← The Dave Gerhardt Show (from Exit Five)
The Dave Gerhardt Show (from Exit Five)
Inside Ramp's Marketing: Creative Bets, Measurement, and AI Agents (with Drew Pinta)
01:01:50
Loading summary
Transcript165 lines
- [00:00]
A
Hey, it's me, Dave. Our friends over at Customer I.O. are sponsors of today's episode. They're a really cool company that helps marketers turn first party data into engaging customer experiences across email, SMS and push. And they built their platform for marketers who actually care about the craft. Because marketing is a craft. It takes creativity, thought and taste. Right now, everyone thinks they're magically a marketer because they have access to AI and the result is kind of painful. More robotic emails, more noise, more bullets. Bl AI isn't magic. It's not going to fix bad strategy or write great copy for you magically. But the best teams also aren't ignoring it. They treat AI as infrastructure. When it's built the right way, it actually makes marketing feel more human, not less. And that's what Customer IO is doing. Their AI handles repetitive work like setup, orchestration and tasks that should be automated so that you can focus on what actually matters. The craft of marketing, the strategy, the creativity. This is how good marketers are using AI right now. Not to replace thinking, but to support it. If this landed with you at all, this idea about the craft of marketing, I want you to go and check out customer IO. It's customer IO. Exit 5. Go and check them out. Customer IO, exit 5. Hey, it's me, Dave. Today's episode is brought to you by Consensys, the interactive demo platform. Look, most of your buyers have already decided whether they like you before your sales team ever gets on a call with them. They've asked ChatGPT and Claude about your product. They found reviews about you online. They've talked to peers who've used your product before. And by the time they hit request a demo on your site, they've often already come to a decision. So you're losing control of the narrative before the first touch point. And now they have to wait three to five days for a demo. Consensus gives you that control back. They help you meet your buyer where they actually are with interactive, personalized AI demos that live on your site. When a buyer shows up wanting to poke around on their own terms, you give them what they want. Plus, you get to see exactly who's watching, what they clicked on and and who the decision makers are. So stop being a bystander. While LLMs sell or even unsell your product. Category leaders like Atlassian and Autodesk use consensus to turn invisible researchers into high intent leads. Go and check them out@goconcensus.com exit5 go consensus.com exit5 you're listening to the Dave Gerhardt show exit. 1221234 ramp might be building one of the best B2B brands of the last decade. They're doing compelling brand and out of home stuff like putting a famous actor in a glass box in Madison Square park to file paper expenses for 12 hours while live streaming the whole thing. And one of their employees actually got married inside the box for real during the stunt. But it's not all just big creative bets. Ramp is also deeply data driven and their marketing team right now is going through an aggress massive AI transformation. Marketers at Ramp are using Claude to do things like build their own dashboards, spin up go to market strategy for new verticals with the click of a button and reinvent what their job looks like. Their CEO told the team last year. Hey, it's okay to drop the ball on some of the things in your current job for a bit if it means learning how to rebuild and do better, work smarter and more efficiently and be more effective with AI. I talked to Drew Pinta, who leads the growth data effort at Ramp and partners with their marketing team about all of it. How they use AI to scan thousands of GONG calls and discovered something that was off in their marketing attribution, even though every attribution model said the opposite of what they ended up finding. Here we learned how they run incrementality testing, their channel maturity framework, how they split budget 7030 between proven channels and experimental bets. And Drew dropped a line that I can't stop thinking about. That was the headline for this whole episode, which is the fastest way to kill creative marketing ideas is to try to measure them like direct response. His whole thing is that measurement should meet the marketing, not the other way around. Then the back half of the interview went somewhere I didn't expect. Drew actually made the bull case for why marketers are in a better position than engineers in the age of AI. And he didn't sugarcoat the bear case either. We talked about both sides of it. I thought this was a great discussion with Drew from Ramp. And if you like B2B marketing, you want to get smarter about it. You're in the world of AI and marketing right now. I think you're going to enjoy this conversation packed with good stuff from Drew. Here it is, my conversation with Drew Pinta from Ramp. So Drew Drew's this is a repeat, repeat visit. We had Drew. Drew was on a webinar that we did all about abm. Turns out that's an awesome way to meet cool guests. Drew is, I guess, technically Not a marketing guy, but you're part of the marketing org Director of growth head of growth data at Ramp, which is a really cool company. I was happy to hear that. The marketing team at Ramp listens to the pod, so let's, we'll represent for them and do a good job today. Drew, can you just kick us off with like a little background about, like in, in your words, tell me about your role at Ramp.
- [04:59]
B
Yeah, for sure. So I lead the data function for all of growth, so that basically means I'm a strategic partner to our marketing team. So everything from like measurement to strategy is kind of in my team's wheelhouse. And although I'm not a marketer, I like to think I've picked up some marketing stuff over the. I've been doing this for like eight years now, so hopefully know a bit about it.
- [05:18]
A
Yeah, you have an interesting background. You were, you were like, did you, I briefly looked at your LinkedIn past. Like, did you work at the Fed or something like that? You had to what, what you do the Treasury. What were you doing before you got into tech?
- [05:30]
B
I did not plan this, Dave. I, I wanted to be an economist when I was an undergrad.
- [05:34]
A
Okay.
- [05:35]
B
I, I, yeah, that's true. So I went and like I did this two year RA program at the Fed where you're just like assisting economists. And within like three months I was like, I definitely don't want to be an economist. I had a great time there. It's just like academia is pretty slow. So that was like, what do I do? And I ended up in like a data science role. I got randomly placed on a growth team and it turned out to be a great fit. There's actually like a ton of parallels between growth and marketing and economics. Like in hindsight, say, tell me, explain that. I mean, I, I guess the way I think about it is in economics, you're trying to model a really complicated system. The economy is probably the most complicated system out there. You've got tons of inputs, tons of outputs. It's really hard to just like measure and know if it's even working without like a year long lag. And a growth marketing team's kind of the same way. Right. Like you have all these channels, you can adjust a ton of things on them. You don't really get great feedback right away. And so it's an interesting like data and modeling problem that kind of carries over.
- [06:32]
A
I like this, I don't know a ton about like the Ramp marketing. Org setup, but I like the idea of I'm just a big believer in like having the marketing, have, having the marketing team be good at marketing. And I rant about this a lot with like something that comes up a lot is like the market, you know, the CMO needs to like really know the PNL and all the finance and the background of the company. I'm kind of, I'm like, kinda. Yeah, like they can't be a moron. They have to know how like the sausage gets made and how much everything costs. But like the two CFOs that I've worked with, they're like, yo, dude, I want you to be great at like getting people to pay attention to us and make it the catch, register, ring and like, I'll be your partner. Is that kind of a parallel to like why this, the data, like your role and how you like support your partner to the marketing team?
- [07:20]
B
Yeah, I would say that's part of it. And then the other part is just giving them the information they need to make these decisions on like where they put budget, for example. Right. It's like, as you know, it's not that easy to know what's working, what's not. So that's like probably our most important role is like, where should we put money?
- [07:35]
A
Why is that? Let's take the like higher level lens and just kind of. Yeah. I talk about this a lot on the pod, that like a lot of the things that are good marketing are a bit harder to measure, but if they work, they end up working and then it's an obvious how they work. So I'm curious how you, how you think about some of those, those things. And I see ramp, like ramp does a, a lot of stuff that's more on the brand side of things. But I'm sure deep down you, you have a way to measure the, the impact there.
- [08:00]
B
We try. It's never perfect and it's not possible to perfectly measure everything, but a lot of times that's where the kind of alpha is, is doing the things that are harder to measure. Right. Everyone can like run Facebook ads and those are not easy to measure, but easier. Not everyone can like put a guy doing like expenses in a park in a glass box for a day and run a ton of ads around that. And we did, we did measure that. Like we did an event study for that. I can't give you exact numbers, but
- [08:25]
A
what's an event study?
