Scaling to Billions: How to Design the Ideal Stack with Natalie Miles Fuerst of Grammarly

"No one's data is perfect. I think the biggest challenge though ends up being more people, because what often happens is that companies of a certain scale, marketing orgs of a certain scale, oftentimes we start adding on products or new verticals and we hire marketers to be in charge of those specific products or verticals, and they work in these silos and isolation of each other." - Natalie Miles Fuerst

In this episode of Making Sense of Martech, we sit down with Natalie Miles, Staff Product Manager for Martech at Grammarly and former Head of Martech at Chime and Credit Karma. She is an expert in managing massive data volumes, driving scalable personalization, and assembling composable marketing technology stacks.

This is a must-listen for leaders and practitioners navigating the complex world of high-volume email, the composable CDP space, and the ever-present "build vs. buy" debate in the age of generative AI.

This episode wouldn't have been possible without the help of our sponsor, Hightouch.

Looking for a smarter way to activate your customer data? See what Hightouch can do for you at hightouch.com/msom.

Highlights

Learn the biggest challenge of scaling audience segmentation and deliverability to billions of emails per year, and why it's a "people problem" before a data problem.

Hear a hilarious and painful "epic fail" story involving a compliance email, a financial services company, and a sex hotline.

Understand the pros and cons of warehouse-native (composable) CDPs versus traditional, off-the-shelf solutions, and why the term "CDP" is fuzzy.

Discover the key difference between real-time and batch data delivery and how to decide what your use case truly needs.

Natalie's "dream scenario" tech stack for a high-volume, modern company, including her top picks for reverse ETL and ESPs.

Learn how to define a North Star metric that aligns with customer engagement signals with long-term LTV and business outcomes.

Episode Breakdown

00:01:35 - Rapid Fire: Night Owl, Favorite Martech Tool (It's not an ESP), and Vibe Coding.

00:03:43 - Professional Admiration: Why Credit Karma's marketing cohort was a special "incubator."

00:05:33 - Epic Fail: The compliance email, the phone number, and the sex hotline.

00:09:20 - Reversing a Major Martech Decision: When a personalization tool isn't worth the price.

00:11:29 - Scaling to Billions of Emails: The challenge of data hygiene vs. the people problem.

00:15:54 - Composable CDPs: Deciding between real-time bvs. batch data.

00:20:20 - Predictions: The future of composability and why "CDP" is a fuzzy term.

00:26:02 - The Age-Old Questions: Build vs. Buy in the age of Generative AI.

00:31:48 - Aligning Customer Engagement: How to define a North Star Metric that drives LTV.

00:35:53 - The Dream Martech Stack: Natalie's top picks for a high-volume, modern comapny.

00:39:25 - The AI Risk: Why you don't want to be the "first adopter" of an AI-powered ESP.

Key Takeaways

The biggest Scaling Challenge is Organizational: At high volume, the main issue isn't data quality but the "tragedy of the commons" where siloed teams send overlapping campaigns, leading to user fatigue and churn. Having an "air traffic control" function is crucial.

Composability Solves for Alignment: The fuzzy definition of "CDP" and it's high cost highlight the need for composability. The warehouse-native approach is "very bullish" because the underlying data should be the single source of truth for both marketing and product/finance.

Build vs. Buy is Fluid: The decision has historically leaned toward buying third- party tools because in-house development often results in difficult-to-use and unreliable tools, diverting precious engineering resources from core product capabilities. However, the emergence of generative AI could dramatically change the "build vs. buy calculation".

The North Star Must Connect to Value: A true North Star metric should be time-bound, align with the natural cadence of product usage (e.g., daily active users for Grammerly)

Subscribe to Making Sense of Martech wherever you get your podcasts. Leave a review, share with your team, and send us your questions and confessions. We may feature you in an upcoming episode. Hope you enjoyed the episode!

Creators and Guests

Host

Jacqueline Freedman

What is Making Sense of Martech?

Unfiltered takes on the biggest shifts in marketing technology. We spotlight what matters, who's leading (or lagging), and what's next. In Martech, clarity is power — and we're here to deliver it.

Welcome to the Making Sense of MarTech podcast, where we interview leaders and put them in the hot seat. I'm Jacqueline Friedman, founder of Monarch and global head of advisory for MarTech Weekly. Let's dove in and meet Natalie Miles first and a little bit about her. Natalie is the staff product manager for Marketing Technology. I grant really where she's shaping our massive volumes of data and messaging drive scalable personalized marketing.

