, of course uh you know she presents it to parliament. Policy change follows. They have cleaner hospitals, better sanitations, mortality rates fall from 42% to 2% for these soldiers. Um, and she becomes a national hero. Uh, so you know, everybody knows her. Oh, she's that famous nurse. But actually, what was her secret weapon for making these

7:17improvements in in medicine that continue through to these this day? It was data. And that is why Florence Nightingale is a hero of data. All right. I actually didn't know that till I was doing this research. I just fig like I knew the story of oh she was a nurse and she like improved some conditions but I didn't realize that she

7:35actually created that visualization type and used it to persuade people that that actually uh it was not you know battle wounds that were leading to all these casualties.

7:47All right. Um so Ryan let's get started. Mother duck and duck DB. Um, let's just start with with what is DUTDB? If let's say someone's never heard of it before, how would you describe it?

8:03>> So, uh, it really just depends on what else you've heard of. The easiest description for it is the SQLite for analytics. Uh, for those of you who don't aware or aren't aware, you know, SQLite probably has a dozen different copies, if not more, running on your phone, dozens or hundreds of copies running on your computer. uh it is an

8:23embedded transactional database. Um and

8:27then you know the folks at at duct DB uh so Hannis and Mark are the the co-creators of duct DB based in the Netherlands and they basically saw all

8:39the the academics around them using a uh

8:44you know using kind of traditional tools for analyzing data. So they would be using data frames, they would use R, they would use, you know, maybe some of the statistics tools like the SAS and SPSS's of the world. And, you know, they were just kind of pissed off that they weren't using real databases. These are database guys, uh, you know, PhDs in

9:04databases. Uh, and so they're like, we need to we need to bring a real database for these folks. Um, and and so they interviewed them and said, hey, what why aren't you using a real database? And you know, it wasn't because they didn't know SQL. It wasn't because uh you know that they didn't believe in real databases. It really just came down to

9:24the idea of nothing was lightweight enough for them. They didn't want to have to launch and and run an extra service to run Postgress or MySQL at the time. Um they didn't want to have to connect out to you know another another service. So you know they weren't using real databases. So HoneySmart created DuckDB. It's an embedded analytics

9:45engine. So what that means is it's super fast at aggregations. It's a polymer database. Um you know it takes advantage of all the latest uh improvements and research into how to make super fast polymer databases. Some people think it's only in memory. Uh that's not true.

10:03Uh it is it is uh in memory but it also has its own file format. So when you launch DuckDB, you can choose to specify the name of a of a file. Um, and it will store your entire database in that one file. Also sort of a design decision from those early days because they wanted to make it easy for the academics

10:23to share data with each other, not have to share like a whole folder of files, but could share a single file. So the single file represents all the tables in a database. Um, >> and yeah, I mean that that's basically duct DB. it's it roots and then it's it's gotten to be extremely popular as this really lightweight tool. You can

10:41interact with it in every programming language most popular uh you know in the in the Python world nowadays and it does actually have a interface with data frames. So you don't really need to know SQL but it's definitely much more powerful if you do know SQL.

10:58>> Awesome. Now of course this is a live show. So if you make comments, questions, we will read them and ask them to Ryan live. So get those questions in, right? Let's not treat this like a webinar, okay? This is way cooler than that. Johnny, >> way cooler than that, Ryan. That's Ryan.

11:14>> Way cooler than a webinar. >> Way cooler. So Johnny says, "Great origin story. Not working with databases makes me angry, too." Um, and so I think there's a lot of um a lot of people feel that one. What?

11:27>> So that's duct DB. So then how does mother duck fit into the ductb cinematic universe?

11:34>> Yeah, I mean the the I I will give the short of it today and then I will also give since we have time a little bit of the origin story as well. Uh the short of it today is it is a cloud data warehouse. Um it is a cloud data warehouse that is used for two purposes.

11:52One is kind of internal BI and analytics and one is customerf facing analytics sometimes called data apps. Um and so we get customers on both of those use cases. Many of our customers have not worked with DuckDB before. Uh they just come to us for the you know the efficiency, ease of use, low cost of of our cloud data warehouse product. Um and

12:17and some some come to us because they they find I mean DuctTV has focused so much on ease of use um and so much on just kind of the developer ergonomics you know that we do have some people come to us because of that but uh the origins really came from uh Jordan myself some of the other founders who

12:37have worked a lot in these big data

12:43spaces right so Jordan and I at working on Google BigQuery. He was one of the original engineers there. Um, you know, he went on to work at Single Store. Uh, I went on to, you know, work at at data bricks and we have other folks on the team, uh, who went on to work at Snowflake and things like that. Um, and

13:04we realized that a lot of these systems are just overarchitected for this universe that we predicted would happen back then. Uh so we predicted when when we were working on Google BigQuery that data will grow exponentially and eventually you will need these 1100 machines executing any given query in order to finish executing that query. Um and we were just wrong. Uh you know we

13:30we kind of marketed through fear. We said you know this is if if you don't attach to this massive distributed universe early you're not going to to succeed in your career. So you might as well do it. And uh it turns out that compute grew a lot faster than data. Uh the power the power of compute whether

