[00:00:00] ​ [00:00:00] Phil: [00:01:00] What's up everyone. Today we have the pleasure of sitting down with Sarah. Sarah Krasnick Bedell, director of growth marketing at Prefect. Sarah studied math and cognitive science before completing a master's in data science. And she started her career at Amsted working on data aggregation and machine learning models, and eventually moved to a customer centric role where she helped engineering data architecture for supply chain optimizations. [00:01:29] She has short stints in financial forecasting and then company wide data architecture, [00:01:34] She then joined per pay as a data engineer focused on product analytics as well as reverse ETL for their marketing team. She was eventually promoted to lead data and managing the full team of data engineers. She's also an analytics and GTM advisor for dev tools. And today she's director of growth marketing at prefect, a workflow orchestration tool for data and ML engineers, Sarah, thanks so much for your time today, really pumped to chat.[00:02:00] [00:02:00] Sarah: It's awesome to be here. Excited to have this conversation. [00:04:00] Phil: I'm excited to chat with you as a data engineer turned marketer. We don't chat with a lot of folks that made that transition. I think a lot of marketers will agree that the relationship between engineering and marketing isn't always the greatest. I actually did a full episode last year, unpacking why Martech is actually for engineers for a lot of folks. [00:04:23] But that overlap is getting really blurry with Gen AI, but we also went super deep into like attribution. And so there's a bunch of topics that we could chat about here. Um, but Daryl, I'll let you, uh, take the first question for Sarah. [00:04:38] Darrell: Yeah, I, uh, great to meet you, Sarah. And, um, yeah, [00:04:41] Unconventional Paths From Data Engineering to Marketing Leadership --- [00:04:41] Darrell: I'm really interested in the, in the transition from engineering and data to marketing and there's, um, it seems to be a big jump. What, what, uh, how did you make that jump? Was it primarily self taught or did you go back to school? Did you take some bootcamps or courses? [00:04:59] Did you read [00:05:00] books? Tell us about your journey there. [00:05:02] Sarah: Absolutely. So when I got my master's in data science, I vowed to myself that that would be like my last formal education. I just knew that that was the furthest I wanted to go is getting master's. Um, I was ready to, uh, Um, learn on the job and that's exactly what I did, um, in this transition. But even going further back, right, in data engineering, a lot of the skills that I learned were also on the job. [00:05:25] So even though I started, right, I took computer science classes during my undergrad, during my cognitive science degree, a lot of, I'm sure software engineers will say this as well, a lot of the computer science classes don't always match to That's kind of like what exactly you're going to be doing day to day. [00:05:42] And I certainly felt that, especially because there was no major for data engineering even, right? It was a certain flavor between data science and software engineering. So I kind of had that learning on the job experience and really enjoyed the practical application. So, [00:06:00] uh. During my time, um, leading the data engineering team at PerPay, one of my biggest projects was working with the marketing team and helping them onboard Iterable as an email and SMS marketing platform. [00:06:14] And as you can imagine, onboarding that type of platform means having customer data, all sorts of different analytics in the tool for audience segmentation, for personalization, for right, Perpay was an e commerce platform, and so adding all the data about the products that we sell, and so there's, that was actually a really big data engineering project. [00:06:38] Through that, I learned a lot about what iterable could actually do. Do and how marketing worked like that was really my first introduction to that and honestly just found it very very interesting Um, that was also the time during which I was kind of trying to Explore what a path in my career could look [00:07:00] like that wasn't tied to how well I could code So with data engineering Similarly to software engineering, it's a lot about the code that you write and what you ship. [00:07:10] And for me, I was much more interested in the output as opposed to the actual process. And that was kind of a learning for myself. And so for me, I just got as deep, um, with the marketing team as I could learning at Perpe, learning from them and really under trying to understand how everything worked mechanically, because that would allow me to understand how the output was possible and the impact of the output as well. Speaking to the actual transition, um, so during, this was all during COVID and I'm sure plenty of us, right, had a lot of time, picked up a lot of hobbies. Um, I'm gluten free and so I didn't bake bread because that would be a lot harder. And so what I did was I started writing. Um, and so one of my, um, one of like the first things that I did was just write about some of the data [00:08:00] engineering work that I was doing. [00:08:01] And. That just started, people started reading it, people started saying this was helpful, or that they liked it or appreciated it, and I just got more and more into that, and at some point, people started, right, people started to pay me to write in the data space, and that kind of started to snowball, and I kind of got the pulse of, there's something here, I don't know exactly what it is, but there's something here. [00:08:24] So that kind of led me to, um, having, to, to leaving my full time role and exploring, uh, consulting full time. And so I, I did that. I had a few, uh, contracts that, um, I set up after, right after I left Purpey in terms of data and analytics work. But then it gave me the time to also go deeper into writing and working with marketing teams at every level. [00:08:49] Several different companies, and I think that action that exposure to several different companies really allowed me to see what it's like a different places as opposed to just being in one place. And [00:09:00] so that's that's really how I landed at Prefect. And it was really more about, um, knowing the product, knowing the people and knowing kind of what that marketing motion looks like from the consumer, right from the end user perspective, just because I know the space and I was previously in the position of being a potential buyer. [00:09:18] Throughout my time at