Formerly, she's the head of MarTech and personalization at Chime, which actually just went public. Congratulations. And she oversaw marketing ops at Credit Karma. She's an expert in assembling composable KPIs and building bridges between engineering and marketing teams. She's a vocal advocate for experimentation, first culture, and she even coauthored a chapter on aligning customer engagement to revenue and the customer engagement book Adapt or Die by Not Engage.

So to start us off one, thank you for being here. So excited to have you. I have a personal bias and excitement because we've done the same job at Great and so I'm so happy you're here and we've got a soul sister for life. But get us started. One Thank you for being here and let's do some Rapid Fire two so I kind of get started as I.

Yeah, sounds like a good plan.

We always do rapid fire with each other.

That is true innovation. That is just.

Always in the hot seat at the time, just like what.

Happened. How do I do this? It feels like we've lived the same life we kind of have.

Exactly. And I am so.

Grateful that you.

All right? Are you a morning person or a night owl?

Definitely a night owl.

I wish I was a morning person. Morning.

People like just have this or about them. Like they've got it together. Honestly, I work fast and I feel like when everything's calm and dead and there's no children, interrupting is when I can do like my real focus work. But we're.

Always.

A night owl.

Same here. All right. What is your favorite martech tool that you can not work without and get specific.

So a year ago, I would have just had like.

A Google she like the joke of like what's the most common.

Use by a marketer a CSP fail honestly chat up or clod use it for everything that's on today. I use it to outsource my thinking and decisions, but using it as a thought partner and basically treat it like a really brilliant intern who is also and if naive and dumb about something so very much still like believe in human in the loop and you know, like we need to be vibe checkers but yeah probably far and away most used tool.

Okay speaking of vibes, vibe coding, yay or nay.

Nay on the term, it does feel a little cringe and.

Like embarrassing.

To say.

It. But yeah, like it's oddly addictive recently. I'm really loving, lovable, honestly. Like it's just so fun to like I've been vibe coding coloring book app for my kids. Like there's so many like applications and the ability to, you know, get out of your own head and execute on something that has lived in this like amorphous space forever is so gratifying.

Like, it.

Is just fun.

Like, I can't remember the last time I've had fun with school before.

I completely agree. And it's also like the playground in which you make and so you get to have a lot of fun there. I'm excited. There's a couple I think lot announcements happening in a couple this week.

Next week. Yeah, yeah, yeah.

So looking forward to it.

A space. Yeah.

Excited for Elena for nice touch. All the things. Yes. In particular. Okay, last rapidfire question, who is someone you admire professionally or personally?

This is a difficult one. I sort of subscribe to.

The like never meet your heroes because like the people who I admire professionally often don't have the personal life I admire and then vice versa, you know, obviously. Jacqueline Friedman.

Okay. Like, let me just say, like, yeah, I. I am so eternally grateful.

Everyone, you know.

Like how wonderful.

You are to work with and like, again, thank you eternally for like setting me up for this role I grammarly and, and like giving me the ins and outs. Like, I always trust your opinions. Like, yeah, like, thank you.

Right back at. I wish I actually got to work together.

So we got to work together on.

The education piece of how to do two things.

Yeah. Two ships passing each other night.

You know, other people. I would say I admire professionally. I had the privilege of working with some really great marketers at Credit Karma. In some ways it spoiled me and made me think marketing was something that maybe it's not.

Just like such an.

Amazing credit. Karma circa 2016 to 2020 was just like a really special marketing award. Like we had an amazing CMO and Greg well, like really amazing marketing leadership. The amount of CMO's that came out of that credit, karma marketing cohort is really impressive. And so yeah, like learned a lot. It was a really great place to really learn marketing and like what marketing should be like, very data driven in my mind, very low ego, very experimentation.

Focused is like it sounds like an incredible incubator. Ultimately it.

Was. Yeah.

That's awesome. All right. This is my personal favorite question to ask whenever I'm hiring. And it immediately tells me if it's someone that would work on my team or in the greater team. So what is your biggest effort? Because I've got plenty, so let's hear it.

Yeah, there's like efforts.

That were, like, very valuable learning.

Opportunities, minds, just kind of things that were so funny or better. Oh, gosh.

So I have recently joined Chime and at the time, you know, as you do, and it's sort of like a scrappy environment, you wear many hats. And so I was also wearing a lifecycle marketing hat and being in the financial services space. You know, we send a lot of transactional emails, especially, you know, change in terms type emails and so like every lifecycle myself.

Little or just.

Like we.

Never want to send those emails but we got to. And so like in the habit of like compliance or legal comes to you with this boilerplate change in terms agree that you need to blast everyone with like sometimes you got to see it up. Sometimes, you know, you just ship it out the door because you got more important things to do and no one raises.