13:51that's Moore's law or one of the other dozens of laws around compute and memory and storage everything grew a lot faster than the amount of of at least useful data that people are processing. So uh it turns out nowadays you can do many things on on a single machine. And that's really where mother duck started is is we said um you know how do we

14:12bring this efficiency uh from single machine compute into the cloud right? How do we take this technology that duct DB made and make it available for uh everyone in the cloud enable collaboration and sharing and all of those things that you would expect of a cloud service. Um, and so, you know, we reached out to to Hanis and Mark and uh

14:37basically said, "Hey, like, you know, we're happy to uh give you a a share of the company uh if you'll you'll work with us as as we build this." Hanis and Mark have their own uh company, DuckDB Labs, uh, and a whole gigantic team of folks working on DuckDB open source project. Um and then we contract them,

14:59you know, with them to to add some features that make it work a lot better in a um you know, cloud computing environment and such. But uh it really started off with that and then you know as we as we tuned over the first couple years of you know we've been around since 20 2022. Um so as we tuned over

15:19the first couple years we really nailed down those two use cases around data warehousing for internal BI and analytics and for the the customerf facing applications. Um so today um you

15:32know we also we don't just scale vertically we have tons of horizontal scaling in there as well and doing other things you know needed for other you know the the five 10% use cases and all um you know in addition to the the 80% use case which is uh you know single single machine compute that's super efficient. Yeah, it really has been my

15:55experience that you don't need the kind of the MPP, you know, scale scale out databases for most of what you're doing.

16:04Um, it's really critical for certain tasks and queries and use cases, but for the most part, you for your breadandut analytics, like a single machine is powerful enough to handle it. Um >> yeah and you just don't have all the overhead of like you know the amount of overhead of of shuffling in the old Hadoop world or um you know that that

16:26concept as it's taken into Spark and other engines like there's just so much overhead for the vast majority of data um and even at data bricks like I think the number one request I heard was for a single single machine Spark instance and it's like why do you have the overhead ahead of Spark if you're going to do

16:41that? >> Exactly. We do we do have a comment here. Sanjie says, "Duckb is one of the key foundational technologies we use for I'm guessing this is pronounced exqua or zqua, I don't know, the conversational data management platform. Happy to do the product placement there if we're there for you." Sen, I'll let you know where to send I'll let you know where to

16:59send the check. Uh we adopted it a few years back. Simple and very powerful.

17:04Love it. I hear this a lot actually. uh duct DB itself is a core component of many of the newer generation of analytics tools whether they're BI tools or some uh you know kind of more use case specific data tools. I know at good data we use DuctTV as part of our data platform. So it's obvious that the

17:24technology itself has been proven in the real world and I guess is it fair to say then mother duck is kind of like duck DB on the cloud and on easy mode? Is that really what you're offering?

17:37>> Yeah. And and a sort of uh you know making it into a multi-user system and things like that like so you know some folks will say well you know it's super easy to put duct DB in the cloud. Like look Postgress you just turn it on and and launch a bunch of VMs and we have Postgress in the cloud. You know duct DB

17:55and Postgress are not equivalent in that way in that you know duct DB was really designed as this SQL processing engine that's super efficient. uh and it does have storage but it actually doesn't have networking right so um yeah we we put it in the cloud we make it you know into this multi-user environment we make it work really well for uh SAS

18:17applications who need to embed analytics in their applications and make it really work really well for the 247 services like BI tools and things like that um yeah >> so then let's imagine that you're um so

18:33there's kind two use cases, right? There's this traditional data warehousing, then there's the analytics engine for SAS apps. Is that accurate?

18:40>> Yep. >> Okay. So, for that traditional data warehousing, let's imagine someone watching today is using Snowflake and they want to know, okay, how is mother duck different for for that use case? I I have to maintain a data warehouse and then I do BI and analytics on top of it.

19:00>> Yeah. I mean, I I'll start with what really hooks a lot of people. You know, as a geek, this isn't isn't my key hook, but uh what hooks a lot of people is I interviewed a customer the other week for a case study. Um and you know, they literally saved 65% off of their Snowflake bill uh when when they went to

19:20Mother Duck. Um and exact same workload. Um so that that certainly helps uh for for a lot of folks. uh the you know the the key things are around simplicity around like you know as as some of these platforms grow uh the snowflakes and the data bricks and the big queries of the world there's just so many knobs and

19:43dials and a lot of people don't need them uh and it just it just makes their uh you know configuration and management a lot more difficult so we aim for simplicity uh the simplicity comes down to the querying side as Well, um, you know, the if you look up on on Google for friendlier SQL, uh, in DuckDB,

20:06DuckDB has done a ton of things, uh, to make SQL easier to use. They say, okay, we're going to start at like the standards. Uh, and really they started with a with a Postgressbased uh, you know, parser for SQL. We're going to start at the standards, but then we're going to say like what are the things that we think, you know, would just make

20:26life easier. So they had, you know, they were the first ones to do group by all for instance and then a lot of the other uh engines including I think Snowflake and BigQuery added group by all. But why do you have to type in the names of all of these columns when you know what they are? Like it just doesn't make any