Prefect, there's been, again, a lot of learning on the job, just like there was, um, when I, right, when I first got into data engineering, and so I've really, really enjoyed that piece. That's supplemented by kind of some extracurricular things, right? I have several advisors that I speak to and learn from. [00:09:37] I read a lot of right newsletters and just read from people in the space about what the trends are. And of course, the Reforge community, um, is, is on Parallon, has been really helpful, um, in, in learning about the space as well. [00:09:50] Darrell: Yeah. I wonder. I think that that's, that's interesting because, you know, I, what I'm hearing is you're focusing more on, you, you, you got more interested in [00:10:00] the outputs and maybe the concepts versus like the actual technical coding, which to me probably is a good move just with the rise in popularity of AI taking over a lot of these, the actual like more menial programming tasks. [00:10:17] And then secondly, [00:10:19] First Principles Marketing Against Common Practices --- [00:10:19] Darrell: It's probably paid off as well that you maybe you didn't start in marketing because, and I, and I'd love to hear your take on this. So many of marketing best practices are just not good. They're just, you know, someone came up with it and everyone follows for some reason, just follows it. [00:10:34] Like send email, uh, 8am on Tuesday, you know, everyone, so everyone just sends it and there's really no data, um, to, to guide that. Um, is that your experience as well? Like, do you find that? When, when you, when you think about these like marketing concepts or like more business concepts, you're kind of like, Oh, is, this is not really from first principles, like I have a better way of, of doing, [00:11:00] uh, of, of thinking about this, do you think about it like that or not really? [00:11:03] Sarah: Darryl, you're really speaking my language. Because for me, I, like, I was just born a natural skeptic. Like, when everyone goes right, I kind of have a tendency to go left a little bit. Um, and so I do that in marketing as well, where the way that I approach things is from a first principles standpoint, which is, what do we want, let's, let's, Take things in two different, right? [00:11:24] Let's in two different ways. The first is what do we want the world to look like? And what is the ideal and then what is possible? And let's find a way and a happy medium between those two. And I think that's really how I approach growth marketing as well. And I think that right to your point about AI. I think there's um, In data engineering and in software engineering, I think there's just more of a stress now on architectural decisions, just like I think in marketing. [00:11:51] There's more of a stress on strategic decisions of who are we gonna go after? How are we gonna go after them? How are we gonna engage them? And then right? [00:12:00] Step two is like, Okay, let's just let's write the copy or something like that, right? So, um, I think it's it's similar in those ways. But, uh, yeah, there's definitely a lot of things that I questioned in marketing, and I'm sure, um, you know, For anyone who continues listening, which I hope you do, right? [00:12:15] You'll hear that in some of my in some of my rhetoric a little bit. [00:12:19] Systems Thinking Applications For Marketing Analytics --- [00:12:19] Phil: What, what if you had to give advice to folks that are considering the opposite transition, like maybe not a full transition from marketing to, to marketing? Data engineering. But what is like some of the advice that comes to mind for you, for marketers that want to improve their data literacy, if you want to be like an analytics or a data leader, but still kind of in marketing, but be respected by data engineers, what skills you really need to know kind of at minimum in your opinion. [00:12:46] Sarah: The answer that I'll give is probably a little bit vague, but it's the most relevant, which is systems thinking. I think that applies a lot in MarTech, but it's, um, How does [00:13:00] data roll through the system? And how are we going to architect the transfer of data through different systems, right? In Martech world, it might be from, um, like signals from a lead into, um, our email platform or CRM into, right, the hands or context of reps reaching out right there into maybe the warehouse to be able to run analytics on it. [00:13:25] So there's a lot of, you know, areas of data transfer and it's trying to think one level above, um, in terms of how to architect that data transfer. And so that is at its core, a data engineering problem, just in the marketing world, right? There are more drag and drop tools and the data engineering world, you're probably going to go just like one layer deeper and closer to the right, the actual code and the actual process. [00:13:52] And that's really fundamentally the difference. I [00:13:57] Understanding Developer Marketing Through Engineering Mindsets --- [00:13:57] Phil: folks that are in data [00:14:00] engineering or maybe like engineering in general that I don't know if like allergic is the right term, but they're kind of like allergic to marketing. There's like this disdain for marketing, maybe because it's chaotic, but also. Because like, it's, it's a bit dirtier, like we're selling stuff and a lot of engineers want to build products and they're excited about building features and improving the user experience. [00:14:24] And not a lot of engineers make the transition you did to going up every morning and. Selling a product and like building a website and creating a way to get people to convert. Like what advice do you have for data engineers that are maybe listening today that like marketing isn't that bad of a transition? [00:14:43] Like why do people have in engineering this disdain for marketing in your opinion? [00:14:50] Sarah: think engineers have a disdain for bad marketing. And that's like 95 percent of marketing and touch points. And I think it's like one [00:15:00] bad apple ruins the bunch where if you have one bad touch point or even just like two bad touch points out of maybe potential potentially 50 touch points from a company, you're going to categorize that whole company is having like in that kind of bad marketing camp. [00:15:13] And I think that's what Um, engineers are are allergic to and frankly, rightfully so right now being in the space and getting under the hood and how these things work like there's there's a lot of moving parts and there's a lot of different ways to to approach marketing. Um, and so my my advice