This is a situation where.

Like, yeah, we had to set it up and then we sent it to our entire addressable audience and it turns out one of the phone numbers in there, because you always have to have like some number and these compliance emails it went to not chime but a sex hotline.

And.

Oh my goodness.

And that fortunately. No, I'm good. What do you mean? It's like, oh, sure. Well, one person did okay.

And that person emailed you immediately.

And so it turns out like and to be fair, this was also posted on our website and the compliance person like, you know, oh no, the Q way failed all across the board. It was sort of like game of telephone.

We were like, you know, we got it from one place and then handed over to the next person. Then they handed over the lesson learned when you're queuing your emails, also.

Queue the phone numbers and make sure that, yeah.

That's amazing. And also a really great case for why you should never have a no reply. That isn't un not monitored and it's not a real email. Yes. How quickly did you guys discover the situation?

So, I mean, it's not like you blew up on social.

Media or anything.

But yeah, it was certainly embarrassing when we found out.

That's fair. Also, just like such a great reminder of a as a team is for.

Right yeah the retro on that was in like you know there's always that temptation that want to point fingers and blame everyone else and.

No one wants to take ownership.

Yeah. Like lifecycle wanted to blame compliance. You know, like compliance was like, well, you know, I'm the fifth person.

Who.

Was in charge of this. And, you know, at the end of the day, like last mile, we're the ones sending out the email. Like we should take responsibility and make sure like all the I's are dotted and t's across.

I agree with the caveat of receipts.

Yeah.

And the rules of.

There were a lot of eyes on this. Before we click send.

I think that like the other valuable part is like of course this was a rush job. Like all these like change in terms emails often are that's where all.

Of them happen, that's where the mistakes happen.

And so like, you know, there's always the opportunity cost of like, you want us to ship this quickly, mistakes are going to happen. And so I think that the real outcome was like we need at least 24 to 48 hours for you know, decent Kua.

That at least was a good lesson learned. Yeah, some good came of it and maybe some extra traffic to.

Sort.

Of open some people's eyes.

I wish I remembered where it went to.

Oh yeah. Maybe quite the plug.

Yeah.

Speaking of mistakes, I don't think there's any mistakes. I think things are always learning opportunities in some capacity, and that's not me being positively optimistic. But have you ever had to deal with reversing a martech decision that was major? And what were the signs? What were the reasons and what did you learn from it?

Yeah, I mean, I think that that's the the nature of the job. I mean, like we've all broken up with thunder. Is it because, like, we all get told promises that don't pan out in reality? I mean, salespeople lie.

I know. I mean, wow.

Maybe they don't necessarily lie, but everyone's, you know, business model is different. Everyone's workflows are different.

Like it stretches the truth in different ways.

Yeah, yeah, yeah.

We all have different needs. Recent situation, like we had a breakup with a vendor. I mean, partly some of this is, you know, on me and the or again understanding what the strategy is. So like we're going all in on personalization. I think personalization and always kills me a little bit. But you know, from a strategy standpoint, what exactly do we want to do.

With.

Personalization? And we were trying to move so fast we onboarded a personalization tool to personalize copy. We successfully implemented it. At the end of the day, did it have an impact on growth and metrics we care about? Not really, because it turns out optimizing a subject line is only going to get you so far and probably what we should have done more due diligence on was understanding the underlying pricing model because when it came time to renew, it just wasn't very scalable to our life.

Wasn't there. And so, you know, we spent all this year like developing this tool, integrating it into our workflows only to find out, you know, it's not really impactful and this isn't something we.

Can continue.

To invest in. So I had to pull the plug. And also it turns out coffee is very commoditized right now and it is your tool is only doing that and is a single use SAS product. Like, you know, how much remote?

Yeah, there's a lot of breaking down of notes lately. Some two very interesting approaches to say, yeah, all right, I want to get into nitty gritty because you, I think, in my opinion, are one of the very few has overseen extremely high volume sending not just Grammarly, but at Chime and credit karma. And there's very few people who've been able to navigate and also succeed in that space.

What is the biggest challenge you've experienced or you see when you're scaling in audience and segmentation and deliverability at such scale where we're thinking, I'm going to say at least a starting line of at minimum 5 billion emails a year. I'm scared. Yeah, we have worked up our numbers. I'm very curious also.

Yeah, I mean, depends which.

Channel to like. The lifecycle scale looks very different from a paid marketing scale where we're often like measuring that in terms of and there's always, you know, the martech line like always the need to get our data house in order. I mean, that's even more true when you're working at scale where, you know, you've got large marketing teams and they often become siloed definitions or metrics, underlying business logic change, and then suddenly, like, we don't know what data we're activating.

