Stella: Why strong ROAS can hide what's really driving revenue

Welcome to Rebuilding SaaS Marketing, a podcast by Digital Hunch. I'm Renata, fractional CMO and co-founder of a marketing agency for B2B SaaS. Together with founders and marketers from across SaaS world, we are figuring out how marketing works in the age of AI. In each episode, we uncover one story: what worked, what failed, and what still keeps them awake at night.

My guest today is Brenden Delarua, co-founder of Stella. He thinks that analysis of which ad really brought you the revenue shouldn't cost $25,000 and a data scientist to run. Today we're going to talk about how to use TikTok for B2B marketing, what is the difference between what looks true in numbers and what is driving the revenue, And why he wants to democratize the incrementality testing. Hi, Brenden.

How are you doing? I'm doing well. How are you? I'm doing great.

Thank you for coming. I'm really glad having you on my podcast Yeah, I'm glad to be here. Thank you for inviting me Brandon, could you please tell me a little bit about yourself and about Stella? Sure, yeah.

I'm Brendan. I'm based out of Florida in the US and I've worked in paid media for most of my career, about 11 years now. And I've got a lot of experience running ads for various different companies, B2B, D2C, lead gen, commercial a lot of different brands. So I saw a need where larger marketing budgets are not properly or effectively measured to understand what's causing new growth to happen, new leads, new opportunities, new revenue.

And I found a statistical methodology called incrementality testing which I found was very important for a lot of the clients at the agency. So eventually, me and the lead data scientist at that agency, we left and formed Stella. So Stella is a B2B platform. We formed it a little less than three years ago now.

Stella's a B2B. We work with mid-market like D2C brands, so brands selling consumer products for the most part. And we help them measure their effectiveness of their advertising by helping them run structured geo holdouts, where we're turning off ads and looking at what happens to their source of truth revenue in the regions that we turned off ads. Hey.

I know that you talk a lot about incrementality, and you post every day, and you've been on a couple of shows, but I want to ask you how you came to that idea even before Stella. As far as I know, you hit every ROAS that client wanted. Every- everything was great in numbers, but you somehow understood that they are lying to you Yeah, yeah. So a lot of my experience obviously is in the D2C world, and now I own a B2B company to try to solve that need.

And part of that problem that I found firsthand from living it was a lot of marketers are assigned a ROAS goal, or in the B2B space, they're assigned like a CPL goal, a cost per lead. Also, for the B2B listeners, ROAS is return on ad spend. So in ad accounts, as a media buyer, that's what you are pushed to. A lot of clients will say, "We need a 2X ROAS and that's our goal."

Or, normally they say, "We can - as long as we-we're at a 2X, we can keep spending forever," which is not exactly what they mean. And and you can hit that as a media buyer. There are certain ways to trick and inflate ROAS. For example, I can start investing more in retargeting campaigns.

I can start investing more in branded search campaigns. That's going to capture people already on their way to convert, whether they're signing up or purchasing something, and I can increase vest - investment there and overall, it'll make the account's ROAS look higher, right? Obviously, you don't wanna do that for the client. You want to grow the client's account.

But when your job depends on a ROAS target, which a lot of marketers do and a lot of marketing teams do, or a CPL target, and you're very close to missing target, you're going to look for some quick wins, is what people call it in the industry, to try to hit that ROAS target. A, a thing that I refer back to in my LinkedIn posts or my TikTok content is there was, I was working at an agency. I did a lot of QBRs or quarterly business reviews with clients, and ultimately, I showed a deck where I showed like record-breaking ROAS across all the platforms.

I smashed target. I'm looking at platform-reported revenue, so revenue pulled from Google and from Meta and the charts are trending up and up. And ultimately, I get pressed a little bit by the actual CFO, who shows me the real numbers, the numbers that I didn't have access to, like the Power BI dashboard. And ultimately it wasn't as dramatic, but ultimately, the charts he was showing was the complete opposite of what I was showing.

The issue, though, was I wasn't lying. I was showing real numbers from Google and Meta. However, what I started to realize is that the revenue in those platforms are not correlated with true revenue, and I think that's what a lot of marketers are still believing or don't understand the concept of incrementality. The concept of incrementality goes deeper than just what can be tracked in an platform It goes into what's causing growth to happen.