would really be to think about marketing as educational. [00:15:33] It's I'm I don't think of myself actually is a selling and I don't think of growth marketing. The team is selling. I think of us as trying to display the value that Prefect can provide to our audience and to potential users and exposing that, making it as obvious to you as possible. And if it's not obvious, then we're not doing our job very well. [00:15:57] And if it feels like [00:16:00] spam, we're also not doing our job very well. Um, and so that's really the way that I think about it. And so the exciting. One of the reasons that I kind of moved into marketing was the, I found, I felt that there was this opportunity to really change, um, how marketing to developers worked, um, and really be part of, of that change because I think it's evolved so much. [00:16:25] It will continue to evolve. And so that I think is the opportunity at hand. [00:16:31] Darrell: Yeah, I just had a thought like, [00:16:32] The Case for Decentralized Data Teams and Embedding Analysts Within Marketing --- [00:16:32] Darrell: so let's say you have your average marketing leader and you know, they didn't come up from data, right? Um, but they, they want to start approaching things like attribution, they want to approach segmentation, they want to improve their reports, like, do you think that it would probably be better for them to maybe hire a data analyst or, or data specialist as their [00:17:00] partner or like right hand person and then approach those? [00:17:03] And like, if, if that is the answer, like bring someone on. What skills does that person need? And the reason I'm asking this is because, you know, I, I think that there are some really critical data problems that, you know, organizations need to solve, and if that, those types of data skills are lacking, you know, I think you, we, we, we would need to do something about that. [00:17:26] We need to fill the gap somehow. Um, what are your thoughts on that? [00:17:30] Sarah: I think this really gets into, uh, how to sell the role of a data team internally, but also the structure of a data team. So there's kind of two different models. One is a very centralized data team, and the other is a very decentralized one, where there are analysts, for example, like a marketing analyst, a sales analyst, a product analyst, right, that are really more focused on different, um, different pieces of the business. [00:17:58] And I kind of [00:18:00] tend to lean on, right. There's so a marketing leader doesn't really care about the underlying data architecture. Like they shouldn't care if right data pipelines run on right within AWS on like an EC2 instance or like something else like that doesn't matter. Right. And so I think that a marketing leader should really think about who's their analysts and how well do they understand the underlying. [00:18:24] Data that they can dig through and grab insights. And then also how well does that analyst, how embedded is that analyst in their team to understand what their goals are? Because an analyst can go on a fishing expedition, go digging really deep, but not find things that will actually move the needle. And so the way that I think about data is what questions are you trying to answer? [00:18:46] And a marketing leader will know that what they won't know is where do you go looking to answer those questions? And I think that is someone that is. You know, a data analyst by trade and has exposure to that is really into the actual what data actually [00:19:00] exists is going to be able to kind of, um, to pair those two. [00:19:03] And I do think that right these days, every company is a data company, right? Every company runs on data. Every company has data, and we're at a point where if you're not using data to your competitive advantage, you either are. Already falling behind or will soon fall behind. And so I do think it's very critical to be very forward thinking in terms of how are we using data not only to better our own decisions, but to understand our actions and how to, um, Create that pairing between the actual engineering with here, the insights that are needed for for my team and here, the goals of marketing, whether it's right driving sales, whether it's driving engagement or whatever it might be. [00:19:49] Phil: Such a good take there, Sarah. Like every company is a data company. It's funny because like 10 years ago, I grew up in marketing where there wasn't even a data team. Like we were [00:20:00] fighting with the product team to get technical engineering support when it came to like APIs and hooking up different things or getting like a script on this site. [00:20:10] And today, Flash forward, just like a couple of years later, most companies, especially in tech, like have a data team. And now we're sitting on a MarTech podcast and most of our audience is marketers, a lot of them are technical marketers, but we're talking about things like pipelines and data warehousing. [00:20:27] Comparing Batch Processing And Webhook Architectures --- [00:20:27] Phil: And I know that you're deep in the process of building a warehouse first tech stack at Prefect. And maybe you can give us an overview of some of the tools that are making this possible for you. I'd be really curious to hear like the decisions. Along the way that you had to make. Um, and maybe you can start off actually with like, what do you think is the role of the data warehouse and Martex? [00:20:49] Like a three part question. [00:20:53] Sarah: I think the data warehouse has to be the center of information, and all [00:21:00] information has to eventually be shared. Get into the warehouse and the reason for that is because there's so much data in so many different places that I think it would be close to impossible to actually get a full view without having that data warehouse at the center. [00:21:17] Um, and even if it were possible, it would be really hard from a, you know, martech standpoint to make sure that all of the data successfully flows from one place to the other and vice versa without the data warehouse kind of being that central hub. And so So that's kind of my, uh, my viewpoint and really the tools that are involved or it's, I think it's, it's a two way street, right? [00:21:43] So it's what are the tools collecting information on the outside, right? Whether it's, um, Amplitude for product analytics, whether it's right. We use common room for understanding the behavior in our community. Um, for for [00:22:00] pre fact, right? Those types of tools. And then how do we getting that data into, um, into the warehouse