I don't think anyone can relate to that at all.

Right. And so this is a.

Very universal problem. Like, I mean, maybe if you're a small, scrappy startup or a slow primary, you know, exactly like every single field and like the two.

Tables you have.

But like for most of us, like, you know, data hygiene and standardization is usually a challenge. Data activation becomes that much harder when you're working with massive amounts of data. You've got thousands of tables, thousands of events with unclear taxonomy or standardization. So that's certainly like one of the challenges to solve. Or I will say though, like there's also a bit of a myth too, that of like having perfect data.

Like no one's data is perfect. I think the biggest challenge though ends up being more people, because what often happens is that companies of a certain scale, marketing orgs of a certain scale, oftentimes we start adding on products or new verticals and we hire marketers to be in charge of those specific products or verticals, and they work in these silos and isolation of each other.

And so take the lifecycle example. We're still like oftentimes emailing the same addressable audience across the board. So let's say you got a 20 or 30 person marketing team. What ends up being the challenges, all of this overlapping targeting and how do you prevent things like email fatigue? Because we're sending, you know, a product offer, for one thing, to the same user and a different product offer because, you know, from an organizational standpoint, we've just decided to, you know, this team is in charge of driving this outcome for this product and this team is in charge of driving this outcome for this product.

And we don't think about the tragedy of the commons that happens when we email all of our users too often. So a big part of that is like, how do we solve for this organizationally, I do think, especially in the lifecycle world, there's a huge need for having that sort of third party referee, whether it's a marketing ops person or whatever title you give it.

Their job is air traffic control from a manual standpoint, whether it's and this becomes very hard to do when you're doing hundreds and thousands of campaigns. But understanding, like at a very high level what's in flight and sort of like very rough targeting. And then I'd say, like, you can start to do more programmatic solutions on top of that.

And so that's where things like ML Solutions become very valuable. So kind of karma, this was a massive problem we had where we had a lot of different verticals. We were sending billions of emails often to overlapping audiences. And so one thing we found was like, hey, even just having like a personalized frequency model, how do we, you know, everyone's tolerance at a11 user level for how many emails they want to receive is going to be different.

And so how can we make sure we're sending the right number of emails to the right user so that we prevent churn? And yeah, like that was not to say it was an easy solve, but this was like circa 2017, 2018 before it was super hot. And so like these solutions have been in place for a while now.

And so yeah, it's a combination of, I'd say manual people operations as well as more programmatic, often ML based solutions are sure.

Okay, aside from email, you're extremely well versed in the world of CDP use and I'm a big fan of them as well. And so we were actually first introduced by our mutual friend Helga Marsh of the Humans of MarTech podcast and I was also a guest. And so he's just the best human special shout out to him. We love our Canadian friends here in the States.

In in your episode you did with him, you championed warehouse native, composable CPUs, but also noted latency issues one let's let dove into that and how do you decide what needs real time delivery versus batch processing. And also, I know pretty much every CDP CDP is working towards that real time quote unquote. I know I still don't fully believe it.

And also, I'm kind of a skeptic in general, but I want to hear all your thoughts.

Yeah. I mean, you know, I would say it depends on the use case and it depends on the channel. So take lifecycle as an example and specifically email. It's a very asynchronous channel and so there should be like a really compelling business case for why we need real time data and especially controlling for we don't actually know when the user is actually going to open and engage with the email.

So, you know, even if we do manage to trigger this email in real time, are they going to open it and see it in real time? Maybe, maybe not. If you've also figured out some time optimization, maybe, maybe not. Obviously, there are some use cases where you do want your real time like password resets or sign up confirmation emails, abandoned cart, probably a good thing.

But for things like engagement or resurrection.

And especially.

Think about like why we're sending an email to begin with. Often it's you know, if we think about like resurrection use cases, right? Like they're already outside of the product. Like are there other channels where we're already interfacing or engaging with? If not, we're doing sort of a channel waterfall approach and then resorting to email like, you know, maybe it doesn't need to be real time.

The other part too is if it is truly based off of a behavioral trigger, those tend to be better near real time use cases than like, Hey, we need to send like this nurture series or this win back campaign. Like, what did they actually do to initiate this? You know, the other thing I would say is from an orchestration standpoint, like where is the campaign orchestration happening?

So for lifecycle, you're probably using some ESP like a breeze or iterable to send out.

The campaign as.