- [08:27]
B
So we basically look at like, how's our marketing going today? We forecast what it would look like over the next week, then we do something crazy like A box done. And then we look at what actually happened and you compare those two things and you can get a read on what the impact was of doing your marketing intervention.
- [08:44]
A
So you're just measuring. I mean I, I love, I like when we can take things like this and make them simple. I often use the example of like there's a lady down the street, she runs an amazing, you know, sandwich shop. If she did some ad like what, what's what I love about small business. It's like it is easy to measure because it's like if you did, if you did a crazy stunt or just did some cool, more people started to come to her store next week, she'd be like huh, that worked right? She wouldn't be like, you know, deeply measuring the metrics on the, on the Facebook ads. But I think something, it's because of how B2B works. And I had the VP of marketing of, on the LinkedIn ads business on a couple weeks ago and he had the stat that was like the average, you know, deal, deal buying cycle in B2B is like 211 days. So that's you know what 70% of the year and there's 22 people involved. So it's like no wonder this is, this is harder to measure. You mentioned even then that briefing you mentioned like marketing's performance right now just baseline, like what it explain ramps, funnel. What metrics are you looking at like right now to know if what's the baseline of ramp this week? Is it like free, free trials, sales demos. How does the business work?
- [09:51]
B
Yeah, we're about 70% sales led. So the marketing team is like primarily trying to drive sales qualified leads. They'll move through the funnel. We don't have a free trial but we'll move them through to like actually signing up, activating and then you know, expansion retention. Kind of a classic funnel.
- [10:07]
A
What is a, what is a sales qualified lead at Ramp?
- [10:10]
B
Yeah, I mean someone who's like actively considering the product they're actually they can buy right now. And I think we talked about the webinar. Like our tam's huge but a lot of our customers just can't buy right now. They may have like a four year contract signed with amx. Right. So we'll catch them in four years. Maybe now we'll try to sell them another product in our suite. But they need to be actually like a potential buyer today.
- [10:33]
A
How do you identify that person?
- [10:35]
B
Sometimes we'll get on the phone with them. Yeah for self serve leads. We'll ask them questions in our onboarding, when we can, we'll try to use data to surface that.
- [10:43]
A
Which would be what, like, signals are there. Is there some level of, like, intent or data? What data would. Would give you a. Like, a breadcrumb of that.
- [10:51]
B
I mean, there's the obvious stuff, like, if we can identify that they're coming to our website, looking at our pricing page, things like that. There's stuff like a new CFO going to the company who maybe used Ramp at their last company, sort of like, those kind of obvious things. But it is a pretty tricky thing to actually parse out. Like, do they have a contract with a competitor they're locked into right now?
- [11:12]
A
Right. Yeah, that's the hard part. Like, they might love Ramp. Love the marketing. Came in. Really want to, like, move to something cooler than whatever they're using now. Ramp has a lot of buzz, and they just can't. They can't buy it. Okay. Have already sent us in a. In a different direction. We. We have prepped for this episode. I've been, well prepped for this. So first topic I want to get into is how we gave leadership the confidence to grow the marketing budget. This is a great topic for a lot of people that listen to this, and I think from where you sit, it'd be really interesting. So it seems like this is a note that I have in here. We use AI to listen to 10,000 sales calls and discovered that our attribution was completely wrong.
- [11:48]
B
Yeah, it's a good one.
- [11:50]
A
Can you tell me about that?
- [11:51]
B
Accountants are, like, the people whose lives we make easier. They go from, like, having to manually process expenses to it's fully automated for them. So we put them in a box for, I think it was 12 hours. It was a glass box in Madison Square park, super busy park in New York. And we had them file expenses, like, paper expenses for 12 hours. We had a lot of special guests come in, pop into the box every hour. Someone on my team actually got legitimately married by Brian in the box on this live stream. So we're doing all this, like, crazy shit. And, like, it wasn't the most expensive campaign we've ever done, but we spent a significant amount of money on it. So we wanted to know if it worked, and we wanted to know we should do similar things. So we do. Like I mentioned, we did the event study on it. It's like the classic thing to do, but there's a lot more you can do now, especially being in B2B. Our sales. Our motion is pretty much sales led. So we have a lot of conversations with our customers as they're kind of like, in the buying process, both over phone and email. And we get transcripts of all of this. Yep, we use gong, so we're getting, like, those gong transcripts. So it's pretty easy nowadays to, like, pipe all of that through an LLM and just be like, tell me every time a lead or customer mentioned the box. Done. Or tell me a time when they said that's how they heard about ramp. Even better. And we can do that and get a chart of like, wow, we've got hundreds of leads mentioning this. And it's hard to quantify the value of that. Like, I don't know how, what percent of that deal the box that should get credit for, but it tells you that it's at least working and you're getting awareness. And the same principle applies to, like, podcasts, for example. I'm sure you have sponsors of your podcasts. There's no. We do that all the time. There's no, like, great way to measure that mathematically. So we'll go look at how much our customers mentioning this podcast, and it does influence our buying decisions going forward.
- [13:44]
A
Yeah, so the podcast ads is, like, a perfect example of this because so we do have sponsors. And I think one thing that we've learned over the years is there's often a gap between what the sponsor wants to do to drive leads in the short term versus what actually works on this medium of podcast ads. And so we still are in this world as marketers where like, okay, we're sponsoring this podcast, and so we. We're going to use this custom URL and there's going to be a form there, and we're going to, like, know that everyone that came in through this page came from Dave's podcast. Okay, but think about your podcast listening experience. Right? I'm listening to a podcast when I'm at the gym, when I'm on a run, when I'm around the house. And so I might hear an ad I often listen to. An example would be, like, I listen to, like, my first million while I'm doing the dishes at night, my kids are up taking a bath. I'm doing the dishes. I hear that. I'm not gonna stop what I'm doing, doing the dishes. Pull up my phone, go to the. Go to the website, type in the form, put in the code, and sign up for, you know, ramp if they advertise that podcast on the spot. But I listen to, you know, founders podcast with David Senra, and I'VE been listening it for, you know, three to six months and ramp, ramp, ramp, ramp. Over time I hear ramp, ramp, ramp, ramp. And so guess what, guess what happened. We, we use ramp now at, at exit 5. And like, that's how we, that's how we got on that. Because it was like, finally got to the point where, like, should we be considering ramp? Like, I don't. I've never thought of this. I said it to Dan. Dan's like, I love ramp. And now we use it. And it's like, ah, that's how marketing works. But the problem is inside of a company, it's often like, Dave is the marketing manager, Dave owns a channel, Dave own. Dave has to deliver. You know, in your case, you mentioned SQL, so I'll just use that. Dave has to deliver. I'm on the hook. I decided to sponsor this podcast, so I'm on the hook for 20 SQLs this month from the podcast. But the actual window of when people would actually convert from the podcast is much longer, which, which throws off the whole thing. And it's like, ah. I got into marketing because I like doing the fun, creative stuff. I want to get people to pay attention. Not everything in marketing is direct response, but we often will try to like, measure it that way.
- [15:44]
B
It's like measuring a Super bowl ad with like, website sessions. Like, if I see a Super bowl ad, I'm not like on my phone going to like the website or buy or thinking about buying or putting my email in right now. A total waste of time. And like, the fastest way to kill creative marketing ideas is to try to measure them, like direct response. And you shouldn't even try to do it. Like you can, but it's a waste of time. I strongly believe that the marketing or the measurement should meet the marketing, not vice versa. I think too often people try to force the marketing to fit into the measurement, meaning they'll do things that they know will look good in the measurement strategy that the company has. But really the measurement should meet the marketing where it is.