and then understanding then merging it with other first party data. [00:22:12] So whether it's product usage or it's Um, sales, behavior, touch points and then surfacing it back into, um, either sometimes those tools to be able to kind of create that, um, flywheel of being able to learn and create segments in something like common room or more simply, it's then right forwarded to something like a, right, some sort of CRM where there's a, a one stop shop for right. [00:22:45] I'm in right b2b world right now. So one stop shop for sales reps or, um, a marketing team or a dev, uh, dev rel team or something like that to really understand, okay, what is the kind of full view of who this [00:23:00] person is and what they've done? Um, The, there's kind of a lot of pieces in that. Um, there's two different approaches in, in how to get data from one place to the other. [00:23:11] The first is, um, just kind of dumping everything into a data warehouse and using something like DBT to transform it, to store it, right? We use Prefect for scheduling and, and putting all those pipelines in production, shameless plug, um, and then using, right, something like Census, for example, to write it out into your email marketing platform, into a CRM, et cetera. [00:23:32] But there's also a second way, which is relying more on webhooks. So, for example, right, someone does something, it triggers a webhook into another system, and that's a little bit more, I hate this word because I think it's really loaded, but it's a little more real time, right? If you want to, um, have, not rely on a batch process that runs, Um, every few hours, but if you really want that, like, okay, one thing happened, now someone gets notified, um, that's a [00:24:00] little bit more of a, of a straight shot to, to that world. [00:24:03] So, that's kind of the two different ways I think about it. I would say the most notable thing that I kind of left out was Google Analytics, because I don't think it really has a place in, in this type of world, but, um, we can kind of get, get a little bit more deeper into that. [00:24:19] Phil: Actually, I'm curious to hear your thoughts too, Daryl. Like, I haven't used Google Analytics in over half a decade. It's been completely replaced by a couple of different tools or, you know, usually like product analytics tools that will also sit on. The web domain and amplitude will give me data connecting on the product, but also on the website and also in the mobile apps, [00:24:43] Reconciling Web Analytics Discrepancies Across Marketing Tools --- [00:24:43] Phil: I know you have a bone to pick with GA Sarah, like, what do you think are good use cases still today for GA and which should be done more by product analytics instead. [00:24:57] Sarah: Yeah, I have a secret, not so [00:25:00] secret, war against GA sometimes. Um, and I think it's really that GA is really good at two different things. The first is, right, because it's a pixel on your wall. On your website, right? They have things like bounce rate and all of those that like user stuff that sometimes if not implemented properly, third party analytics tools, um, might not do as good of a job of, but with the caveat, like if implemented properly, usually they, they can compete. [00:25:32] Um, ga also, right, has. Like, it's Google, and Google kind of knows, knows all, like, I am part of the problem, I have a Google Pixel, I've had the Pixel 7, right, so it's like, it's this ecosystem of data that Google has, whether it's from the data they have on display ads, right, on, on other sites, and, and this whole ecosystem on the internet that they can then, right, pick up different referrers, or pick up other [00:26:00] information, not just third party tools that aren't, that aren't As massive as Google don't have, but I do think that it's the question is, what is the value of that compared to the drawbacks? [00:26:11] And I think the drawbacks are pretty huge. I think the biggest drawback is when you think about conversions. So. The immediate response of someone that's used Google Analytics recently in a lot, they would say, well, you can just send conversions to Google, to Google Analytics. And yes, you're correct about that, right? [00:26:28] But you send conversions at a starting point, moving forward. And what I mean by that is what if at some point we want to add, like, we want to look at Website behavior and then overlay some salesforce field based on like what outreach a sales rep did before Then or something like that and then we want to run an experiment and it's maybe a different field We want to go look into mapping, right? [00:26:54] just Web traffic and web behavior to some first [00:27:00] party data that we have from somewhere that maybe we're just digging and experimenting. GA takes that ability away from us a little bit by, for example, not. De anonymizing so that we can't merge, um, directly on like who this lead is or what, what they did in other systems or what it looks like down funnel. [00:27:19] And so to me, that's a huge, huge drawback, particularly, I mean, I've worked most of my career in, in startups. And so particularly in startups, when you're digging a lot, you don't really know what you're going to find. It's really hard without knowing in advance kind of what, what to do. Um, I also think that GA like It's really hard for people who aren't marketers to get value out of GA. [00:27:45] And so when you have the other question to ask is like how many quote unquote like B. I. like tools do you want to have within an organization, right? So you have your traditional B. I. tool like. Like a tableau or like a [00:28:00] looker, et cetera, that sits on top of a data warehouse where you have all your product data. [00:28:04] You have all your marketing data, all of this stuff. Hopefully the data team supports you, or you have an analyst like we discussed earlier. GA is kind of like a BI tool, but specific for marketing. But if you think, well, sales has their reports in Salesforce. And product might have the reports and mix panel and but what about when, like, we have conversions where we have company goals that we then have to align between all of these tools to make sure that our conversions are the same and we think about them the same way. [00:28:33] And so to me, that sprawl is very detrimental because we spend time talking about, well, you have five conversions on this on this day and I have six and really it's like a tracking problem. And why are we doing that? Three or four different times for however many flavors or, you know, team specific, uh, BI, basically BI