20:41sense. Um >> you know and they've done a huge number of other things to improve SQL u to make it a lot easier to query. So you know that aspect you get that with with motherdoc. The other thing that you get is the uh you know our web interface a lot of people love it for the instant

21:00SQL uh is one of the features we launched I don't earlier this year maybe March May um and instant SQL basically takes advantage of the idea that duct DB can actually run in the browser uh through web assembly you don't need to know those details but like it's running in the browser which enables you to do things like you read ahead and cache

21:23from the database so that when you type a query, it instantly populates the the query results. When you add a column, you instantly see that column. You have literally milliseconds uh between you adding a column name or or an operation in SQL and seeing what that what that does to your data results. Uh and that just changes the way we work. It gives

21:47us um you know what the creator Hamilton calls the flow state. Uh Hamilton is a a Grammy award-winning uh music uh artist.

21:57Uh and he now leads our UI team. Um and he wanted to take what you know the this music synthesizer world of hardware and bring that to data. And that's where really where you know instant SQL came about. Um so there's a lot of these sort of nicities that that we're adding on top of it to make it you know easier to

22:17do your analytics. Um you know another core one that that I would talk about would be uh sort of the model of compute basically what we call hypertendency. So the idea that every user within an organization um actually gets their own compute uh and the admins can adjust how large that compute is uh but they get their own compute so that you know

22:42they're not stepping on the toes of the other users within the organization um and you don't have to worry about uh about that shared resource problem that you do with the other data warehouses.

22:54Um, so I'd say, you know, it's between ease of use, lower cost, um, you know, also just kind of investing in the future that the, you know, the DuckDB folks will see an academic paper and within weeks they'll have the PR into DUTD DB to implement a new compression algorithm or a new way to speed up execution. Um and you know they and we

23:16are working with a lot of the the you know leading database professors out there um you know to to apply these

23:25things to and do additional research.

23:29I >> have a comment here from uh Olamei. He he says simple is great and and this is like a meta comment because it's a great simple comment.

23:38>> Yeah it's perfect. >> Um so thanks for joining us today again a lot of magi. Yeah. And then um I've let me pipe in here for I know you like talking a lot, Ry. Let me pipe in here for a second.

23:48>> So in in the world of of duck DB and and Mother Duck, have you seen a lot of like crossover of adoption in like the world of like bigger enterprise and you know ClassBI that maybe they have an Oracle database that is feeding SSAS cubes that's then being reported off of PowerBI? Have you seen like these big

24:11slow organizations kind of adopt to be in mother duck? And like if you have like where does it slot in? I asked because I work for a big slow organization.

24:19>> Yeah. >> That has the exact setup I described. >> So right now as mother duck we're focused on SMB and mid-market for the most place. Uh you know there are some features that larger organizations want to have that we don't yet have >> but also it's just it's a really good market that's growing really fast. So

24:39from the business side uh the TAM is there in the SMB and mid-market space and and we can provide a great level of customer support uh that people aren't getting when working with the bigger players. People in the SMB and mid-market space are not getting when working with the big players. So you know we have been focused there. there

24:58has been a lot of interest in the uh larger enterprise space in sort of the R&D groups and the leading leading edge folks of you know the big banks and the big uh logistics companies and other things that we've talked to. Um we haven't we haven't yet really really focused on those. Um I would say that some of them that have huge engineering

25:21teams uh have definitely adopted duck DB. So you'll see great talks about how octa uses duct db for instance uh where they break down their big data problems into little chunks um and they run

25:38thousands if not tens of thousands of simultaneous uh ductb lambda jobs uh in

25:45in AWS uh that kind of process each chunk of data as it's received all around the world at their different endpoints. This is kind of like for security log auditing, things like that. Um, and you know, they then use um, you know, they then use Snowflake as as sort of their gold tier, but they got rid of Snowflake

26:05in the in the processing all of the data. They're just having DuckDB do that at a much much much lower cost. Um, so the bigger organizations are able to adopt it. uh if they're sort of the you know R&D focused bigger organizations they're adopting DuckDB and kind of rolling it on their own. Um and then we actually have plenty of customers who

26:28started rolling it on their own and then came to Motherduck because they just didn't want someone they didn't want to manage it dayto day.

26:34>> Yeah. Um, but you know, we we're not yet into the, you know, Fortune 10, Fortune 100, you know, uh, even Fortune 500 space. Um, just because that's not really where where we're focusing our energy. Um, >> our our teams for, you know, on the sales side, uh, like to build strong technical relationships with practitioners. Um, and and we haven't

26:59yet done the, you know, take people out to golf yet, so maybe one day.

27:02>> Oh, yeah. Yeah, that's uh that's that's the next phase, right? uh eventually that you know the thing about that enterprise market is like the they can write big checks but with big you know mo money problems they say >> um and and it's it's the like the requirements list will have a lot of stuff that you know for mid-market it's

27:22not as important I I think it lets you guys move faster right >> yeah and I mean you know we've done the the like sock 2 compliance and we've done you know GDPR audits and we've done other things like that uh we just actually announced our our European data center. Um so, you know, we're making it easier for our European customers to

27:43adopt us. Uh but yeah, right right now we're focused on on the practitioners and not focused on the sell from the top that you often need to do in the large enterprises.