Well. Not all of that data needs to come from your warehouse via composable CDP like you can also real time stream events. But the challenge there is it is then often hard to enrich those events with additional customer attributes that you might have in your warehouse. And this is where I would love for someone.

To solve for that. Like, Yeah, you can change the world.

Please create a solution here. Natalie and I will be very happy.

If High.

Touch, I think, has come out with something recently. I haven't had a chance to investigate it. I do think like this is top of mind for a lot of people. Like how do we solve real time personalization? I think the first question is, do you actually need a software real time personalization? I do think there's a ton of need, especially on web and in product experiences and the solutions there I've been pretty unsatisfied with so far.

Underwhelming to say, yeah, yeah.

And like it still gets you in this very like segmented persona based model of personalization. I'd say it's still very hard to do that. Like we all aspire to that, like Netflix level personalization. Like you log in and you see, you know, movie recommendations and even the actual content cards are like dynamically created to, to like engage with you and based off of who you are and your history like that is very hard to do.

And I don't know if there's like any good third party tools that can deliver on that just yet. We'll see. I'm sure like someone's working on that. But yeah, I mean, depending on the channel, depending on the use case, warehouse native gets you very far if anything else, like going back to data hygiene and data standardization, it solves for a lot of that versus a lot of the traditional off the shelf.

CDP is where you're just recreating data that already exists in your warehouse. It's also very expensive because you're paying for that additional storage costs and then of course, like you're actually going to end up with like misalignment. You know, what is the data source of truth? Is it in that third party CDP, or is it in your station tables, in your data warehouse somewhere?

So like from that standpoint, I think the data warehouse or warehouse native approach solves a lot of those problems.

I agree.

It will be interesting to see if someone is able to resolve like how do we do real time and then streaming and enrich it with the data that's already in your customer 360 or data warehouse. So someone please solve for this. I'm waiting.

I have a feeling it's coming. And so kind of in that same vein, speaking of predictions within a composable space, where do you see the next 2 to 3 years going, especially now that our decisioning is more of a prominent option? I'm curious what your headset I mean.

CDP.

As a category space.

Has been very fuzzy in general. So first off, there's been this trend of like everything is converging into CDP. It's like.

Everything's a360.

AI.

Everything's trying to be a CDP, but going back to it, like I think Composability is more important than ever, especially in the world of Agent IC AI and with software development costs going down to, you know, quite cheaply.

Like.

The build versus buy calculation has probably started to shift quite a bit. And so why would you buy an off the shelf solution with all these bells and whistles that you don't really need? So the Composability piece is going to be more important than ever because people are going to have more opportunity to pick and choose the things they actually need for their business needs.

Composability will continue to be a trend.

Everything is going to converge into a CDP. I kind of hate CDP as a category space because like personalization, it means everything and nothing. At the same time. You talk to data engineering their concept of what a CDP is, it's very different from a marketer going back to why Composability is important is like when we talk about CDP and you break it down to its core components, you know, often it's maybe data ingestion, data storage, data transformation, and then data activation like which CDP service are we actually talking about and honing in on going back to the like, what is the actual CDP?

Is it your warehouse? Is it your data activation tool? It's just like a very confusing name and jargon that we don't need. And as tools like ISP's become CBP's like what does it even mean to be a CDP at this point? Like which part of the data value chain and data orchestration and activation value chain are they solving for?

Is it that the data transformation part? Is it data activation? So yeah, like again, composability I think will be more important because what we're really trying to solve are the various different services that CDP is are meant to solve for one.

Hundred percent agreed in that one. CDP is super fuzzy. It's always been fuzzy. The first time I encountered CDP as a concept and then did a demo with a very large vendor at the end of it, I still had absolutely no idea what was said. Yeah, I was the most technical person on the call. Yeah, and that's not a good look when you can't even explain what your product does.

Well. And so whenever I'm talking to folks about CDP, I always have to level set and ask, What are you thinking of? Yeah, I think of CDP because often times it's a pie in the sky concept or it's truly just one singular use case and they don't need a CVP. Yeah, but it's all so dependent. Yeah. I mean.

Even from like a B2B sales standpoint, like who's their ICP or ideal customer profile? And it's hard to solve for when it's this like amorphous bundled solution. Is it data engineering? Is it marketing? It's very hard to sell marketers, especially in scrappy or startups like the value add of the CDP from like a campaign owner or a campaign operator.

Like what is this actually solving for me? Like, it's actually like very difficult to sell.

It is. And that's where I've found the most success. Whenever I've had to build business cases or champion on behalf of folks, it's hey, this is a tool that's not going to just serve marketing because I think any and everyone knows the feeling of I hate the concept of like the marketing stink where it's just like, oh, it's just marketers sending pretty little emails over these things.