- [16:24]
A
Hell yeah. That's a, that's a quotable right there. The fastest way to kill creative. Damn, I don't even. I'll play it back. But yes, man, this is exactly what we want to talk about. This is. This is the stuff. Okay, so, so how do you, how do you build a culture? How do you, how do you, like, how have you enabled that at the team? Like, I want people to listen to this and be like, I get it. Most people will say, yeah, that will. Must be nice. If you're ramp and you have a data science guy on the marketing team and you can do all this stuff, but, like, let's kind of make this, like, approachable to people. Like, that's a core. That's a core philosophy, right? You're not just doing marketing in a vacuum. You're having a. You're having a philosophy on this. So how do we make this, like, approachable for most people?
- [17:01]
B
Yeah, that's a fair question. I don't think you need a data science team to do a lot of this. I go back to my econ roots where, like, you're creating hypotheses about the economy or your marketing, and then you figure out ways to test them. And that test is going to be different depending on the channel. So if you're doing something like direct mail, it's pretty straightforward to hold out 10% of your audience and do an A B test and see what the impact is. If you're. If you're doing something like a podcast, it's way harder to measure it, but you can still form a hypothesis and test it. And that could be around, like, maybe you have a how did you hear about us? Survey. I think most companies have that. You don't need a data scientist for that. How often are people going to mention this podcast in that survey? And I kind of go from there, but I think where you kind of ground yourself, because on the flip side, it's like you don't measure anything. You just do a bunch of random stuff. So you need to, like, form hypotheses and test them, but you need to be flexible about how you're testing them. We have had a lot of success with incrementality testing. That's kind of for us. What to get back to your headline gave leadership the confidence to 2 or 3x our budget.
- [18:05]
A
Can you define and explain incrementality testing for people that might not be familiar with that concept?
- [18:09]
B
Yeah, for sure. I mean, it can kind of like vary depending on the channel. But the idea is you're trying to figure out the causal impact of spending money on a channel, not like I spent more money and then we got more leads, more MTA said that. But if I did an experiment and was actually able to say, like, 50% of my audience can't see these ads, what would have happened? That's kind of the gold standard. And so how you do that depends on the channel. For something like billboards, you do a geolift test, you run them in certain areas, other areas you don't. And Then you compare those two geographies over time and see if the one that you were running billboards in is now doing much better. And that tells you like what would have happened in absence of that marketing spend. And you can back out an ROI from that.
- [18:49]
A
In that example of the billboards, like, would the measure of success be like sales? Did Ramp close more customers in Vermont or where we were now? Or is there some other leading indicator that might also prove to be successful? Because to your point earlier, just because you show the ads to those people, you might have just shown them to a whole group of people who now know who RAMP is and likes Ramp. But like is stuck in a contract or is not ready to buy. It's not. I think one of the hardest things that marketing can. One of the biggest things the challenges is to create is to create urgency. Like why, how do, how does that. That billboard is not going to get you to buy Ramp now. Okay, but we're going to look at it and be like, well, we spent all this money in this geo. Like how do we prove that out?
- [19:30]
B
Yeah, it's a great question. This is one of the hardest parts is like what do you use in a B2B company as your sort of metric you're optimizing to. We'd love to do like actual closed deals, but that takes way too long for these tests. So we'll look at SQLs. We'll also look at some upper funnel stuff. We'll look at website sessions even though it's not the best thing. We'll also look at brand awareness. So we, we have a brand awareness survey survey we run weekly and it has a GEO component to it and we can see a brand awareness has increased. We know the value of increasing brand awareness by one point on like down funnel metrics based on historicals. That doesn't necessarily mean it 100% carries through to any experiment we do, but we have a rough idea of like how that'll impact us going forward.
- [20:11]
A
So it's kind of like the whole game is like you're trying to. It's like a portfolio management, right? Like you have all these things that like all these people. We have the baseline. If we turned everything off, here's how many people would just show up to Ramp because we exist, right? And I know this as an example for my, you know, small little business. That's exit five. We don't really do much outside of like LinkedIn ads. And so I kind of know what website traffic is going to be steady state. And if we Turn that off. And then you're like, okay, we're going to place a bet here and we're going to measure that. We're going to place a bet here and we're going to measure that. And then you're. It's just like this kind of game, this constant game of like, this thing's working, let's do more, right?
- [20:47]
B
Yeah. And you're going to have a lot of Ls. Most of them probably aren't going to work. And that can be like a frustrating part of incrementality testing. But you got to think about them as like actually valuable lessons. A, because you can reallocate that money and B, because it may just mean that you're not as mature on a platform as you thought you were. Yeah, we've had some, some losses where we were like, how did this lose? We were sure it worked. And then we go look at the platform, we dig in, we're like, oh, we're, our audiences are fucked up or we're optimizing to something stupid. And it's like we actually weren't even ready to run ink mortality tests. But we did it. But now we know we need to improve on this product.
- [21:21]
A
So. One of the biggest mistakes I made in my marketing career was I was running marketing at a company that was growing really fast. And I was so caught up in the. There's just so much going on that I only was able to focus on like right now, like the leads this month, the leads this quarter, and then when you get the plan for next year and it's like, okay, Great, we did 15 million next year. We're gonna grow to, you know, 20, 26 next year. It doesn't work like you just drag the spreadsheet of the existing channels. But I didn't have enough, like I wasn't forward thinking enough to be testing enough channels. So then when we got into the new year, it wasn't like, okay, cool, we're gonna go dump all this. We're gonna spend this money here and spend this money here because we hadn't tested and learned. And so it's like, it's almost like you need to have, you have the main thing and this thing is working, but you're kind of always like running experiments behind the scenes to figure out what the next channel could be. And I love what you said about accepting that things are not always if everything worked, like it's just not reality, that all your tests, all your ideas are gonna work. And so it's like you gotta have 10 ideas, and maybe two of them are gonna hit and we're gonna kill eight, and then we're gonna double down. Do you see it as that type of game?
- [22:28]
B
Yeah, I mean, it's totally shots on that, I think. I'm constantly surprised by what works and what. Which is why, like, I love being in growth and marketing is I'm constantly surprised by what works and what does.
- [22:39]
A
What's something that was surprising?
- [22:40]
B
I mean, I'll go back to the example of, like, scanning our gong calls. So we scan all our gong calls. We're, like, finding mentions of how people heard about us. And when we Looked at that, LinkedIn was mentioned, I think, three times more than meta, even though all of our prior measurement using, like, an MTA model, things like that had said Meadow was like, three times more effective than LinkedIn. So we saw that and we were kind of, like, shocked. And it. What it led to was, like, we didn't just go reallocate the budget overnight, but we did a bunch of incrementality tests to actually, like, prove that hypothesis out. And it did lead to us making, like, substantial decisions. That's not, like, a cool example. I feel like you wanted a cool example.
- [23:14]
A
No, it's okay. It's all right. Whatever. It's okay. You're doing a good job either. Right. This is kind of what's really interesting now about. Especially if you're in a company where you have that touch of, like, someone has to get on the phone and talk to sales because it's almost. It's making a big dent on the attribution puzzle because you're getting that data without having to ask it? So is there a best practice there? Like, is there a standard operating procedure with sales where you want them to ask, like, hey, how did you. How did you hear about us and get some of that data so then we know we can get the transcript and kind of crawl that stuff later?
- [23:48]
B
We haven't done that at all. Our sales team's hungry. Right. They're just. The main thing they care about is hitting their quota.
- [23:54]
A
And, yeah, they're not like, let me. Let me make sure I ask. Let me make sure Drew knows the sort. Well, what was the source of your lead, sir?
- [24:00]
B
Yeah, I could ask. It would be nice, but I have a feeling it may not totally land.
- [24:04]
A
Okay, but don't you. I. I feel like even in our little business example, I feel like that stuff just comes up all the time. We have this product that we sell to CMOs, and we ask every, like, people just oh, you know, I, I decided to apply for this thing because like I heard about you on the, I heard about, on, on Dave's LinkedIn or I heard about podcast. I'm, I'm always surprised at how willing people just kind of offer that up in, in conversations.
- [24:24]
B
Right. I mean, and that's arguably a better read is when they just. Yeah, right. Versus us asking them.