27:54>> So, here we have a a comment. This is from Bora. Uh thank you for joining us Bora by the way uh on LinkedIn. He says, "Thanks to Mother Duck, our non-data scientist customers who have big CSVs are able to analyze that data inside of chatgpta via their AIDBMCP server and and by converting the CSVs into mother duck."

28:17And thank you. So I think this is I mean this is actually a use case I'm really interested in which is uh when it comes to ductb and mother duck is um is how to pair it with things like chat gpt or claude to do analysis, right? like it just seems so like the time is so right

28:34>> for this kind of technology and its ability to quickly fit into these AI enhanced developer workflows compared to like a snowflake or data bricks and this is a perfect example of that. Yeah, it's so funny because like you know we do you guys know when when chat GPT was first announced like I feel I don't remember

28:54the date but >> I mean I I remember I kind of like 35 came out in like which is when it hit my consciousness and that was in October like was that October of 202 or 23 something.

29:08>> Yeah. So I mean basically >> November 2022. >> Thank you Mike. um you know when when Motherduck first started which was June of 22 um you know we certainly weren't conscious of of this uh and we did have discussions um literally on the first day we started in the Madrona offices in Seattle for those you don't know like Madna is a big VC

29:32firm focused on the Pacific Northwest which is you know where our headquarters are. I'm in Boulder. Don't ask. But you know uh the uh uh we started in those offices in Madrona and on the first day we were talking about like how does you know can we do text to to SQL natural language to SQL? Can we make it easier

29:52for you know the average business user. Um and at that time like you know the the whole founding team was pretty uh pretty down on that idea. um you know they had spent many years doing research on it uh in other in other companies and never really you know even putting the smartest people in the world never

30:14really doing it well nowadays like it's getting a lot better with with LLMs uh with you know things like we do have an MCP server that you can attach to um and

30:27you know if you add a little bit of semantic modeling uh semantic layers whether manually or asking the LLM to do it for you. Um, you know, you eventually get to a spot where it's it's it's decent. Uh, it's still not good enough to say, "Give me my numbers for my public company, uh, you know, uh, quarterly call, uh, quarterly earnings

30:49call." Like, "Oh my god, please don't do that." Uh, but, uh, it's it's getting good for a lot of day-to-day stuff. Uh where I particularly like it is just we have this thing called fix it where basically you can type what you think is a SQL query and it will fix it. Um so you know I get syntax wrong all the time

31:09or mix up the datetime formats between you know what is the Postgress versus the duct DB versus the you know my SQL datetime formats. Um you know and it just fixes it and so that stuff like is amazing. Uh and the text to SQL is getting better. We actually have uh customers in that customerf facing analytics world that run

31:30uh agents on top of motherduck and they're just building all sorts of queries themselves as agents and running a lot which is great for us on a uh consumptionbased business model but the agents are just kind of querying away and and keep on trying to learn new new facts from the data um and uh in a much

31:51faster way than humans could uh and and mother duck turns out to be really well suited for that.

31:56>> Yeah, I want to come back to that agent conversation and make sure we we spend a good chunk of time on that before we wrap up today. But but I do want to share this comment again. This is from Bora.

32:06>> He he says, "Here's a Slack message from from the customer he referenced earlier >> who was able to analyze this data via Mother Duck and he says it should make you feel good." So >> let let me read this one. Let me read this one.

32:18>> That data analysis tool you showed, Ryan, gamechanging for us. We just found a way to get data on our seating which has been which is something we had highly paid data analysts trying to solve all year without success. Bro, you have no idea.

32:31>> Well, you should definitely reach out to me and share more about this story.

32:35Yeah, like you're doing you're doing my customer case study work that for me. So, thanks. Uh definitely reach out on LinkedIn.

32:43>> And I I I feel like unlike many other companies, you guys could really run with the bro, you have no idea quote.

32:50>> Bro, you have no idea. Yeah. Which is >> Are ducks bros? I don't know. Like ducks, you know, we are also, you know, mother duck, right? So, is is the mother duck the bro? I don't know. But I I love that the simplicity of that statement.

33:06Let's put it that way. >> Yes. Yeah. Bro, you have no idea. This >> You have no idea.

33:11>> I mean, let's talk about that real quick. Just the uh the the branding. So, so going out like deciding to go to market as mother duck was um was

33:22obviously provocative. So, what what did that decision look like? Like how did you guys come to that you know what, we're going to do it. Let's be mother duck.

33:33>> Um so, the the name Mother Duck, uh you know, actually came before before I signed on to to join the founding team.

33:44Um, so this was I think it came in like April, May. Uh, we started the company officially in in June where we had the kickoff. Um, and the name came from Lloyd Tab. Uh, for those of you in the data world that that don't know Lloyd, he's the creator of Looker, which was bought by Google. Um, you know, he's now

34:05doing his own thing at at Facebook. Um, and you know, it was created by Lloyd and you know, it was just you have all the ducks who keeps all the ducks in order, right? And it's the mother duck that keeps all the ducks in order. So, it didn't come out of profanity that a lot of people try to to say that it

34:24comes out of uh that's that's not true. It really came out of like how do we we organize all the ducks in the cloud? Um, and so Lloyd came up with a name and my first call with Jordan. Um, basically I had taken a break after after data bricks and and went sailing for a bit.