tools we have. [00:28:54] ​ [00:28:54] Darrell: [00:29:00] [00:30:00] [00:31:00] Yeah, I, I, that really resonates with me and I also think that, that, um, my theory is that Google analytics really shines with maybe SEO on Google as well as like Google ads, but increasingly those are becoming less effective, especially with, you know, products and services being discovered on other channels and other, channels. [00:31:33] Uh, types of, of advertisements being more, um, More effective. So for example, consumer is heavily moving toward like Instagram and TikTok. And then with with B to B, the search volume is just so low for especially if you're, you know, marketing a startup that I almost feel like the percentage of time that you spend in Google Analytics is it's not really worth it. [00:31:56] It's not really worth, worth the return. And, you know, not to mention we can go [00:32:00] down the rabbit hole of, you know, AI being most of new, new products and services being discovered through, through like AI. So I think that, that in general, um, there, unless you, you're in a very sort of like search friendly category. [00:32:16] Um, you're not really going to get too far with Google Analytics and you, you'd much, your time would much better be spent on either like product analytics, optimizing your website and, you know, gathering all of the data from all the different channels that people are discovering you on, um, and bringing that together in a data warehouse and then on, and then onto, you know, like a data visualization tool, like a tableau or power BI or something like that. [00:32:42] Phil: Yeah. I'm curious to get your enterprise take on this, Darrell, because I resonate a lot with what Sarah said from the startup world. Like there's, [00:32:49] Challenging the Notion That There Must Be a Single Source of Truth for Conversion Data --- [00:32:49] Phil: I've been part of so many debates on where should that conversion data live? Is it in Amplitude? Is it in Looker? That's sitting on [00:33:00] top of things? Is it okay that it's an iterable or customer IO, or should it be in something else? [00:33:06] Like the sales team is fighting for CRM being the source of truth, but the data team is pushing for all this work being done in snowflake and end redshift. Like, how does that work for you, Sarah? Like at Prefect, like being. A person who's on the marketing team now on the growth team, you get a sense of say into what is going to be that endpoint. [00:33:27] Is it for you like the BI tool that sits on top of the warehouse? Cause you just kind of said that the warehouse should be the source of truth for the whole company. So is your Tableau and your Looker sitting on top of the warehouse and it's getting product analytics data and web data and conversion event data? [00:33:44] Like what are you using for that use case? [00:33:48] Sarah: Yeah, I think the question of what's the source of truth is really important one, because I do think that Kind of imposing the warehouse on everyone is also it's going to be like moving [00:34:00] mountains And so it's gonna it can't it it might be hard, right? Um, and so right so for example in sales, there's a rhetoric that frankly I fully understand Um right within sales teams, which is if it's not in salesforce, it doesn't exist. [00:34:15] Um, and I get it right if that's the tool that you're in that's the tool that you're in and why would you You Why would you want to switch tools? You don't you don't want that. And so the way that we've approached solving that at prefect is making sure that we have a view of very sales specific metrics. [00:34:34] So around right number of leads or pipeline or opportunities or meetings or something that's very sales focus in sales force, but being very, very We want to be clean and careful to be able to reproduce that within our BI tool so that, right, if it's for like an AE or a director of sales, right, who's literally just in Salesforce, it's there. [00:34:59] But then if we want [00:35:00] to tie that information with broader things that we want to look into, such as how did people become leads? Like how did our traffic, our website traffic, um, influence our lead count, right? Those are questions that are made more relevant for marketing and made more relevant on, like, outside of just one team, and it's really a funnel question. [00:35:21] And so that's kind of how I think about, um, what the, what, what the source of truth might be. [00:35:29] How Machine Learning Teams Approach Personalization Differently From Analytics Teams --- [00:35:29] Phil: curious to ask you about, uh, personalization because this whole like one of the main use cases of doing all of this work, getting most of the data in the warehouse is getting the source of truth, like a full profile. Most people call it like 360 view of the customer and usually like hyper personalization or personalization is like the outcome that that use case of that. [00:35:53] What does that mean for you exactly? And like, maybe walk us through examples of what you were able to accomplish at [00:36:00] Prefect or something that's on the roadmap, cause I know you're, you're in the process of, of building out that, that warehouse native stack, maybe chat about that for a bit. [00:36:08] Sarah: So with the risk of sounding too vague, um, I'm going to start, I'm going to start with something high level and then we'll just, like, unpack it until we get to something a little bit more tactical. Um. To me, personalization means reaching the right people, so who have a specific problem, at the right time, when they're also ready to solve this problem. [00:36:29] And usually, right, um, the, the flavors of that are, right, about the product, and it's a big, you could take that as, Something around product positioning or right problem aware versus solution aware and at what, or right in the sales sense of what, where are people at in their buying journey. But really, I think that every, every product right at some point, a startup starts with one very, [00:37:00] very narrow ICP where it's just one type of person trying to solve one specific problem in one specific way. [00:37:07] But pretty quickly that expands and it becomes two it becomes three it becomes multiple and usually you know Hopefully that doesn't expand too far where you're still Where you're still you have you actually have an ICP you're not just marketing to everyone but that grows quickly with with the growth of the company and What that really translates to from a marketing perspective is being able to understand where your people are and what