- [24:29]
A
Well, right, but, but this is a whole game of like triangulating. It's never going to be like this like paint by numbers approach of like how to get more customers. It's like you're looking at, you're matching up all the data, you're like looking at what you're seeing and then you're hearing. Well, actually, but everyone's kind of telling us they're hearing about us from this thing. So this thing is clearly working and you know, it's never going to be perfect. But.
- [24:48]
B
Yeah, I agree.
- [24:50]
A
This is an interesting note I have. In our doc here, LinkedIn was mentioned three times more than meta, even though your, your MTA showed the opposite.
- [24:57]
B
Yeah, no, it was very surprising to us. I mean, in hindsight, looking back, like, I think LinkedIn's more of a view through channel than a click through channel. And we have much better data when people are like clicking ads and coming to our site. So it wasn't surprising. But this is kind of why, like I wouldn't trust any sort of deterministic attribution model like MTA or Last Touch or anything like that, especially today with all the privacy stuff that's going on. So you really need to be testing and looking at like, like talking to your customers and figuring out, well, that's
- [25:27]
A
the other, that's the other wrinkle in this thing. Like there's so much, you know, so many things that you can't do anymore or are just straight up illegal now because of data and privacy. How can you perfectly track all of the touches that led to Dave becoming, you know, an SQL and ended up buying from you?
- [25:44]
B
You can't. And it's a, it's crazy to try. You'll never get it right. So it can be a useful signal and it's useful for like goal tracking. We have an MTA model, but we don't use it for like allocating budget across channels. We use it for like setting goals within a channel and understanding like within a channel what campaigns are working, things like that.
- [26:01]
A
Let's talk about that. The out the, the budget allocation piece. So yeah, how does that play into. Give me some lessons learned. Advice for those listening out there, a lot of people ask, hey, my company gave me a million dollars in the marketing budget this year and I got to figure out where to spend it. How would you go about solving that problem based on what you've learned from your, from your work at Ramp?
- [26:21]
B
I'd go start spending it and see what works based on tests. Like, if you have no data to go off of, I'd go start spending it. You got to start to figure out what are the efficiency curves on each of these channels. And the only way to do that is to go spend and get some data points.
- [26:33]
A
Well, but, okay, I want to, I want to like, crack into, like, the real reason of you saying that, though. I think people ask because it's like a defense. Like, I don't want to, oh, my God, I got all this, I got this budget to spend. I don't want to go spend it in the wrong places yet. Or even if you have data we often want, data is very rarely, it's very rarely going to be a perfect path for like, if you go and spend this money here. If I go and spend this money on LinkedIn, I'm absolutely going to get X out and I'm going to hit our goals. And it's like, well, but think about how many variables there are in our. There are in there, like, there's the copy, there's the creative, there's the audience, there's the competitive situation, there's the market conditions. There's. There's so many. So I want to try to just like, give people that, that attitude other than just like, yeah, go spend the money. There's actually like a real, the bigger lesson in there is like, you got, you got to like, be in the game to, to know some of these things. Part of it is like, they give you the money to, they give you the budget because they want you. You need to spend it. Right. Like, don't be afraid to spend the budget.
- [27:29]
B
Yeah, I mean, to go back to your, to exit five, you mentioned, like, you guys are just spending on LinkedIn, right. So I, I wouldn't, I don't think you need any of this fancy stuff when you're starting. If you, if you're just spending on like one or two channels, you can just do that and see if your metric goes up or down every month and you kind of understand if it's working better or worse, that's probably fine to start. And it's not until you start getting into like maybe three plus channels more than I don't know, a few hundred thousand a month that you really need to go beyond. Just like the top line. Are we getting better or worse? And what are the in platform numbers? Show me in matter, LinkedIn or wherever you're spending.
- [28:03]
A
Do you have any perspective on, like, how much you should be focusing on from a budget standpoint? Today's growth versus, like, testing into those new channels? I think we've often talked about it's like, is it like 70% on today and 30% on the future? 80, 20, actually exactly where we're at.
- [28:18]
B
We're 70, 30. Yeah.
- [28:20]
A
What is it? 70, 30, Rick?
- [28:21]
B
Roughly.
- [28:22]
A
Yeah. And so that, that 30%, is that for like next quarter or next year or like just future things that you want to take off? Like, can you, can you explain take me through that 70, 30 split and how you think about it from a marketing budget standpoint.
- [28:37]
B
We'd love for it to be next quarter. Like, you know, perfect world. We test a new channel and it goes great. And we're like, boom, this is a core channel now we're, we're moving it to core. We're, we're scaling it up a lot. Realistically, it's probably more like 6 to 12 months out for those channels. We have a sort of channel maturity framework. We actually take new channels through. And this is because, like, you actually don't want to jump to incrementality testing. Like, incrementality testing is great, but I don't want to come off as being like, it's going to solve all your problems. It's really expensive and slow. So we start by just kind of like getting the fundamentals right on a channel. We'll allocate like, some amount of our experimental budget to it. And I do think, by the way, having an experimental budget, having that worked out with finance is key. You need a budget that's not. They're not going to be analyzing. And being like, your LTV to CAC overall is terrible because you tested a new channel and it was terrible. That's healthy and that's fine. You need a way to measure, like, what is our LTV to CAC on the core stuff that we really believe in and that's mature versus, like, be able to take bets and not have it tank your overall numbers. Sorry to get back to your question. So I don't know. Did I answer it, Dave?
- [29:37]
A
Yeah, you got, you got me. Like, so I wrote down this nugget that I was, I wanted to come back to, which is like, so you mentioned, like, you ended up saying this channel, you have a channel maturity framework. So you mentioned like, okay, a channel becomes core, so we got 70, we're spending now. 30 is to like scale up in the future. And we're testing new channels and you know, Reddit ads or whatever, different stunts or we're testing a different trade show. What actually goes into that channel maturity framework? You don't have to give away like the secrets at Ramp. But I'm just curious, how do you, how do you think about that? Like there's all the marketing channels, these are the three mature ones. So what do you want from the others? I think just giving people some kind of like general framework on how to think of all of the marketing channels and put a wrapper on them is interesting. So I'd love to dive into that a little bit.
- [30:22]
B
Yeah, we kind of have gates on what it takes to graduate through each step in the framework. So there are things like you need to have solid optimization events set up, you need to have solid like just baseline reporting set up to move from like the earliest stage. We need to be seeing it come up in those customer calls and the how to do your about us survey and then we'll sort of move it through and then after that it's like you're kind of actually running this thing. So we'll try to do. Some channels will let you do an in platform incrementality test and these are a lot cheaper, probably a lot less trustworthy. But right, this is like you go into meta and Meta is running a test for you where they take half your audience and they say we're not going to show them any ramp ads and then we're going to show the other half ramp ads. We're going to tell you what the impact of that was. It's great because you don't need to do one of these long geo holdout expensive tests. It's bad because you have to trust meta that they're going to tell you the truth about like impact. Right. But we'll do that first to just prove out that a channel actually works. And then after that we'll go to sort of that the true incrementality test that we trust and compare across channels and once it passes that then it's sort of mature for us.
- [31:26]
A
So if you're looking at how you're going to hit the plan for like a quarter or a year. I made the mistake of like making too big assumptions about like new channels and it's like, well actually we're going to get a million new from this channel and it's like, well, we haven't even proven that out yet. So do you build the plan with assumptions around core channels? And like, how do you factor in only core? What? Only core.
- [31:47]
B
Only core. But we're, we're further along, I think, right? We have enough core channels where we can do that? Sure. If I was just starting out, I don't think you, you probably wouldn't have the luxury of being like, oh, we're going to drive 0sqls in the plan, but we're going to spend a bunch of money.