No, in reality, we're actually making transformational campaigns that can really do something and move the needle. And we're the largest are a Y bringer of things like ACH is spending all the money to get folks in, but we have to keep them as receipts. Okay. And so I found great success where not only are you supporting and empowering marketing, you're empowering sales, product engineering.

I was it all the mixture of variables and ingredients are different at every company and depending on your business model. But the more you're able to make a tool multi-purpose and a Swiss Army knife, not only is the business case easier, but the understanding of how everyone's going to use it in a different way, but in a singular same way at the same time, makes the sell a lot easier.

And I think that's the missing component for both internally selling, but also externally with vendors too.

Yeah, that's why I'm also very bullish on the warehouse native CDP or composable CDP approach really because like you look at a lot of the off the shelf CDP is they're very much in service of marketing use cases. If you think more broadly, data activation is not just the purview of marketing. The same data that powers all of our buy dashboards that finance uses or product uses like that should be the same underlying data we use to build audiences in marketing.

And that should all start with our source of truth, which is in the warehouse.

I mean, we should have a standardized understanding and definition of everything across every department. What do you.

Mean?

I mean, that might be wishful thinking. It's a good and aspirational. It's aspiration. And what we got there, I don't know.

I mean, astride is strike. You know, you got to take five steps forward to get four steps back, but you're still a step forward. All right. Shifting gears slightly, you kind of mentioned about building versus buying in this current sphere of potentially bad coding or just reduce engineering costs. It's my favorite age old question, but I'm curious from your perspective, because I know you've worked with in-house homegrown solutions as well.

You invest internally or do you go with a third party? Like what is your approach to deciding when to pivot between one or the other? Very serious and yourself.

Yeah. I mean, even before.

The emergence of generative, I still say like the dichotomy between build versus I. It's never been either or. It's always been a little bit of both. Like even when you buy a third party solution, you still always need developer solutions to to build the integration to connect it to your data. So there's always like a component of build with buy, like we're never going to get away from it.

Talked about this before but you know like depending on the company I, you know, often engineering and developer resources are often your most precious resources. And most of us don't have the luxury of working at a company where we have dedicated developers or engineers, especially, to help solve marketing use cases. And so unless what you're building internally ends up being a core capability that you can ship externally.

And my point of view, like historically, it hasn't really been worth it to build. I can say like as someone who has worked with a homegrown CRM tool before, it's not very fun.

And it's not reliable. I mean, it's so gratifying to you from.

Like a campaign operator lens. Partly like often the UI or UX lags a little bit. You know, if you think about like an iterable or braze, these are massive companies that have engineering resources specifically focused on making this like a delightful I mean, maybe that's a stretch, but making these, like, usable tools that were seamless, right. A little bit harder to replicate internally if like the CRM tool is just sort of an afterthought.

And so like from a fintech lens, like the question I would always throw out like, you know, are we a fintech company or are we an ad tech company? Like, do we want to be spending our engineering resources, building internal marketing and ad tech tools when we could be focused on building core product capabilities that will provide a better user experience overall?

And so that's historically been my perspective. I do think we're very much in flux.

If Sam Altman's.

Right and we're already at the start of the singularity, I don't know what's going to happen.

But like where MCP is come.

Into play.

Yeah, all of the above.

Right? You know, the cost of.

Building internally may have just changed dramatically. I mean, are we going to be able to vibe code our way to a workable CRM tool? I haven't seen that quite yet. What I would hope is like an intermediary step is even when we buy these internal tools, interoperability is still a massive problem. Data hygiene is still a massive problem.

I think there's a ton of opportunity to solve this first using a gigantic AI, basic core gen AI capabilities before we start like using a cursor or a windsurf to like replicate that third party tool. Like let's start with like, hey, like the martech paradigm has always been around. Like how do we get our tools to talk to each other and connect with each other?

We still haven't solve for that. And so like if we can start there, that would be a great place to start.

Would be well. And I also think of the longer term consequences, not just of building it in-house, all of the potential tech debt and things you've already mentioned, but also from a skillset hiring standpoint, if we're not on some aspects of standardized platforms and tools, the transferable skills, obviously they're there. But it's hard to recognize from a hiring perspective, oh, that person has used marketing, cloud and iterable so they understand scale and personalization and complexities versus not a dish that MailChimp or any of the smaller ESPs.

It's a different capacity and understanding and level of technicality. And you won't be looking for someone who has five years of match up experience. You bring on omni channel, multi-channel step in the same capacity. And so it's by no means a standardized test because I don't those are not even acceptable in their own regard. But it is a barometer for understanding who you're working with and what they're capable of.