So it's regardless of what your ROAS or CPL says, what's hitting the bank account? What leads are we seeing in HubSpot or Salesforce? If we increase spend on, I don't know, Capterra or Google Ads or Microsoft Ads in Arkansas, do we all of a sudden start seeing a bunch of leads from Arkansas, like qualified leads? It doesn't matter if they never clicked on the ad.

Are we seeing true impact from our marketing investment? And that's kind of the mission I've been on, is trying to educate marketers and business owners into looking at what's truly causing growth to happen, not just what the platforms say Could you give an example the situation that the light on all of this in your early career before Stella? Maybe with exact numbers - if you remember them Yeah. There's this one time I was a media buyer for a publicly traded company.

They worked with an agency I worked at. And I was held to a ROAS goal. I met specifically with the CEO, CFO, and the VP of marketing, like on a weekly basis. And I was like a - I wasn't a junior media buyer, but, I wasn't an executive.

And I would meet with them on a weekly basis and we'd talk through the ad account, and they're holding me to a ROAS goal. And I think that's just a flaw in how CEOs or CFOs or board members run a lot of companies because they don't know any better, and they don't spend a lot of time thinking about it because they have so many other things that they're doing. So they hold you to a ROAS goal. I meet with them weekly.

I'm showing them ROAS, record-breaking ROAS. They're a publicly traded company. I basically start looking into all their ad accounts, and we're running on a lot of different channels, and I see campaigns with high ROAS, and then I see campaigns with low ROAS. I do a lot of optimizations, but ultimately what I do is I divest from the campaigns with low ROAS and I invest in the campaigns with high ROAS.

That makes sense. That's the goal of the company. Their blended ROAS increases from a three to a six, and for any large like company that's very consumer focused or e-com, their largest expense is advertising. So I just made their advertising much more, quote-unquote, profitable.

It's a publicly traded company. Everybody looks good. The CFO's looking good. The C- CEO's looking good.

And even though I'm not getting credit publicly, the CEO and CFO and VP of marketing are all praising me of "This is amazing what you've done." And it was amazing for three months Where all of a sudden performance started to drop and I'm being pressed questioned and "Hey, we need to hop on an urgent call right now. Like revenue's down. What's happening?

What did you change in the ad accounts?" And ultimately it's like I didn't change too much that would affect revenue. And long story short, the conclusion we came back to was like I ultimately cut a lot of top of funnel spending that couldn't be attributed to conversions like linear television, CTV, display audio YouTube, like a l- like all the top of funnel things that are typically hard to track. And we were h- very profitable for a while because these top of funnel tactics have a lag effect.

They have a delay between someone seeing a TV ad to when they convert. So while we pulled back budget immediately, we were still seeing the benefit of all the demand that was generated from all the top of funnel efforts. So all of a sudden when revenue was starting to drop, we didn't know what it was because the change that truly caused that happened three months ago and we're not thinking about those changes. We're thinking about what h- what did we do in the last week that's causing this revenue drop?

And then it's the same thing but by turning it back on. It takes a little while. Like you don't turn it on and see sales immediately. It takes a little while for that lag to kick in.

But again, that was another way I was looking at in-platform metrics and it led me astray. And I think the biggest thing there is not fully understanding what ROAS or CPL is. It's a metric based on correlation. Someone saw or clicked an ad and then later converted, but it doesn't tell you what caused that conversion to happen.

So platforms like YouTube or TV - You can't click on a TV. I've tried for a while and it just doesn't work. But those types of channels get deprioritized because everyone is looking at some sort of multi-touch attribution dashboard and trying to attribute conversions and revenue to every single touch point, which is what I think is holding a lot of companies back, B2B and D2C companies. They're trying to over attribute and they're not fully understanding how how these platforms are tracking.

Yeah. In my experience, this is a never-ending conversation with founders or with business owners that how do we invest in something that we can't measure? For example, brand awareness, PR or something that we only assume that works, but we can't see that in numbers. So that's why my next question is my listeners are mostly B2B founders who don't have huge budgets because it's usually hard to track, and they don't have thousands of orders per month as D2C brands, for example, At what point does a company actually spend enough to track these things?