they're doing, who fit in this category, who are the right people and what are the signals that they're exhibiting when it's the right time when they're ready to kind of solve this problem that they inevitably have, right? [00:37:45] Because that's like the crux of the company, uh, existing in the company messaging. And so at Prefect, right? I mentioned Common Room earlier. That's a tool that we use, um, to really understand what are people doing out in the ecosystem. It could be [00:38:00] with monitoring our own ecosystem, whether it's like an internal community or an external ecosystem like LinkedIn or like a third party, right, uh, platform in terms of we have some philosophies and theories, um, on who the ICP is, identifying them and then matching that with signals, but that doesn't, um, Go anywhere unless right. [00:38:22] Just saying here's an ICP with these signals like that doesn't mean too much to the sales team, right? There needs to be a lot more context. And so I think that needs to be matched with how can we as a company? So we talked about developers like hating marketing. And so I think that needs to be matched is how can we as a company right in our outreach be helpful to that person at that time with what we know. [00:38:48] And so that to me is what personalization means. It's about being helpful and being hyper targeted in that help where it's not, we're helping everyone the same way and here's five different things and hopefully one, one fits in terms of [00:39:00] case studies or, right, docs or, um, content or whatever. It's really about, hey, this might be really helpful for you, or I hope it is, or whatever, and being very hyper targeted. [00:39:10] For prefect, for example, um, one of the kind of, I think, canonical use cases is around The like ML team versus the, um, analytics team. So just to dive into that, um, a little bit, so Prefect were all of our users primarily use Python or have used Python and use Python to deploy data pipelines or any sort of data related work. [00:39:38] For analytics teams, it's usually some sort of ETL process, right? You extract data, you load data, you do something with it. It's usually warehouse native. For an ML team, it looks completely different. It's, they have data and they're building models or training models or testing models and the requirements around how those models get trained and how they're [00:40:00] run on infrastructure GPUs sometimes depending on the field that you're in, right? [00:40:06] Um, You also, you just have a completely different view of the world versus a data team that's actually not working in Python, for example, in Prefect's case, using mostly SQL, but needs some Python to just take their work into production and maybe do some like random ETL processes. And those two use cases are fundamentally different. [00:40:27] And so if ML engineer versus someone who's right, just doing a lot more stuff in the warehouse, we're going to talk about our product a little bit differently because. The problem is similar, and the solution is similar, and it's still in the realm of what we're doing and what we're trying to solve as a company, but they're just two different personas, and I think it's really, that's really the work with product marketing to define what those personas are, how you reach out to them, but also, right, the work in growth and the work in MarTech is having the tooling to identify and being able to properly [00:41:00] segment on those personas. [00:41:04] Darrell: Nice. Nice. I'm going to, uh, break up the next question into two. So in my notes, I wanted to ask you that [00:41:12] The Venn Diagram Overlap Between UX-happy and Customization Tools Is Nearly Empty --- [00:41:12] Darrell: you've said the Venn diagram overlap between UX happy tools and customization happy tools is very small and non existent. So I think the first question is, what do you mean by that? [00:41:24] Sarah: So what I mean by, uh, customization, happy tools is really tools that like expose the innards of what you're doing. So I'm a really big fan of a tool called pipe dream. I take every opportunity I can to like just shout from the rooftops about it. It's basically a Zapier competitor, but I can't even put my finger on why it's better. [00:41:45] It's just like, it's just amazing. The controls that you're exposed are like. For example, dealing with rate limits, if I'm trying to, like, enrich my CRM with, like, from Apollo or something, it's very easy because I [00:42:00] can limit how many times a particular workflow runs within a second. And so then I can very easily say, okay, well, Apollo has a rate limit of, I don't know, I don't remember what it is, let's say five requests a second. [00:42:10] It's very easy, right, to control, to control that. However, right, with power, it's comes responsibility. And so there it takes time to figure out how do we actually hook everything together and to test it. And you have to be a little bit deeper in the tech to be able to do that successfully versus UX tools, which I would maybe put Zapier in, um, a little bit sometimes, right? [00:42:35] There are probably better examples, but where you say connect to Apollo, run this go, and you don't need to know what that is. In that case, right, if you don't need to know what that is, it's probably a good generalization and the UX path in that case is probably a good, um, a good layer and a good abstraction over the things you [00:43:00] probably don't really want to spend time thinking about. [00:43:02] However, um, there are times when, I'll give another example, um, Account matching in Salesforce. Um, sometimes you want to, um, Only match accounts that are like only consider accounts existing if they're assigned to someone that works at the company today, right? as opposed to just existing in salesforce in general and so A ux happy tool might not give you that flexibility and might say do you want to match the account? [00:43:31] Yes. No, give us the website and then we'll match it or give us the email domain or whatever and we'll match it without saying Okay, well filter on the account Existing, you know, or having activity within the last six months or being assigned to someone that works in the company or whatever Versus in the custom is in the customization camp. [00:43:49] You'd likely have a lot of those controls, but just need to know How that information is represented in salesforce and be able to kind of know your criteria a lot better [00:43:59] Darrell: Interesting. [00:44:00] Interesting. So it sounds like. On one hand, there's very customizable solutions