34:42Uh, needed needed that break and um, uh,

34:47I reached out to Jordan and was just like, "Hey, what are you up to since you left Single Store?" And he's like, "Oh, we should chat." And and that's really how me joining the the founding team came about. But uh in in my first call he he mentioned the name Mother Duck and I'm like you know we're going to have to

35:01talk about that name but you know that that's beside the point and uh let's let's talk about the rest of the company. So I was not a believer in the first five minutes. Um, a day later I was a believer and and the reason I was a believer a day later is just talking to uh friends and family like uh their

35:23ears would perk up as soon as I mentioned that name. Uh it was it was shocking enough that it was memorable and uh that's you know 99% of of doing a

35:35startup in a in a B2B space is standing out. Um, and if you go to a lot of these conferences, it's a bunch of black and blue booths with names that I like could never remember. Uh, I even have products that I use once a day that, you know, every other day I ping someone on the team, what is the name of this product

35:55again because I need to find it. Um, and so like mother duck is a name you remember. And um, so it wasn't that hard. I mean I think basically we recognize that um the stigma against provocative names

36:14exists in you know in my generation in earlier generations uh but the people making the decisions certainly in the SMB midmarket space and and and the enterprise space by the time we get to it like these people are younger than us uh they are millennials or uh you know gen whatever else we're at now. And uh you know they don't care

36:38that the the name is is provocative. Uh it it stands out for them. And um

36:45basically we wanted to change the world of data to be fun and happy and golucky and make you know if you see a duck quacking uh you can have an error in the in the product and still have a smile on your face. like you might be pissed a little bit, but like the smile on your face helps helps you balance that out,

37:05right? Um and and honestly that's bring joy to data. Uh stop these sort of incessant battles between different companies putting each other down and just bring a little joy. Um so I think it's worked out great.

37:19>> Yeah, it's super memorable. Everything about it, the name, the you know the the images you guys use. I mean certainly if you if you can give me a a cute duck quacking on an error screen I am much happier than with your typical error screen with a big red X. Um so better at that at putting some of the ducks in.

37:39Right. So >> um the the irony is some of the founders are you know still against the name to this day. Um and you know it it is what

37:51it is. You're always going to have some some disagreements. Uh, but I think it served us super super well and and even those that are kind of against the name also recognize that it served us super super super well. So >> yeah, I mean I think when it first hit my consciousness was at Day-Tay Texas.

38:08>> Yeah. >> Where you you were on stage in like a duck outfit if I recall correctly. And I had no idea what Duck DB was. But the name and the presentation and everything stuck in my head like from that day forward I was paying attention even though I I really it was like at that point it was like yeah it's some new

38:26database right I I didn't even that's all the attention I had given to it but I was like I'm I was paying attention >> I mean I think the uh so just to to refresh your memory data council is where people were on stage in in

38:42>> uh prior to that there was

38:47uh New Orleans uh what was the New Orleans event? So Tino went around in a blowup duck thing in in New Orleans. Uh I think it was Coales. It was Coales in New Orleans.

38:59>> Oh yeah. >> Jordan did an onstage presentation in a duck outfit. Um, and and yeah, I so the reason I'm I'm actually even bothering correcting that is uh that uh day-to-day Texas, an event I really love dearly. Uh it's an amazing event. Um run by by a gentleman named Lynn Bender. Uh Lynn Bender is is still a little because you

39:24know we've talked about sponsoring two years ago I think we talked about sponsoring and I'm like I'm only going to do it if I can bring the duck. Um, and he was like, "I don't know about that for our audience, for our whatever." And I'm like, "Well, let me know when you're when you're okay with that and and I'll sponsor." This this

39:41next year we're sponsoring. Uh, I think he's he's been convinced. Um, but it does take some people a little bit, you know, to get to get accustomed to it, but uh, then it works works out really well.

39:53>> Yeah. And folks, for those of you interested in attending Data Texas, you can get, uh, 20% off your registration by using code super data brothers.

40:01brothers. >> So, uh, thanks for the plug, Ryan. >> Yeah. >> Yeah, absolutely. And I'll plug my own conference then, too. We have small data San Francisco in two weeks. Um, and, uh, I don't have a code for you guys all custom. I can make the Super Data Brothers code uh, after I'm done on this show, >> please.

40:21>> But feel free to reach out to me on LinkedIn if you want a discount for that. It's two days.

40:25>> Yeah. Yeah. >> Efficiency in data. >> Yeah. And while we're at it, data in the D conference, [Laughter] >> we're I think we have we have a talk there, too.

40:36>> You do? You do? Yeah. Yeah. That's coming up November 8th here in Detroit.

40:40So, um check it out, data.org. You can find the conference info there. Uh no discount code for that. It's only a hundred bucks. So, you know, I mean, it's it's a highly affordable one-day conference here in the D. So, and Alex Monahan is the guy that's giving the talk there for for Motherduck. Um, and Alex is the only employee ever uh that

41:03has been an employee both of DuckDB Labs and Motherduck. Now, he's exclusively duck uh but he's a super knowledgeable guy. He's actually the one that wrote a lot of the articles on uh the friendlier sequel with DuckDB. So, uh show up show up there and ask him all the questions.