So what's the budget, you would say, or maybe not, it's not about the budget, but about the amount of information that you get? At what point you can really measure this incrementality? Sure. Yeah.

People ask about the rule of thumb. I think you're right with like you're hitting on something with the volume. Ultimately, what we do is complex statistical analysis, so we need a lot of volume. So if you're a lead gen company and leads for you are like $10, is your cost per lead, and you're spending $50,000 a month you can likely run a proper holdout study and get a statistically significant result.

Now, if you're a corporation, if your sales cycles are six months long or even more than that, and a deal value might be like $50,000 or $100,000, like longer deal cycles, that's gonna be much harder to track. So usually what we say is like for D2C brands, you should be making over 10 million annual revenue minimum. For lead gen brands, it should be over 50 million annual revenue And for B2B with like larger commercial contracts, it should be over 100 million annual revenue, which I know is like for B2B startups, that's a very high goal to get to.

But I think the issue is like there, there is a right way of doing it, right? And you use a platform like Stella to properly do holdouts. Even before you're ready for holdouts, the concepts of incrementality still apply to you. So it's really important to understand that like we should not just be grading things based on what someone clicked and converted, 'cause if we did, we would just give Google all of our money as B2B marketers.

'Cause that's usually where people go to search for things and search for products. Now, ChatGPT and Claude are starting to come up more and more like in the referral. But there's other ways to track, right? Like you don't have to reach statistical significance to turn on a new channel in just one or two regions and see what happens in those regions.

Do we s- do we all of a sudden just start to see more people reaching out to us from that region above what we would normally expect? Now, that's not a proper holdout, but that does give you an idea of, is this channel incremental or not? The other thing that I love that any brand can do is have a post-purchase survey or post-sign-up survey. So once a lead comes in, you ask them, "Where did you hear about us?

Like what caused you to convert today? Like what problem are you trying to solve that we, that you think that we can solve for you?" Or what makes you apprehensive about our product or our pricing?" Or like getting those questions, getting people to tell you where they heard about you is super important, especially in B2B.

If you're doing like field marketing, if you're going to events if you had a PR about you, if you post organically, there's so many different things. None of these things work in a silo. Nu-no marketing channel works in a silo. So it's good to have multiple data points.

So typically I recommend for any brand to have a multi-touch attribution tool. I'm not anti multi-touch attribution. I just think it has a place. And then definitely having a post-purchase survey so you can collect qualitative data on, the reason why someone is converting or booking a call with you.

I've even seen some p-post purchase surveys that'll say like, "How did you hear about us?" And someone might say like, "YouTube." And then the follow-up question will be, "Do you remember which YouTube ad you saw?" Um, So you can get more data of like what's actually working.

The value of that is like you might not see any clicks on YouTube videos. Usually, you rarely see a click-based conversion from YouTube. But you might find that YouTube's actually driving like 25% of your MQLs. Or you might find, maybe Google's driving like 80% of your MQLs, but that the MQLs that qualify to SQLs or opportunities actually came from YouTube, to where YouTube is driving more quality leads that come in.

That type of data you can't always see with a multi-touch attribution tool alone. Okay, and my next question is, you have a B2B SaaS, you're trying to grow. What would you measure in your own company, and maybe what you'd already measure? And do you use those recommendations that you just gave, for example?.

Yeah, tell me about it So yeah, so we use Amplitude, and it basically tracks everything between the website and and the app itself when people sign up for the app. We also have post-purchase surveys, and when someone signs up for a demo, we ask "How'd you hear about us? What other tools have you used? What did you like about them?

What don't you not like about them? What are you looking to - looking forward to talking about on this demo call?" So we can understand their pain points. So we have that infrastructure set up.

We are too small to run proper holdout studies, which is ironic, obviously, 'cause one, we're B2B, right? We're too small to run proper holdout studies ourselves. But to be honest, w-we don't - So we have the infrastructure in place. We do not look at the infrastructure as much, like we do, but the idea is we know all of this doesn't work in a silo, right?

I post a lot on TikTok. I have massive reach on that platform. From TikTok, like I've gotten random accounts reaching out to me, and they work at Meta on the Meta Ads product or Google at the Google Ads product, and I've been able to collaborate with them and grow Stella even further with like strategic partnerships. And then those strategic partnerships send us referrals, or they just list us on their website.