- [32:01]
A
What are some of the core channels today? Hey, it's me, Dave. This episode is brought to you by our friends at Knack. Knack is a no code email and landing page creation platform focused on a problem every marketing team runs into. Have you ever had a really good marketing idea, but then it takes forever to actually ship it out the door? It's usually not because your idea is bad, but because the process in the middle is slow. Briefs, more briefs. Approvals, reviews, tiny fixes that somehow turn into weeks. And by the time the campaign is finally ready to go out, it barely even looks like what you originally wanted to ship. Yep, that right there, that is the gap that Knack exists to close. Knack is a no code email platform built for modern marketing teams. They have AI built into the platform that lets you prompt ideas and instantly generate on brand email assets so you can create, review QA and launch your email all in the same place. No jumping between tools or messy handoffs halfway through after the email goes live. Knack also gives you performance insights and recommendations so you can see what worked and how you can make the next send better. So if execution is the thing slowing your marketing down or you just want one system that takes you from idea to shipped, to learning, to improving, you should check out knack. Go to knack.com/exit5. That's k n a k.com/exit5.
- [33:24]
B
Yeah, I mean Google huge lame LinkedIn lame meta. Yeah, I mean they're the big ones. Like I think we meant I mentioned on the webinar, we're pretty big.
- [33:35]
A
The biggest scam in the world. I'm awake to this now. The biggest scam in the world is venture capital dollars trading hands. Because all the. We all just, we all spend our money on Google. It goes back to Google. It's. It's insane. Let's just do. I want more guerrilla marketing. I want more actors in boxes. Okay. Take that back to the RAM team.
- [33:54]
B
I would love a world where everyone Declares truth and says we're not going to spend any money on like Google branded stuff or maybe any Google stuff. Like, how nice would that be?
- [34:03]
A
Like back, this is for my first time running marketing, the bane of my existence was like not spending on paid and then having to have the VP of sales and cfo literally. No disrespect, I don't think either of these people could use their cell phones effectively. But they were like, look at. Why are we, why is this, why is this ad from Intercom appearing ahead of the drift ad? I'm like, well, because they're spending like an ungodly amount of money on this keyword that we don't even want to. I don't care. Why are we showing up? And it's like, how many companies have to play that game because of internal politics?
- [34:35]
B
And did you incrementality test it? Did.
- [34:36]
A
No, I'm not that smart.
- [34:37]
B
That's what I would have told if I was your data partner. I would have been great like that. They've got a good hypothesis. Let's test this.
- [34:43]
A
So what, what would we have done there? Turn it off? Like, turn it off and see or run some and see. What would the incremental incrementality test be?
- [34:49]
B
We probably, if we really wanted to be rigorous, we would have done a geolift test, right? Like where we say we're going to exclude these 25 DMAs or whatever from this campaign and then we're going to compare those DMAs to the ones where we kept it on and we're going to see if there was like actually noticeable and back. My guess is there probably wouldn't.
- [35:06]
A
But if you were like this was early stage, like, I don't know, a couple million in revenue. Do you pick geographies where you feel like a lot of customer? You're not going to run them in Mobile, Alabama, you're going to run them in Boston and San Francisco and you know, New York. Do you pick, is that how you choose the markets?
- [35:19]
B
You need to do it randomly, actually. Oh, well, making sure you're, you're sort of preserving some assumptions. And this, this is where like it can get a little complicated. But there are great vendors out there that can help you with this stuff. So you really don't need a data science team to do this. And we use a vendor too, by the way, because it's like they're pros at it. It's not the highest leverage thing for my team to be spending our time on.
- [35:39]
A
That's my takeaway. Incremental we, we actually just were talking about something in a, in a newsletter with a newsletter ad that we ran yesterday and Dan and I were like, we should just test it. And I think that's the, that's my mentality is I'm always like, I move really fast. I have a bias for action. But I don't think like a scientist or a data science person. I don't go back to the scientific method. And so then I'm always like, damn it, I guess we should have isolated that variable and test it. But. And then we had. I don't know if you've ever heard Dave, Dave Kellogg speak or familiar with his, his content or all like, he's been a CEO CMO in SAS for a long time. He spoke at our event two weeks ago and I love this. He said the job of the marketing leader is to be a dispassionate analyst, which I thought was great. Which, like, I'm looking at all, I'm, I'm, we're placing all these bets, we're doing all this and I, but I don't have a dog in the fight. I don't care if we win in this market or that. I don't care if this channel wins or not. I'm going to be like a, the arbiter of the, of the truth. And I kind of feel like that's what's cool about your role in partnership of the marketing team. You're like, I don't really care if LinkedIn works or doesn't work. I'm just going to tell you the truth about like what, what we're seeing from this.
- [36:39]
B
Yeah. I mean, I think of a marketing leader as like sitting in the cockpit of a plane. My team's job is to give them more instruments. Right. Tools in their sort of toolbox to know what's going on. And then they're trying to like take actions based on that and increasingly maybe like there's machines and AI making those calls. But at least now that's kind of the world we're in. They're going to take an action and to go back to the complicated system modeling problem. It's like some of those instruments are going to react to it. They're going to react on a lag and then we have to kind of interpret what the impact of it was.
- [37:11]
A
So the marketing leader is flying the plane, but there's like some ladies in like row seven and she's like, why don't we have oat milk on this flight? And then there's some Guy in, you know, row 10, he's like, it's too hot in here. Like, everyone's unhappy, everyone wants something different. Your job is just like, get me there on time.
- [37:28]
B
Yeah, I think that's about right.
- [37:30]
A
Yeah, yeah, something like that. You mentioned machines and AI, and I do want to ask you about AI because where do you think a lot of this stuff is going? I'm sure a lot of your work and growth and data science is, you know, you're spending high volume, you got high volume activity. With a company like ramp on LinkedIn, on Meta, are we getting, are we going to get to the point where you're really just like letting the. And I know a lot of this is happening already. We're just letting the AI and the AI ad systems at these companies just choose which ads to run and it's just going to happen for us. Like verse, man. I remember trying to figure out AdWords, you know, 15 years ago and it was like learning a new language and trying to figure out like, you know, you got a spreadsheet, you got all the bids and you got all the keywords and the exclusion. It was insane. Now we've kind of shifted more where it's like you're going to press a button. Is that a positive trend? Are you seeing that? Am I making this up? It's kind of like, why can't Claude just do. Claude is just doing all my work for me. Can I just log into LinkedIn and say, like, run these ads to these people and like, I'm just gonna sit back.
- [38:24]
B
I mean, I don't think you could do that LinkedIn because I don't think they're as advanced as like Meta or Google, but I think we're pretty close to that on Meta and Google. It's easy for me to say that in my chair where I'm not the one sitting there like in the platforms every day. But yeah, we're doing a lot of work to automate both on our side with like launching campaigns and creative, but then also just kind of trusting the black box on their end. It's like the most important thing is we give them a really good signal to optimize on and then we sort of just let them run.
- [38:50]
A
What's the signal that they. What do you mean?
- [38:52]
B
What I mean is like a few years ago we did a lot of work on setting up really custom audiences and segments and targeting, trying to basically tell Meta or Google, like, here's exactly who we want to show our ads to and what you want to show Them and now I think the machines are smart enough on their end where it's like, you should just tell them anytime someone converts, you should just tell those platforms, hey, so this person converted. Here's who they are in an anonymized way, of course. And like, here's how valuable they were to us. And other than that, you just take the, take the reins off and let the machine kind of run and figure out like who looks like that person, what ads will resonate with them. That makes sense.
- [39:32]
A
Yeah, makes sense. I'm just happy to be looking at your LinkedIn while you just. You have a, you have a master's degree in computer science.
- [39:38]
B
Yeah, I do.
- [39:39]
A
Where's the world going? You specialize in machine learning. What's your opinion about where marketing is going? Like my, this audience is all marketers. Yeah. You work with a high performing marketing team. You have a background in computer science.
- [39:50]
B
Yeah, I have some takes.
- [39:51]
A
Can we, yeah, give me some takes, man. And I hope they're, I hope they're like, I want to make the, I hope we're making like the bull case for marketers.