And so there's so many question marks, even if we get it to be a reasonably priced, then how do we know folks can actually handle what they're supposed to be in charge of?

Yeah, I mean, that's also just the reality of the hiring market. Everyone wants a plug and play employee who has experience working with the exact tech stack that, you know, this brand may have. Yeah, you see that a lot with people who've worked at Metta where they're very focused on internally built solutions. You know, like you hear a lot of grumblings from employees where like it becomes harder to sell that experience externally when they're on the job market, for sure.

And they're not alone. Airbnb and Lyft also. I mean, each one of those companies also has an instance of one of the major platforms for very specific use case and of department. But when you have all these internal tools, it gets harder for observability and alerting when things go wrong because you might not have the mechanisms in the same capacity if it was your primary product that lots of top tier.

All right. As previously mentioned and the very beginning of the episode, you were part of million pages, adapter time book and one I was really excited to see your name there and also Graham release name there. And so what you discuss there is that there's so many tech players, part of every tech sector. And how do you consider and think through a North Star metric that aligns both engagement signals but also bottom line impact?

Yeah, I mean, that's always tough. Like how do you find that thing that unites everyone? And like we talked about before, like everyone's kind of chasing silos and optimizing for different outcomes, which you know, creates incongruent and fragmented user journeys and experiences. So like on one hand we're all sort of optimizing for LTV, but you can't optimize for LTV.

It's not it's too big of an indicator to actually activate and operationalize. What you want to find are sort of those proxy metrics and signals that indicate someone is going to become LTV or the interventions and experiments you're designing have an influence on long term LTV and retention curves. And so what you usually want to find is one in each to be time bound.

And that time amp should align with sort of the natural frequency or cadence of the product. So Grammarly as an example, like very much oriented around daily active users because people rate on a daily basis something like, you know, an Airbnb, the natural frequency of that product is going to be much longer because how often are people in the market to buy a vacation rental?

And so one that time AMP should align with how users should be engaging with the product. The other part is the the actual metric that you're measuring or optimizing for should connect to some business value or tell you that users are getting some kind of business value out of this. So, you know, like going back to the or the ride share example like an Uber, right?

Like you would do like number of rides completed or within the context of Grammarly number of suggestions accepted. That tells us that someone's getting value out of the product and we'll continue to use it and is less likely to churn. Yeah, I mean, what I would say Reforge has some amazing frameworks around how to arrive at your North Star metric.

Alina Verna, who I think is both of our heroes, like also has a lot of great content on how to come up with their Northstar Metric. I think what's important is when you do come up with that Northstar Metric, making sure everyone's.

Roadmaps.

Actually roll up to those Northstar metric outcomes versus everyone chasing their own independent outcomes. And so what often happens is marketing is choosing sort of surface level channel level engagement even within product, right? We're often chasing specific growth, funnel outcomes. And if we want that cohesive experience, we need to make sure it's still rolling up into that Northstar metric that's highly correlated with long term retention and LTV.

And I totally agree. I also think there's such unrecognized value in actually having company wide goals that every single thing you do ladders into that and I definitely have seen the failures where there maybe is in a really risky sort of company wide goal. There's more departmental goals which a lot of times contradict each other. Yeah, and it's not to the benefit of really anyone but the employees, customers.

You name it. Yeah. And when you're in a very cross-functional role like MarTech, when you have myriad stakeholders that you're trying to keep happy, that's especially challenging to operate in because how do you prioritize against opportunities and initiatives from lifecycle versus paid versus end product? Even within lifecycle, often lifecycle spans the gamut of multiple growth funnels. And so even there it becomes difficult to prioritize against different outcomes.

So you need a northstar metric to normalize the outcomes of those initiatives so that you can actually and think through opportunity cost and make sure you're working on the most impactful thing.

And it gets more complicated if you aren't just A straight B to C or B to B business, if you're a, B to C to B or B to B to C, it gets real muddy real quick. And that's why the juicy stuff, the fun stuff is, but also the harder uphill battle as well. All right. I want you to dream up your most ideal martech stack.

You're working at a brand new high volume company and you are solely in charge building the stack because nothing exists. Yet they've got such product market fit. So many folks are on it outside of, you know, logging in, everything is in product. What is your dream scenario set of tools? And maybe you can do two options of each category if you want to spice it up, have some options, but I want to hear what you would dream of.

Yeah, I mean, one data warehouse.

Is out there. I don't mean on martech, you know.

Argue. It's partially definitely.