I post a lot on LinkedIn as well, my more buttoned-up content. Anyways, I have a - we have a very o-organic marketing strategy that seems to be working. But how do I tie back, is it a LinkedIn post that got you to convert, or did you see my TikTok and then follow me on LinkedIn? Did you do that and see like a PR that we did?

Did you see a podcast? There's so many things. So ultimately, we're doing - like we're not trying to pay too much attention to tracking individually, is this working or is that working? 'Cause it's all working together, and then we can start assessing the latest post-purchase surveys that we've gotten or the post-demo surveys to see if something's changed and we need to update.

Like lately, we have been getting more people telling us they've heard about us from Claude or a GPT recommendation, which is very cool, and that was new, and that's not something that was happening when we first started Stella. So now, like we have, a, a - one of our marketing employees working on AEO agentic engine optimization or something like that. I actually don't know. AI engine optimization?

I don't know it's G-O-A-O. Yeah. Way too many acronyms. But anyways so that's how we start pivoting our strategy, where we know our sales cycles take a while.

We're selling. Stella is the most affordable in the industry of what we sell. However, it's still relatively a high-priced SaaS product, so it's gonna take. It's not like an impulse buy.

It's gonna take some dis- like some time for decision-makers to f-figure out their need and then convert. And luckily, we've been doing a good job at our ideas, like filling up the umbrella of paid, owned, and earned media. As long as we're doing that, we don't really look too far into where are people coming in from, why did they convert, and, like, how do we get more of it? We're just continuing to push our content.

We're trying to provide as much free value, and it's working for us. Yeah, but the problem with that it's hard to scale after, like, when you're stuck, maybe you hit a plateau or something. If you don't know exactly what drives your growth, it's harder to scale. Yeah, this is m- usually the problem that I see Yeah, exactly.

You're exactly right. I think the idea is yeah, we're - right now we are growing good it's less of a concern. But if we, if growth starts to stall, we have all the data, we have that infrastructure set up where we can start looking into this, this data. But how we like to see it is we take a more holistic approach in how we grow.

So we have the infrastructure set up, but we are involved in so many things. We're running lead generation ads, right? That are driving top-of-funnel leads. From that, we have a nurture sequence.

From the nurture sequence, we're also providing free value, like free case studies. There's a lot of value that we're providing. Ultimately, what we do is very complex, and a lot of people don't want to do it themselves. So we're allowing the door to be open when those conversations arise, and that's what's been working.

So we can double down on that. We can increase budget on lead gen to fill the top of the funnel and start working them through with more email, more more value, more organic content, better organic content, which is part of our strategy for the second half of 2026. So that's the idea. The, the - instead of looking at individually what's working to scale, we bucket it into those three main media channels, paid, earned, and owned and we continue to increase it.

But yeah, I see what you're saying and maybe we should have a better plan. What would you suggest for that? It's, it depends. Usually, for B2B, and especially with higher ACV, higher than 2,000, it's usually mostly ABM marketing.

But ABM is a strategy as actually on my previous podcast, we discussed a lot ABM. And under this ABM str- strategy, you can get a lot of channels that just support this strategy. But yeah, mostly I would say that for higher val- value clients this is what you need to do Sure. Yeah we have a, we have an approach where we are looking at, The idea is we're trying to fill the funnel first, and I've actually, a lot of my TikTok videos I talk about stepping away from the funnel, and I go more towards this journey framework of early journey, middle journey, late journey.

But the idea is we don't wanna market to anyone that's not showing a signal to us that they need what we sell. So we have these very specific lead gen offers, where the only people that would, in their right mind, download these offers are people that are very qualified to work with us. And th- uh, they convert really well, so we have a large list of people, and, we QA the quality of people that come in, and they are all, very qualified buyers or potential buyers. And then that's where our marketing starts.

So we start reaching out directly to the companies that are are qualified that we do wanna work with whether it be, like, through an email, like a direct email or a direct LinkedIn message. At the same time, they're getting emails from us, and then they start getting into a retargeting campaign as well. So we do have a full funnel strategy like that. We just see it working as it's all contributing together because this is very complex technology.