41:21Yeah, we're we're really happy to have you guys out here. Especially I think um you know the if you look at Detroit's economy, I think we have a lot of of uh we have a lot of firms that are looking to start to modernize and kind of have, you know, are still operating in a pretty old school enterprise way,

41:41but I think that's changing. And so, um we're happy to be able to bring you and and other more innovative firms in to kind of show people in Detroit like, hey, here's some of the stuff you could be doing, you know? Honestly, like I I grew up in the tech world in sort of the Silicon Valley bubble. I was an engineer

41:56for five years in in Rochester, New York, but then moved to San Francisco in 2006. Um, and

42:04it it is a bubble and it is a environment where you just get accustomed to all these new technologies and all that and they aren't as as as exciting. Um, and then you leave that world and you go to day-to-day Texas or presumably you go to Detroit and everyone is super excited. There was a DuckDB meetup in uh was Tennessee or

42:26something like that that like filled the room, right? And so um yeah, I I think that getting outside of that bubble is always fantastic.

42:35>> Yeah. So, let me uh let's read a couple comments here. First of all, we've got um Johnny Winter saying, "Mother Duck is a better name than Looker." um and that he's he's having conference FOMO already. So yeah, uh all three of these conferences are great. Johnny, come join us here in the United States. Um and then I do want to address this this

42:53question for Puja. So Puja says, "For collaborative teams, how should we structure a mother duck environment? Do you have supports for roles, something like shared workspaces? You know, do you use naming conventions for data sets and queries? Like just what is the standard for keeping things clean?" So the the basic structure so we'll talk about you know mother duck for what I

43:15presume is an internal BI data warehousing uh space you know there's there's different structures around the customerf facing analytics but you know the general structure so we use mother ducks motherduck ourselves for our internal data warehouse um and uh we

43:31actually use omni as as our bi tool on top of that uh but everyone in the company has access to the internal data warehouse house. Um, and you know, for those of you at thinking about size, I think it's like a three terabyte data warehouse at this point. Like, you know, duct DB does scale. Um, I will I will

43:50say that very strongly here. A lot of people assume that it doesn't, but um, and so the way that we have it structured is is basically we have uh service accounts that go in and build the data. So we have uh I think it's still airflow jobs that go and uh process the data from all the different

44:10data sources HubSpot and uh our own product and data dog and and other things um and you know do all the ETL and load the data in uh owned by that that service account uh and then what happens is that there's a a readonly share uh for what we call as MDW which is motherduck data warehouse else. Um,

44:34and you know, there's a there's a readonly share that everyone in the company has access to. It's delayed by a minute, two minutes, something like that. Uh, but it it's basically a uh read replica that can scale, you know, fairly infinitely for all the users within the organization to access. Um, and then every user when they log in,

44:56they get their own duct DB instance. uh the the share is is available to them because we share it organizationwide. So anyone that has an account within our organization. Uh so we do have so the notion of an organization and admins in that organization and then a set of users and then you know you can share things with the whole organization. So

45:18people will just see it in their left nav uh when they log in as as one of the databases available to them uh and they can just start querying it. And uh you know the beauty of this this architecture is they're quering it on their own compute. So it's a separation of storage and compute. They're querying on their own compute accessing the data

45:37uh in this share uh running their own reports and and you know however they want to do it. Um and then you know the BI tool also connects uh through that.

45:48Uh right now you know we have a thing called read scaling. So the BI tools, you know, can uh you know, end up using many many machines to connect and and uh you know, and query the data. Um and we're working with with a lot of the BI tools to roll that capability out because they kind of need to pass an

46:07identifier of of who the end user is in order to for that to happen well. Um but you know, the the core concepts of organizations, users, shares are all there. Uh there's roles in terms of admins or non-admins today. Uh we're working on expanding roles. We're working on things like rowle security, column level security, and you know, all

46:30sorts of other things that as as we go higher up mid-market that folks need.

46:35>> Yeah. Integration with active directory.

46:39>> Uh that is SSO stuff is being worked on. Uh presumably that you know active directory is in that scope. to be honest, we haven't heard it that much or I haven't heard it that much. Um, I think as we expand more in Europe, we just launched our our European data center. Um, so as as we expand there,

47:00we'll see Microsoft uh Microsoft products pop up more and more because they're more dominated uh in in that world. today. You know, you can log in with your Google Apps account uh through through uh you know, Open ID or OOTH uh

47:15an old friend of mine and uh uh do that. But, um yeah, eventually we'll get to to full like Active Directory and and all of the other larger enterprise features.

47:26>> We got a couple more great questions here. Before I ask them, folks, if you like this kind of content where we do live interviews, where you can get your questions in to real leaders in the data space, I'm going to need you to click the like button on YouTube, subscribe, hit the bell, and if you're watching us

47:44on LinkedIn, why don't you go ahead and share this to your LinkedIn feed so we can we can spread the love of the Super Data Brothers throughout the community.