that you can do, and you can build maybe a more robust setup and, uh, solve for, Um, you know, these maybe long tail, uh, uh, problems, uh, and that's, that's on, that's on one side and on the other side, you have a very easy to use UI that can solve the problem for you, but you're not going to be able to get very far if maybe something unexpected happens, or is that kind of, uh, what you're saying? [00:44:34] Sarah: Yeah, exactly [00:44:36] Darrell: Gotcha. Gotcha. And then, so why, why do you think that? When we purchased Martek or we set up Martek that we, that we're pretty optimistic about, Hey, everything's going to go the way that we want, we're going to be able to solve for this solution. And how do we kind of better set our expectations? Like, do you, do you have experience with that? [00:44:56] And, and what do you think? [00:44:58] Sarah: I think there are really two [00:45:00] things that we need to think about the first is You Who are we optimizing for when we think about a tool? Who's going to be using it, right? Is it going to be, um, An A. E. Or S. D. R. That really doesn't care about the tool at all, but just wants to, like, set their vacation or just wants to look at some account or whatever, right? [00:45:19] Or is it going to be, um, someone that's one layer deeper and setting up the workflow or really cares about account matching and all the rules and being able to, uh, granularly match accounts, right? Is that a business requirement or not? Um, and if the answer to both is yes, then we need to be able to both appeal to this more business persona that's just trying to get the job done and also these technical requirements. [00:45:48] And I think we're overly optimistic about trying to do both, but what I have found time and time again, and I am generally an optimistic person, and I have gone multiple [00:46:00] times with these very optimistic, ideas about having one tool and one approach kind of fit both, but it very, very rarely does. So that is, I think the, the expectation, the expectation that a tool will fit both is, I think that's just an unrealistic expectation is what I've kind of encountered time and time again. [00:46:23] Um, and so the way to solve for that is making sure we have the UX happy tool, right? For someone that just. They just want a, not to have a steep learning curve, they want it to be intuitive, they want to be able to update a certain setting or just view a certain report or whatever, but then having that integrate with some parallel process or some process that supports this tool to be able to add all the levers for a smaller group of people usually that are changing those settings, but not exposing that to everyone because it's going to be information overload. [00:46:59] Phil: It's been a [00:47:00] super fun conversation, sir. I love your, your perspective on. These like growth marketing ops topics, but from a much more, like you just said, like sales folks using a tool, but there's like a layer lower of things that are like more technical, but like things that like other people care about a bit more, like, I feel like you're. [00:47:21] Like marketing ops, folks care a bit more about certain things than your average marketer, your average salesperson, but from your data practitioner experience, like you even care more than a marketing ops person about some of these technical things. Cause you built the data pipelines before you've architected data teams, you've ran data teams. [00:47:41] So your perspective on this is super, super cool. We had two last questions for you. Um, we can't get through a podcast without. Chatting about, uh, ChatGPT or Gen AI stuff. We're like a few years into this like sea of content written by ChatGPT and Claude. And it's like very apparent to [00:48:00] most when like something is written by AI. [00:48:02] I don't know for you, but like, you know, it's shallow. It's filled with like similar words. Uh, GPT loves em dashes every couple of sentences, but it lacks personality. And most importantly, it lacks a ton of context. Right. And I'm really close to this personally. Uh, we repurpose a lot of our, uh, podcast episode transcripts into blog posts versions. [00:48:22] And it's, it's all chat, GPT driven stuff. I do a lot of editing and catching like hallucinations and bad transcriptions, but like, we're not asking AI to write something from scratch about. Someone that became a marketer from a data engineering background. And we have a conversation and with an expert, we ask a question and through specific prompting, we ask tragedy to turn that blog posts, uh, to turn that conversation into a blog post passage and have like a specific style around it. [00:48:51] So we're giving a ton of context and the output is way different than if I was to say, like, You write me a blog post passage [00:49:00] about Sarah becoming a marketer from her data engineering background. Um, but the, this whole ramp up to the, the question I want to ask you is like [00:49:08] How AI Mimics the Limitless Drug for Content Creation --- [00:49:08] Phil: when it comes to writing content specifically for your background, like creating content for a product that you're selling into developers, um, like you said that you struggle. [00:49:20] Both with agencies outsourcing this to agencies, but also using AI for this because neither of those are embedded in the product itself. Maybe unpack that for us. I'm really curious to get your take there. [00:49:33] Sarah: So I think there's a fundamental question of what would you outsource? It doesn't matter whether it's to an AI tool, to a contractor, to an agency, right? And I think, right, there's, so for example, would you outsource a product marketing role? No, you wouldn't, because that role is actually, it's mostly, people work. [00:49:52] It's making sure people are aligned on your product division, on your messaging, on your positioning, on your personas, and then building [00:50:00] context and disseminating that across the company. And it's really an alignment role. And there's a lot of people work in that, right? But would a lot of people contract out, um, doing like running ads? [00:50:15] Why is that? It's because there's a lot of technical work that's very similar across different companies. And so I think there's a first question. There's two questions here. The first is what would you contract out? Anything that you wouldn't contract out, you should be very careful using AI for, um, to, to, as a, for a generative purpose, as opposed to summarization, editing, whatever, right? [00:50:35] Like that's a little bit, that's a little