I guess when I used to run a lot of accounts for B2B marketing, I would run into that issue where if we just look at what's driving growth, which is usually Google Ads or Capterra you run into issues where it's not infinitely scalable. Maybe if you're in a competitive industry and you can become the number one SERP position. But a lot of the Google demand would've happened regardless. And then there's also the constant conflict between sales and marketing of "I cold prospected tell me about it.

he, yeah he filled out a Google form and that's why it says Google, but really I drove that lead," right? So but the idea is like maybe it all work together. And I think, like, when you try to segment too much, I get the reasoning for it, but I see a lot of businesses get to a point where no one really n- can tell what's causing growth, and that's what stalls them. But yeah.

But to be fair Stella is not a $100 million SaaS company yet. But maybe one day we'll be. Tell me about your marketing team. So as far as I know that you have a really small team, so you do it all by yourself Yeah.

It's me and then I - we have a full-time marketing hire. Her name's Kaylee. She's crushing it. So we.

Basically, the marketing team is just me and Kaylee. And yeah, we do most of the stuff ourselves. Yeah, I do have a lot of partnerships with a lot of cool people. I've been in the industry and I've met a lot of really cool people that are advocates for Stella, so they'll post about Stella.

I just collaborated with some other really large marketing measurement companies on this like really large case study that will be put out and obviously Stella's logo will be next to their logos which will be very cool. So I think it's. It is me and Kaylee, but I think it takes a village to market a company, especially a SaaS company. And I'm lucky to have that village around me of like people that really like the Stella platform and they talk about it to their audience, which helps me expand more.

But yeah, the internal team is just me and Kaylee Yeah, as far as I can hear, you're doing a great job because sometimes companies with larger marketing departments can't handle all, everything that you're doing Yeah. I also think, to be honest I've been in the e-com space for a long time. I don't have an MBA, right? I have a bachelor's degree in marketing.

I think a lot of what does well for me on calls is I'm very informal and I make a lot of jokes, maybe some jokes that you probably shouldn't be saying in a work setting environment. But part of that is we're willing to post things that. Ultimately, all our posts, we try to provide as much value as possible, but we don't have a big QA process of we need to make sure the board members approve of this post before we launch. I get to post whatever I want, and I s- think there's value in posting things that might go viral for the wrong reasons or just getting a lot of attention on your name.

On my TikTok video, sometimes I'll post a really silly video that has nothing to do with marketing, and it'll get millions of views. And what happens is people click on the profile, they look at my profile and see the links to work with me, and then it drives more people to Stella and my consultancy as well. So even if I'm posting content that doesn't have anything to do with the company, I. It still drives revenue for me.

And I've worked at B2B companies where we've made really compelling ads. Maybe they weren't funny, but they were just very l- compelling. And then when you send it to the review process or the legal team, it just gets diluted and diluted. And by the time you launch the ad, it's much less effective than it was in the first version.

So I think right now we have the luxury of being bootstrapped to where we don't have too many cooks in the kitchen telling us what to say and what not to say, and it seems to be a net positive overall i'm reaching through LinkedIn, 'cause everyone on LinkedIn is pushing AI and buttoned-up posts, right?. No one talks about real failures. No one talks about being a human or the anxiety of running a company. So it's very easy to break through on platforms like LinkedIn just by being yourself, just by being human.

Which again, if I had a board of directors, they probably would not approve some of the posts I post, but they're all true. They're all valuable. They all have some sort of lesson that hopefully the audience can take away from it, and that seems to be working really well. Yeah.

And Kaylee helps me with that as well with what we post and how we post and all that. Cool. Actually, I had a podcast uh, with Carol a couple of episodes ago, and he said that every time he posts on LinkedIn, his investors call them and they say "Okay, you shouldn't post that because it's too informal or maybe it's too controversial." So you don't have this problem at least.

Yeah. And honestly I'm still new. Stella is less than three years old, right? Will this come back to bite me?

Maybe. But a lot of times when I talk to Kaylee when we're, like, making ads for anything, I tell her to look at our competitors' ad libraries. It's all public. You can go to LinkedIn ad library, Meta ads library, Google ad library, and just type in your competitor's name and see all of the active ads that they're running.