47:51>> You also need them to share it to their LinkedIn feed, right? Not correct. You want them to.

47:56>> Yes. Do it now. Um uh so uh a couple

48:00questions. So Michael Duny asks a question. Could you share why you chose Omni over other BI vendors? He's looking at alternatives t to Tableau. Very popular topic right now >> and trying to learn more about Omni. And I would say Michael um if you go back in the show archive the very first interview I ever did on the Super Data

48:21Brothers was with Colin the CEO of Omni. So uh so go check that out if you're interested. But but Ryan, tell us about Omni.

48:29>> Uh well, first of all, if you're if we're doing plugs here, Colin and the Omni team will be at Small Data SF in two weeks. Uh and you can ask them directly. But uh why did we choose Omni?

48:39I think that there's there's a couple things is uh you know there's a there's a team culture uh around this. A lot of the Omni team came from Looker. So after Looker was bought by Google um and kind of Google we'll say changed their philosophy on how to support Looker uh and the community um and uh so our head

49:01of customer success ran the department of customer love uh at Looker um and our head of bisdev uh came from Looker uh and there there's a few others. So, um, you know, it's only, I don't know, 10% of Mother Duck as a company are are exookers. Uh, but, you know, they're a, uh, a very motivated 10%, let's put it

49:23that way. Uh, and so, you know, we had those folks and, um, you know, looking at what they did in Looker and what they were planning on doing in Omni to build upon that in their next company.

49:39um you know there's a lot of just belief in the team and where the product is going and what the team is doing and there's a lot of excitement around that.

49:47It's hard to say that for a lot of other BI tools. Um and uh you know that's what

49:54what we believe startups are all about. Like I think that a lot of people will choose a product that is not yet, you know, where they need it to be, but they have strong belief in the team and where it's going. And uh that's much better than choosing a product that, you know, is quasi where you need it to be right

50:13now, but it's on a downhill path. Uh which is what is happening a lot in in industry due to acquisitions and things.

50:20So um you know, I'd say that's a lot of it. We do believe in sort of their visualizations and and just you know the the beauty that they attach to their visualizations their semantic model way of uh of referencing data within their visualizations. Um and uh but you know

50:41we also do have a lot of other BI partners and I don't want to I don't want to expand on this too much further because I don't want to be critical of them. We want to be the the happy good folks in in data. Um and I guess I'm talking to one of the BI partners uh right now here too. So

50:58>> I wasn't.

51:03>> All right, we have uh we have one question here and then and then I do want to make sure we talk a little bit about kind of what you see coming from agents. Um but I want to ask this because this is something I'm wondering too. Sanjie asks in your experience what

51:18are the typical real life use cases where people are doing in imbrowser analytics and and what concerns if any do you hear about that way of doing it and this is something I'm I'm really interested in I mean even purely from a BI perspective I think if you could ship off a data model to the browser you know

51:37just the speed of interactivity is is better than sending queries to a server right um so this is something I'm be curious about.

51:46>> Yeah, I mean I I think DuckDB is all about how do you be super lightweight and you're getting lightweight enough that it can run in the browser in web assembly. Um, you know, the

51:57a lot of BI tools do it already and you don't know about it necessarily, but you see them be super responsive in the browser and and Duck DB is what's enabling that. Um, and the, you know,

52:12the the challenge with many of those BI tools is then, you know, they have to build a second layer of SQL translation or whatever because the SQL that's running inside the browser is duct DBS SQL and if it's connecting to then a Snowflake instance or something like that, there there's a translation that occurs that just makes it a little bit

52:30harder to implement and maintain from an engineering perspective. Um, but you know, a lot of the BI tools do do it. uh what we are seeing is more and more of the customerf facing analytics want to have it. So you have a bunch of end users. Let's say, you know, you're looking at uh we have a lot of logistics

52:48companies. So you're looking at, you know, um some of the the package delivery services and things like that.

52:56And you want to see a map of where things got delivered and you want to be able to zoom in and out of that map or you want to be able to like, you know, as you expand your view to query more data from from the database.

53:10um or you know you want to see charts uh over time and you want to enable you just to kind of scroll right across that chart and see the data instantly. You know that's the type of stuff where uh where it really makes a difference. Um especially kind of in that sort of consumer or modern B2B tools space. Um

53:31and uh some of those have not yet implemented it. But you know the reason they choose to be based off of DuckDB and and mother duck is because that is available down the line. You know I'll be brutally honest here is uh WASM itself as a technology is kind of a pain in the ass. Um and to get to the other

53:51part of this question the reason it's a pain in the ass uh is that it does put tons of security restrictions on on what your your web application can do. and you start start to need to do, you know, lists of every permission that you're granting the web application to do uh instead of sort of the default allow all

54:09uh that you're used to as a as a web app developer. So, it can be kind of painful to implement. Um, but you know, luckily a lot of other folks have done that for you if you if you choose one of the BI tools, especially one of the BI tools that does sort of embedded analytics.

54:25>> Awesome. So, let's uh we got about five minutes left, folks. You've been a great audience. Again, I would ask you, you know, like, share, subscribe, all that good stuff. Um, go do it.