- [39:58]
B
I'm overall bullish, but I'm going to be, I am also going to be a little bit maybe harsh.
- [40:03]
A
Okay, hit me.
- [40:04]
B
Do you want the bull? I'll give the bull case first. So AI and LLMs is trained to give you the median output. That's how these things are built. They're statistical machines and they're spitting out the average or the median. And in marketing the median is death. If you just run like average ads, you're dead. Like you're not going to stand out. You need to be unique and like the alpha is being creative and having taste and things like that that I don't think AI is great at today. It could be in the future, but I don't know, I think there'll always be a human element there. And I compare that to like backend engineering. Right. Where they're just like writing backend code all day. For a backend engineer, average is good enough. Right. If the code runs and it passes the tests, they succeeded. And honestly, being average when you're writing code is actually better than being like above average because you want to write standard, easy to understand, easy to test code. You don't want to write this like crazy fancy code. So I would argue that like, compared to engineers, I think marketers are in a much better place where there's still an important place for the human element. That's my bull case.
- [41:14]
A
Okay, I'm into that. I believe it. My bias is that I am having a lot of fun right now. Look, I. The. The thing that I. And apologies if somebody listen to this. They've listened to my podcast a lot. I've talk this a lot. But I got promoted into, like, running the marketing team because I was good at marketing. And so eventually that. That's what happens in a lot of people's career. Like you. You do good at the job, and then all of a sudden you're the leader of the, of the, you know, you're the manager of the people that do the job. But what I'm actually good at is like, the storytelling, the creative, the ideas, the landing page, the. And what's so cool to me about what's possible with AI tools right now is like, I can have an idea for this. I can have an idea for something. And like, the time from idea to me shipping that is like, you know, I don't need a designer, I don't need a developer. I don't need to do this or that I can, like, actually make stuff. And so I'm like, that's. That's the case for me. And I do think sprinkling, like, even I've noticed in our writing, like, I'm using AI to like, help me write in research. But the things that people react to are what the, The. The person, the little injections of personality or a joke that I make or something like that, which is the whole taste piece.
- [42:12]
B
Yeah.
- [42:13]
A
The only problem, the only issue. One of the things that I am worried about, though, is like, in B2B, what happens if they just cut everything out of the middle where it's like, it's ramps AI. Dave is. Dave is buying. However you would define ramp. I don't know, expense management, a credit card. I need a company credit card. I'm not going to do the research for that. I'm going to have my. I'm going to have my AI agent go do the research and do all that. And so it's really just. My AI agent is talking to ramps AI and then they're going to give me a list and it's like, all right, Dave, you should go buy one of these. And so there's no human involved at all. That's the. That's the middle that I'm worried about where it's like, I think if. As long as humans are involved in the buying process, I think there's always going to be some. You know what? I kind of like that. I like that brand. I like that brand better. You know, I want to, I want to. I'm going to buy them. I like, I, I went to HubSpot's inbound conference and I really like them. I'm going to buy HubSpot, you know? Yeah, I don't know.
- [43:01]
B
I don't have an answer for that one.
- [43:02]
A
Nobody knows.
- [43:03]
B
I mean, if we get that world, then like what, what, what is any human doing if it's, if it's all just machines talking to each other?
- [43:09]
A
That's why. Dude, I'm dead serious about this. I am. I'm gonna be like, I want to be, I want to be in both. I want to have a foot in each. I'm going to be. I'm going to build a local business in my town. Fact, it's going to happen guaranteed. I'm going to be in the community. I'm going to build something in, in real life and then I' an Internet business because, because I want to have, I want, I need to have both sides of it.
- [43:26]
B
What's your real life business going to be? What are you going to do?
- [43:29]
A
It's going to be like a gym. Not, not like a gym, a gym, but slash, community connection. Like, I think people want to hang out in person. There's a huge. I don't have the. I, I don't have the perfect thing yet, but. And I never will. But there's like a Venn diagram that overlaps with like a lot of trends right now, which is like community run clubs, sauna golf simulators, hanging out in real life, book clubs. Like people want, we want to hang out with, not going away. We're going to hang there. And so I would love to own a. My wife and I were just talking about this morning. I'd love to build and create an awesome gym. Maybe there's an element to this where like a huge part of my youth was like growing up and playing basketball. My, my parents would drop me off on a Saturday morning and like me and like six of my friends, we play three on three or pick up basketball at the YMCA like for the entire day. And I think that's an important part of our culture and society that I don't want to miss out. And so I want to be a part of that. And I think it's. I also really want my kids to see me do something in, in person and not just sit in my office and talk. Talk to them.
- [44:27]
B
Do they listen to the podcast, by the way?
- [44:29]
A
Well, they're 8 and 6 and so my, they both just actually troll me. They're like, what the heck is B2B? Marketing and why? My son, my son literally said to me, he's like, dad, how boring is your job? You just literally sit in front of the commuter and you go, marketing. Oh, marketing, blah, blah, marketing. I'm like, fair, fair. That's fair. I'm like, but you, but you like the house that we live in, don't you, buddy? Okay, what's the flip side, Case? What's, what's the bear with the bear, case?
- [44:54]
B
Okay, so, I mean, I think the, the harsh. I think marketers are in a pretty good position, but we. Probably 80% of what we're doing today is going away and going to be automated, which means we need to like. And you're already doing it. Like, you're telling me you're using cloud code all the time. But like, we, we need to change our jobs. And even like, my job has drastically changed in the last nine months. I don't do 90% of what I was doing nine months ago.
- [45:19]
A
Well, right. I mean, you're, you're the perfect role of this. I was thinking about this. Normally it'd be like, you gotta know, you gotta learn SQL and you got to be able to, you know, know all the, all the like, technical stuff that you, that you went and got a degree in computer science. Because you're the only guy, an average Joe like me inside the company can't run the queries and get the data that you can now. I'm like, yeah, I used to troll my, my, my, my best friend, best man in my wedding. For years. I'm text. We had a, you know, we were in G G chatter. I'd be texting all the time, like, yo, how do you, how do I do a vlookup? You know, like now, like, that's a simple example. But for, for processing data, like, I can just do that in Claude without anybody else's help.
- [45:59]
B
Oh, for sure. And like, our marketers are very good at using Claude now, so they're vibe coding their own dashboards and apps, like data apps. They don't really need me for that stuff anymore.
- [46:10]
A
So.
- [46:10]
B
And I think everyone will face this eventually and you can either accept it now and be excited that you don't have to do shitty stuff like making dashboards anymore and then figure out what your new job is, or you can wait for the market to kind of dictate that to you. And I personally would much rather get ahead of it and reinvent myself now than wait for it to be forced on me.
- [46:29]
A
That's a good message. I like that. That's a good message because it's like, it's the bear case, but it's giving something actionable. You're not saying like, I don't know, go back to trade school and become a plumber?
- [46:39]
B
No, no, no. I think there is still place for marketers especially.
- [46:42]
A
So how are you? What have you been doing? So what is, what is Drew's like reinvention? What's different, you know, versus last year? And.
- [46:47]
B
Well, honestly, right now I'm focusing on how do I make our entire growth and go to market org agentic and so how do I help all these marketers get through this transition?
- [46:56]
A
What does that mean? What does that mean agentic, though? Yeah, everybody uses Claude. What does it mean?
- [47:00]
B
It means where we're going is like everyone on the team. Instead of like marketing and like going into the platforms and clicking buttons, they are building agents that do marketing at tremendous scale. All these marketers are shifting from running campaigns to building. Like they are becoming builders enabled by cloud code, but also able to build something that is much more effective than what a person who has no marketing expertise would build because they deeply understand the domain and they understand the platforms.