Like a modern cloud based data warehouse, databricks or snowflake, to be honest. Like, I don't have skin in the game for either one, but like worked with both are the ones good enough on top of that, a reverse ETL or data activation tool? I'm a big fan of high touch for what it's worth. Like when I love their content, I think they have some of the most forward thinking content around MarTech.

I think they were very forward thinking as far as the shift from the traditional ISP's to the composable CDP space. So a reverse ETL tool like a high touch or census on top of your data warehouse, I assume there's probably some lifecycle channel because it's always good to have a direct line of communication with your users. So certainly an easy, probably a braise or an edible.

Again, like either one I think is probably going to get the job done. CMS certainly like I'm a big fan of the headless CMS is. So a content for experimentation is an interesting space. I've seen a lot of success with home grown experimentation tools. There are some like interesting experimentation platforms out there. Static is interesting. That said, like I would love to see a third party experimentation tool that like really like one of the things to solve for with experimentation is making sure user assignments and bucketing truly centralized.

I still see this all the time where like you onboard an experimentation tool and marketers are still setting up experiments and user assignments outside of the experimentation tool in their ISP or you know, using some other tool to do like randomization. And so then you end up with like a lot of data interference and, you know, issues with like not being able to isolate like which experiment intervention actually drove like the outcome.

Yeah. I mean to be honest, like hate advertising, I think that's going to be very dependent on the type of business and business model you need to figure out which networks, where you have channel marketing, you know, depending on your.

Audience, your customers.

Are. Yeah, exactly. So, you know, obviously like all the major ad networks, but like take it or leave it based off of like who your users are and like where you're most likely to find and engage with them. But yeah, what I'm really excited about too is like, how do we start building a more AI native or a AI growth native infrastructure and tech stack?

And by that I mean like how can we start using, you know, MCP servers to connect all of our tools? Like there's still going to be like myriad little third party tools that we're probably using on top of these major platforms. Again, like how can we do a better job of connecting all of them to make sure we're truly sharing the same audiences, that we can truly orchestrate these cross-platform journeys, which is still very difficult to do.

I think like that's going to be the exciting and emerging new development that we see in MarTech stacks in the next year or two.

A great I think that's the biggest unlock and we all just have to start experimenting, but also with the big asterisks of there's no regulation, there's no formalized ethics or even moral code around where privacy is in place. So there's definitely a lot of risk mitigation involved with that experimentation.

And how do you see the MCP connection with HubSpot? Like on the one hand, like it looks super cool. On the other.

Hand, like, I don't know.

Like having some agent interface with all of your CRM data and then potentially taking action, which in this case is like sending an email to you like your entire user base, like little high risk.

Yes, without a doubt. I think it's one of those you want to be an early adopter, but you don't want to be the first adopter. You don't.

Want to be the.

First one to fail.

Epically. Yeah, exactly. Well, it's.

Kind of like you don't want to be your ISPs. Largest volume center is you're the reason things are going to break. You don't ever want to be that number one dog in that regard. But yet on the same page. Okay. To wrap us up, who else should we have on the podcast?

Oh, well, obviously we need to do it. It's manifest. It.

Keith Jones from Open Air. He was on Phil's podcast recently and like you had a lot of really interesting takes. I think he would like.

I love to.

Talk or hear his perspective on the future of the MarTech stack. Anyone from Credit Karma marketing is great.

Sounds like I need some introduction.

I'm happy.

I'm happy to make some intros, although like they may not be like Credit Karma anymore. I need to like.

This.

Find in stalking. I think we need to get some like who's really dialed into the h a.k.a space right now and separating leisure. What is hype and theoretical?

What is fact versus fiction like?

What can we actually do right now? Especially like what can we operationalize today? Because I think it's still really hard to figure out in this space what is like still aspirational versus I can make this initiative in my upcoming quarterly.

Roadmap 100%. Agreed. Thank you much, Natalie, for being on here. Where can folks and you and learn more hear more of your hot takes?

Hmm.

Well, I did just vibe code my own website.

Oh, you want to plug it?

Yeah, like everyone you should be doing that.

You put your own website, it's like nothing else. Like it's good for.

Your own brand. It's good, like, exposure to some of the tools out there. So I have the domain. Natalie Miles dot me. Hopefully you can remember that or else can you find me LinkedIn? Although I'm not the best, LinkedIn's gotten like very noisy. So, you know.

There are.

Apologies for contributing.

And then I'm often.

At it like a lot of the are tech industry thing so you know feel free to say oh.

Well thank you so much for being on the pod.

Yeah, my pleasure. Always a pleasure speaking with you, Jacqueline.

I appreciate you.