bit different. I think there's a, a second question, which is how, like, if you go down the path of using AI, right, um, how should you interact? With Chachi, BT, or Claude, or whatever, right? And this phrasing I'm about to say actually came from our CEO, Jeremiah, um, who's really into the AI space and very bullish [00:51:00] on AI, rightfully so, and I agree with him. [00:51:02] And the way that he phrased it made so much sense to me. And it was, think about an LLM as an intern. Would you, would you give an intern That's a company, right? Like the 19 year old who's like transitioning from a freshman to sophomore year in college. A really open ended question where they're supposed to be creative about and have all this context that someone who's 20 years their senior has on the space. [00:51:31] No, you wouldn't. And then, furthermore, even if you give them a well scoped task, would you give that output to your CEO without reviewing it first? No, of course you wouldn't. And so that, I think, is the approach that we need to take with AI right now, where it's not, it's not good enough to be, like, a tool. [00:51:50] a staff marketer, right? But it is good enough to take some context and, um, do some manipulation, right? [00:52:00] And put it in different formats and then have someone senior with the context edit it. And it's still a lot faster because it's a lot easier to do something when you're not starting from zero. It's like, you're not saying paint me a painting. [00:52:12] You're saying, Here's a stencil, right? Fill it in, um, which is a very, very different, um, which is a very different problem and a very different approach is, is a lot more, um, is a lot more feasible for where AI is today. There are two caveats. The first is, um, It's like technology is changing so quickly. So I don't think that this is always how it's going to be, but I also think that there's, um, there's right. [00:52:38] And then there's the, the second caveat, which is. It all depends on how you set it up. So I personally really like Claude for setting up context where you can add, right? So for example, like I can add a bunch of blog posts, or I can add a bunch of LinkedIn posts, or I add our company positioning and you start to build that context. [00:52:55] And so you just get to that point, um, a lot faster of building context, but [00:53:00] we're still not quite fully there yet. So, um, that's kind of my take. [00:53:05] Darrell: Yeah, I, I had this insight and, um, Before, have you, about, about J. I., Gen A. I. and, and its, and its use, have you both seen the movie Limitless, by chance? [00:53:17] Sarah: I have [00:53:18] Darrell: Phil has, okay, [00:53:19] Phil: that the one where they have like a watch for like how long they have left in their life my [00:53:23] Darrell: so, Limitless, Limitless is about a pill that you take. And it increases your intelligence. Like, several magnitudes. It's with Bradley Cooper. Anyway, it, anyway, in the movie, The, the core, one of the, one of the key insights is, The pill works better if you're already smart. Like it, it makes like an average person pretty good, but it makes a really smart person, genius level. [00:53:51] And I feel like that very much, in my opinion, um, applies to AI. I think that if you have a good [00:54:00] strategy, if you're, if you, if you, if you, if you're thinking about your customers the right way, if you have a good product offering and you know, you're approaching customer engagement and messaging the right way. [00:54:10] It can get you really far, but for most people, they don't have that. So AI will just help them write their content a little faster or generate the reports a little bit faster. It doesn't take you where you want to go. Um, anyway, so, so I just wanted to share that. Um, and I'll now move to the last question. [00:54:32] Phil: you get the honor of this is the first time you get to ask that question, too [00:54:36] Darrell: I do. I am honored. I am honored. Okay. So [00:54:38] Finding Work Life Balance Through Outdoor Adventure --- [00:54:38] Darrell: we asked this of all our guests, uh, Sarah, you're a marketing leader, a data practitioner, a newsletter, off newsletter, author, conference speaker, and also a renowned TV show binger. Yeah, me too. Ski expert, avid swimmer. Um, so our question is how do you remain happy and successful in your career? [00:54:58] How do you find balance between [00:55:00] all things you're working on while staying happy? [00:55:04] Sarah: Balance, I think, is really the key to life. And so for me, um, my job is in front of a computer and I'm inside all of the time. So I try to balance that by going outside, um, whenever I can. So we're recording this right in winter. I ski a lot. It's about to be my life. And basically all of my life outside of work for the next five months, which I'm very excited about. [00:55:27] In the summer, um, I live in Vermont. And so there's skiing and we're right on a lake. And so do paddle boarding, try to get out on the water, but really it's just trying to get that change of scenery. Just trying to get the blood flowing and, um, just getting a little bit, a little bit of adrenaline in there too. [00:55:45] Phil: Love it. Awesome. Awesome answer. I think we're we're all Uh, TV show bingers, but yeah, it's, it's, it's a good balance to go outside. Cause you spend the day on your laptop and then like you spend part of your evening in front of the TV [00:56:00] watching a show. And my wife is also is like, yeah, having a dog is good for that. [00:56:03] Like even in the winter, I'm just like, Oh man, we didn't walk you today yet. Like, let's, let's go. And it's always a drag, but then you're outside and you're breathing the fresh air. And you're like, man, like I need to do this more often. I'm in front of a screen way too much. [00:56:18] Sarah: Yeah. The fresh air just, it really just like changes your whole perspective. [00:56:22] Phil: Definitely. We'll, we'll link out to your newsletter. Uh, sir. I think there's some, um, awesome articles in there. There's a bunch of topics we, we didn't get to cover. We could have, uh, done like three or four more, uh, parts, but, uh, yeah, thank you for, uh, all the stuff you're, you're writing about. I think you're, you're doing a service and helping bridge that gap between growth marketers and, uh, all the data engineering concepts out there for, thanks for, thanks for being here today. [00:56:47] Thanks for sharing your time. Really appreciate it. [00:56:50] Darrell: Thank you, Sarah. [00:56:51] Sarah: My pleasure. Thanks so much, Phil. Thanks, Darrell.