- [47:30]
A
Yeah, Dan and I have this. We work with this guy and he's just kind of like a moonlights on some side projects for us with AI and he was telling us the company that he's working with, they didn't want to go spend, you know, whatever money on some martech vendor for, for this tool that they were going to do. And so they just built it. And I'm like, oh man, this is the future. Which is like, wait a second, I'm not going to go spend, you know, 50 grand on a contract for this, this thing because I could just build my own agent that could do roughly the same thing. That's, that's an interest that was like, oh, shit, that's kind of interesting.
- [48:01]
B
Yeah, I don't think SaaS is going to. I mean, we're seeing that too. I don't think SaaS is going to die, but I think the margins are going to be incredibly compressed.
- [48:08]
A
Yeah, look, I, I could also poke holes in the argument that I just said. And I, I just wrote about this on LinkedIn the other day and people got really mad, really mad about it. But I was like, look, I do think there's a lot of performance theater right now where like people are spending weekends building. Personally, I would rather just spend the $20 a month for like calendly or whatever. Than like vibe code, my own calendaring app. And I think that's always going to be the case and eventually when that ship breaks, who's going to fix it? Yada, yada yada. We're probably still years from that. But you, you're going to mention a bunch of agentic workflows in example.
- [48:38]
B
Yeah. Let me tell you. And we're, we're honestly early stages in this. But let me tell you some of the cooler things that I'm seeing. The marketing team ship and these are things that like, to the point about marketers still having a place in subject matter, expertise mattering. I could have built these six months ago, but I never would have thought to build them. So what the real unlock was for us was getting everyone on cloud code, getting everyone in that builder mindset and giving them also honestly, giving them permission to stop doing their old day jobs. That had to come from the top, where we said, hey, we are okay with the numbers going down for a couple months while everyone learns this stuff and shifts into like builder mode. And this shit's been dramatic. So, okay, example one, it's called the vertical machine. I'm going to give credit to Elana on our team for coming up with this. So Elana runs our verticalization effort, which is where we pick a vertical, like construction, and we try to sell into it. And so that means, like, we just go super hard on, like, what are the pain points in construction? How do we build parts of our website to rank in SEO for those people and what they're searching for and come up in like AI agent results? What sales materials do we need? All of this, like, supporting stuff where you want to go sell into a new vertical. Also, like, what product do we have to build to like support them? And Alana started about a year and a half ago, did an amazing job on construction and got that like up and running. And then we were like, all right, what next? Should Alana move on to the next vertical? Should we hire three more people to each focus on three verticals? Like, what do we do? And instead of that, we said, let's build a end to end agent to replicate what Ilana did for construction. And then that will allow us to scale out to fifty or a hundred verticals with the click of a button. So concretely, what that looks like is an agent that you tell it, hey, I want to spin up a vertical verticalization effort for, let's say healthcare or medical manufacturing, whatever. It goes and does a bunch of research on, like, what are companies like in the space, what do they care about? What are their pain points? It uses like public resources but also like will scan our sales calls. Like we talked about gone calls before.
- [50:30]
A
Right.
- [50:30]
B
This stuff is like incredibly valuable.
- [50:32]
A
What's crazy is like that used to, it used to be like hiring product, like a, hiring a product marketer example. You'd be like we're going to hire a product marketer that is specific, that has deep experience in healthcare.
- [50:42]
B
Right.
- [50:42]
A
But it's like who has that experience? Realistically? No one. And something that I've always like had a beef with in marketing is like we, this is not an ageist comment, please don't take it this way. But like we have a 22 year old fresh graduate trying to sell to the CMO of a billion dollar company about an industry that they have no deep expertise in. Now you have this like insane product marketing assistant where I can go and I can go and do all the research, I can do all the sales enablement. So we're going to go to market in this vertical. I can have the messaging down, I can have the landing page. Right. I can have the deck right. Like that's a, that's a very cool use case.
- [51:16]
B
Yeah. Automatically. And I don't need to go talk
- [51:19]
A
to sales, figure out who's the healthcare, which healthcare guy are you going to put in a box in, in New York, you know.
- [51:26]
B
Yeah, maybe they will come up with those ideas. But we've got this machine. So it does the research, it then it goes into, then it goes into each channel and figures out like how do we market to this vertical. So for example, on the SEO side it'll go look at based on the research, it'll sort of be like what are the search terms or kind of keywords that someone in this vertical would be searching for while they're kind of like looking for our product. It'll go hit a bunch of external tools to get more similar keywords to that and then it will spin up hundreds of pages on our website that it thinks will rank for those searches in Google or AI searches as well to like seed there. And all of a sudden we have like our full SEO strategy just like deployed with the click of a button for a new vertical. And we do that across channels. Right. Then it's like how do you prep sales? All you can think about it for every channel. Now Elana has gone from like manually marketing to one vertical at a time to building this thing and continuing to add on parts to it so that we can click a button and market to new verticals.
- [52:23]
A
And is this all done within Claude code specifically? Or is it like the person owns this and there might be multiple tools that you're stitching together to make this happen.
- [52:32]
B
It's all in Claude code. But the key insight that kind of made this possible was sitting down with Alana and being like, what does your actual workflow look like here? And how do we decompose it into like, atomic pieces? And then for each of those, we sort of said, okay, here's an atomic piece. Let's go create a cloud skill for that. Right. And a cloud skill is like a reusable sort of instruction set for Claude that tells it how to do a thing. So you break it down into these small pieces, which you can write a really clear skill for with defined inputs and outputs. You can write evals, which are kind of like test cases on that skill, and then you can chain them all together and you get an end to end workflow. But the key piece is you're not just like, Claude, go spin up a vertical. You're like, here are all of these small pieces and they're really well tested and defined.
- [53:17]
A
What's been the enablement inside of the team to like, get everybody in the marketing team fluent with Claude. Claude code being able to build stuff? Was there. Did someone come in and do a training? Did everybody just learn it on their own? One person learning and teach everybody else? How is it working?
- [53:32]
B
It's been a journey. It's a mix of like a mandate from the top, all the way from the top of our company and through our head of growth as well. And then like, I think the thing. There's also like a bottom stuff.
- [53:42]
A
What, what's the. How do you paraphrase the mandate? What, what was the mandate?
- [53:45]
B
I mean, our CEO is literally like said what I just said to you. I'm okay with you dropping the ball a little bit on your. I'm paraphrasing. Sure. But like, I'm, I, I want you to invest time in this. Now is the time to do it and it is okay if that means you'd have to like, take away from your.
- [54:01]
A
Yeah, yeah. That's interesting because, like, I, I came up in an era where like, if someone kind of like rogue built a landing page for a new vertical, everybody
- [54:09]
B
like, whoa, whoa, whoa, whoa, whoa.
- [54:10]
A
Where does this fit in? But now it's like the way to learn is by doing. And so if Ilana wants to figure that out, she's going to go build a landing page. And like, I think part of it is letting go a little bit of like the guardrails from the creative and brand team and be like, well, this isn't perfectly on brand. It's like, well, that's not the point. Think about what we just spun up and we can always create a skill. We can. I'm sure you all have it. We can. We have a skill that has our brand guidelines in there and we're spitting out stuff that looks on brand total.
- [54:35]
B
What else? The other piece on enablement I think is really important there. I guarantee in any org there are pockets of people who are like super AI pelled in using this stuff and they might just be quiet about it. So we spun up sort of like a simple dashboard that's like, tells us in each team who's using it the most. And then we went and found those people and we were like, how are you using it? How did you start using it? And we turned them into kind of like evangelists for their team. So we were like, I basically went to all beat them over the head. I was like, record a loom video showing what you built and how you did this and share it with the team. Constant demos and like that bottoms up piece really work too. People get like super excited when they see someone like them being able to actually use this stuff and build with it.
- [55:14]
A
Yeah, for sure. I think the biggest piece on the enablement is you need to do it for yourself.
- [55:20]
B
Say more.