54:37>> Yes. Uh, I do want to close by just asking you about what are you seeing in the early days of of agents and the impact agents are having on data loads or data structures? Just like what do you see today and where do you see this headed?

54:57Yeah, I mean today we're starting to see a lot more agents interacting with databases. Um I I think there's been some quotes around like some of the bigger transactional databases uh out there in the world talking about how agents have uh taken over a lot of the workloads. But we're also seeing it in analytics. Um, LLMs and and kind of the

55:18the modern uh AI tools are are really sort of really bad at aggregations and really bad at facts in general. Um, and

55:29so, you know, I I've seen this when, you know, putting in a a list of things and saying like how many are in this list and it's like there's three even though there's seven in the list. I mean, you just get all sorts of wrong data.

55:43uh but yet data is important for making a lot of decisions, right? And so uh the advantage then of hooking up a LLM or an agent to an analytics database is that you know you have the analytics database doing the facts um and looking up the the results and doing the aggregations so the AI doesn't have to. So it's

56:06really kind of a specialization of interest uh you know idea there. Um and

56:12and like I said earlier, a lot of times what they'll do is is they can execute queries a lot faster than a human. So they'll execute, you know, 10, 15, 20 queries and see, you know, uh just to explore the data and understand the data better. Um and then, you know, they can they can then take meaningful action

56:30from those uh those different results. Um, you know, where we're seeing it really awesome with with duct DB and mother tug is is that hypertendency architecture. Uh, whereas you might not want to pu push a agent or point an agent at your snowflake instance and let it run away like wild on your snowflake instance. Uh, from a cost perspective

56:53alone. um when you restrict it to uh

56:58essentially a you know single machine sandboxed environment that you can choose what data you're sharing to etc.

57:06Um you know that works a lot better in the agent world and you don't need to really worry about that sort of runaway costs. Um and so we're seeing a number of companies doing that with mother duck today. Um, and you know, we, you know, as individuals on the at the company also are, you know, looking at agents to

57:26sort of power our workflows based off of mother duck data using things like the, you know, MDW that I talked about earlier.

57:33>> Awesome. All right, so that brings us to the top of the hour. Uh, Ryan, I want to thank you for joining us today. Awesome show. Uh, where can people keep up with you?

57:44>> LinkedIn is probably the best. Um, I'm really bad at most other socials uh nowadays, but so keep up with me on LinkedIn. Uh, feel free to reach out to me at any point. I'm also just Ryan at Motherduck. Kind of bad at email, but I will try. Um, and would love to hear from you. And, uh, we also do have our

58:04own Motherduck community Slack. Uh, there's DuckDB discord. There's, you know, all sorts of other ways the community can interact with each other, too.

58:12>> Awesome. Well, uh, folks, check out DuckDB. I am a believer in the technology and a big fan of Mother Duck.

58:18So, um, really glad you could join us here today. Thanks, Ryan. >> Thanks, Ryan. And thanks, Eric. So, great. I did not know until now that you're real life brothers. So, thanks.

58:28>> We are real life brothers. >> Yep, we are real life brothers. That's right.

58:32>> That's why we say it at the opening every time. >> Uhhuh. Yeah, because that's the number one question we get asked. Are you really brothers? Like, yeah. Yeah, we're really brothers. All right, see you, Ryan.

58:39>> Cheers. Bye. >> All right, man. Um, I just really, you know, I realized when I was writing the show, not the show notes, but I was writing the promo post for for this episode that it was going to look like sponsored content >> because I like because I like >> Duck TV and Mother You're too.

59:00>> Yeah, really. But this was not sponsored content, folks. Um, or rather not by Mother Duck. It was sponsored content by Good Data. Uh, so coming up November 12th, 9:00 Pacific time, 12 o'clock Eastern, 6 PM Central European time over at Good Data, we are talking to you all about Agentic AI in the financial industry, particularly how people are

59:22using good data to create customerf facing agents that do things like analyze data, suggest outliers, you know, help customers rebalance portfolios, uh, evaluate credit risk and take actions based on the credit risk.

59:37all these sorts of things in customerf facing data applications. And so, uh, if you're particularly if you're in the finance industry, you got to come check it out. Even if you aren't, I think there's a lot you're going to learn about the cutting edge of what people are doing with agents and customerf facing data apps. So, join us on

59:53November 12th for that. Um, okay. Uh,

59:58last thing to plug. I feel like this we're just like plugging stuff like crazy here this episode, but um, this is how it goes. It would be awesome if we were uh getting paid for all this. But um the last thing I wanted to plug, did you know we have a blog, superdatablog.substack.com?

60:13Head over there and subscribe to to check it out. The latest thing we posted was about if you're a BI person and you're freaking out because your BI tool is dying, what can you do to insulate your career and take the next step, right? And it's it's not just wait for someone to get good data on Omni.

60:29There's things you can do today to help take the next step in your career. So check that out.

60:33superdatablog.substack.com. Okay. Um I think that's it, right? That's got to be it. >> That's it.

60:39>> All right. >> I think that's all right. Thank you everyone. >> All right. Until uh next week, >> we are the Super Data Brothers. Take care everybody.

60:47>> Thank you. Bye.