- [55:20]
A
You know, I think there's a lot of like, if I just listen to podcasts and hear people talk about Claude, cloud code, whatever codex, whatever you're building. But I think it's more of like this challenge of like, what's something that's happening in your. And this is like a call to action for somebody listening to this before we wrap. Like, what's something that's happening in your current workflow and in your life and marketing inside of your company. And like, you know what, I'm going to use this as an opportunity to go deep here and build this thing. Has there any been any like negative, negative pieces of this? Like if you have this org where everyone has Claude code, everyone's enabled to like create and ship stuff. You publish a landing page, it doesn't work or isn't on brand, or you send out a bad email. Like, has there. Has there been any negative, you know, or like something broke the whole. The whole marketing funnel one day?
- [56:02]
B
I don't know. This is a LinkedIn example, so I don't know how true it is, but it luckily didn't happen to us. But I saw someone who was like, yeah, I had like a Google Ads or maybe it was Meta MCP they built and it was, you know, bidding for them and then they, their whole account got like locked down. They were spending like over a million a month and they can't get their, you know, the support for those things is not great. They're like fully just locked out. Their marketing's off. So that's something to watch out for. Is like anytime these things are interacting, we classify things according to like, low risk and high risk. So if it's just like you're generating creative or you're doing something internal, that's low risk and it's like, go crazy. If your agent is interacting with the outside world or with customers, like sending them emails, for example, that's where we try to put some more guardrails on things. So for example, our Lifecycle team, super AI pill, they're doing a ton with cloud code, interacting with like HubSpot and sending emails through it. We don't really want that to be just like wild, wild west. So any HubSpot interactions go through an MCP that we've built internally that puts guardrails on things and make sure we're not like blasting the same person with 100 emails, things like that.
- [57:05]
A
So does. Does someone have to manually review that or if it, like, if it goes through that, it's okay. Like, has to. It's a like a code check of some kind to.
- [57:12]
B
Code check? Yeah, it's not perfect.
- [57:15]
A
Do you guys have opinions on Claude versus Claude code specifically?
- [57:19]
B
Right now we're in like, use whatever tool is best for you.
- [57:22]
A
Let me tell you where. Why I'm asking this question, like, yeah, please. Part of me is like, everything's just getting so good so fast. So so much better. I'm like, there, you know, you have Claude, you have cowork, you have Claude code. I'm like, you should see what I can do with just regular vanilla Claude. Like, are we eventually, like a year from now, is it just going to be like, there isn't Claude code or Claude cowork? There's just this tool and it does it all in there for like, I kind of. I keep going. Like, I feel like that's where we're going to end up.
- [57:49]
B
Yeah, wouldn't shock me. A lot of the marketers are using Claude Claude cowork. Yeah, it's sort of like, if you like that better, go for it. You can use Claude code through the desktop app now too. So some of them are doing that. We have some internal tools that make it easier to work with cloud code too. But, yeah, it's kind of choose your own adventure right now.
- [58:05]
A
There's also some cool shit. Like, there's some cool stuff. This is not, you know, I think I saw. I think maybe it was last week or so, like, somebody shared something about. Ramp is an interesting company of, like, having your own. I'm not a technical guy. Forgive me if this is the wrong way to say it, but, like, Ramp has their own MCP server now. Every company does. And, like, here's all the stuff that you can do with expense reports now. Because I can connect Ramp to Claude and I can say, like, you know, do these expense reports for me. And this is what. This is. What's. What's getting really fun. You know, our example, too, is, like, we use hubs. We're a HubSpot customer. We use HubSpot for our marketing automation. We use Claude. It's amazing that I can be like, hey, Phil, I feel like our. Our email list growth has slowed over the 90 days. Is that true? And I can get an amazing, you know, analysis from that, or I think it's cool to have this. I can be this super marketer where I have all these things plug into my AI tool of choice, and I have all this access to stuff being like, before, it literally be like, all right, let me go down to the third floor and find Drew. I'm gonna need him to, like, go find this data for me, because I want to do this thing.
- [59:07]
B
Great.
- [59:07]
A
All right, Drew, we're gonna wrap, man. This was awesome. Thank you very much for coming on the show. Big fan of yours. Fan of Ramp. Although my only beef with Ramp is since we moved the. We moved the company to Ramp, I get way less credit, personal credit card points, which is one of the benefits of being a founder. But I don't have the headache of, like, having to file expense reports or do any of that nonsense.
- [59:27]
B
Yeah. Would you rather get points or save your whole company time and money?
- [59:32]
A
It depends. I got a lot of points. I. You know, I can fly nicely with those points. You know, I flew my parents. I actually. We did. I'll give this. I flew my parents in first class on my points to our event. They've never. They've never flown in the. In the front of the plane before, and it was amazing, an amazing experience. I wish I. I wish. I wish I vlogged it it. But no, I'm just being silly.
- [59:55]
B
That is priceless.
- [59:56]
A
So I got Less points, but it's okay. Ramp has been great. And that's not even a Ramp ad. Just good, good product. And I also like the way you guys use the data. Speaking of data science, I think the best Ramp ads are, like, the ads on the podcast. Ads about, like, how Ramp only hires some absurdly low percentage of engineers, and therefore the talent is so good at the company, therefore their product must be great. I love that. Great marketing play.
- [60:20]
B
Thanks, Dave. Love it. All right.
- [60:22]
A
And look, I'm still waiting for, like, a Ramp hat or something. My God.
- [60:25]
B
You know, we'll get you some swag. Don't worry.
- [60:27]
A
God, no, I'm going to the wrong people, man. I.
- [60:31]
B
No, no, we'll make it happen. I know that. The swag person. So I'll go.
- [60:34]
A
All right. And I just shout out, shout out to the marketing team at Ramp. I know you all listen to this podcast.
- [60:39]
B
We do. No, they love you. I'll send you some screenshots. When I said I was going on.
- [60:43]
A
Yeah.
- [60:43]
B
Going crazy in Slack.
- [60:45]
A
All right, thank you for coming on. I appreciate it, Drew. Good. Good to have you on the pod. And I'll talk to you soon. All right, thanks, Dave.
- [60:51]
B
Have a good one.
- [60:51]
A
Yeah. Hey, thanks for listening to this podcast. If you like this episode. You know what? I'm not even going to ask you to subscribe and leave a review, because I don't really care about that. I have something better for you. So we've built the number one private community for B2B marketers at Exit 5. And you can go and check that out. Instead of leaving a rating or review, go check it out right now on our website, exit5.com. Our mission at Exit 5 is to help you grow your career in B2B marketing. And there's no better place to do that than with us at exit 5. There's nearly 5,000 members now in our community. People are in there posting every day, asking questions about things like marketing, planning, ideas, inspiration, asking questions and getting feedback from your peers. Building your own own network of marketers who are doing the same thing you are. So you can have a peer group or maybe just venting about your boss when you need to get in there and get something off your chest. It's 100% free to join for seven days, so you can go and check it out risk free. And then there's a small annual fee to pay if you want to become a member for the year. Go check it out. Learn more exit5.com and I will see you over there in the community. Foreign. Is brought to you by Converter. They're an enterprise lead data management platform. If you're running marketing at a large B2B company, like many of you listening right now are, you know this problem well. Leads come in from LinkedIn, webinars, events, content, and the data is a mess because one form captures job title, another one doesn't. One says United States, another one says usa. By the time it hits your CRM, records are incomplete, fields don't match, and your routing is broken before the lead ever touches a sales rep. It's annoying. And now that everyone's plugging AI into their tech stack, bad data isn't just an inconvenience, it's actually a real liability. Because AI is going to scale whatever you feed it, feed it garbage, it's going to scale garbage. Converter is the layer that sits between your lead sources and your systems. With Converter, every lead gets validated, enriched and standardized before it touches your CRM or marketing automation. This gives you clean data every single time. Companies like Microsoft, Amazon, Oracle and Stripe use Converter today. So if you're dealing with this, you're not alone, and there's a great fix. You can check out Converter right now at converter IO exit 5. That's C O N V E R T R IO exit 5 converter IO exit 5.