And I tell Kaylee to do the exact opposite of whatever they're running, 'cause a lot of times they are very buttoned up. They have to look professional. But what that comes off as is very boring. Humans like content.

Even if you're running an ad should be looked at as content. So I'm trying to get people to look and stop and care. I'm trying to relate everything to them. I'm trying to make UGC that feels like native TikTok content, where I'm talking about advanced statistical analysis that doesn't normally do well on TikTok, but it finds the right audience, right?

Like when you dilute everything, you drop into this like, bucket of noise 'cause you wanna look professional. But I think people wanna work with people. We're B2B, right? We're business to business, but really we're selling human to human.

And at the end of the day, at least in my industry, people work with who they trust. And if they get on a call with me, I try to be funny, right? But I know what I'm talking about. I've spent a long time building this platform.

I've had the problem that they're trying to solve right now. So I can talk, circles around, like, how our methodology works, how it's different from everybody else, why it's better and, like, why it's cheaper and more valuable and all the things you can do with it. So I think the content opens the door, and then we're able to close just based on our knowledge and the value we provide. I want to talk about the pricing.

You said that you looked at your competitors and you saw that they provide pricing for more than $10,000 per month, and you decided to go down significantly to 2K. Tell me how it worked out, because I think this is very interesting for B2B SaaS founders who are trying to solve this pricing problem, pricing mystery Yeah. I just, I'm competitive, and if there's a race to the bottom, I'm gonna win. You know what I mean?

I'm gonna release Stella for negative money. I will pay everyone. Everyone listening, I will pay you $2,000 a month, and Renata will actually pay for it. She approves of it, and she agreed to it before the call that she will pay for everyone's subscription.

No there's a few things. I think pricing in general with B2B. I worked for tons of brands tons of B2B brands where we would look at Hotjar recordings of their website, and the highest intent page was their pricing page. And a lot of pricing pages for B2B marketing is a joke.

They - It's just a form to request pricing. And, hate that Exactly. It's like a lead gen form. And, there's something to be said about the people who are requesting pricing, then maybe you follow up - the sales team follows up and hits them - gets them on a call, and then you go through a whole 30-minute deck, and then at the end of the call, you tell them the pricing.

Or maybe you have to schedule the second call, and then at that time, they're like, "Oh yeah, we can't afford this." Where that an- that could've been answered immediately by just showing the price on the page. I've worked for a bunch of B2B companies that had mainly one ACV. It was the same ACV across the board.

There's very rare use cases where they had a custom larger ACV. And like, why don't we just put that's the ACV? And they're like, "We'd be leaving money on the table." And I was like, "Literally everyone pays the same price."

I'm looking in HubSpot, everyone's paying the exact same. Anyways, that was always a point of friction, so when I got to make a B2B company, and I talked about this with my co-founder of like I think we just put the pricing on the website. Oh, and the other thing is like, clients would say "We don't wanna give out the pricing 'cause then the competitors can see it." I'm like, "Man, you are silly if you don't think a competitor already got a little weasel in there to find it."

Like you're sharing a PDF that's getting passed around. They're using it to negotiate with the other vendor that they actually wanna go with, like I would rather just put it on my website, and the idea is I know my competitors cannot compete with our pricing. Word on the street, I don't know this to be fact, so disclaimer word on the street, obviously they're VC-backed, we're not, but they are not yet profitable. So we know that they can't drop their prices as low as we can yet because they are probably under a lot of stress to get a return for their investors, right?

So since we're bootstrapped, we are not losing money by dropping our price low. We're not making as much money as we could be. But how that has worked so far is, one, it's disrupting the industry, which is what we wanna do. We wanna be disruptors.

And two, it is actually being effective. We're driving people into our pipeline. Now, for a correction on the $2,000, when we launched Stella, we were $2,000 a month. We actually have updated our pricing, so we don't have a $2,000 a month pricing anymore.

We have a free tier, completely free. So if you're doing holdouts or MMMs, media mix models, there's no reason to not be using Stella 'cause we have a completely free tier. We have a $750 a month tier that has a lot more features. It's very cool and very worth the price.

Then we have a $3,000 tier and then a $6,000 tier. The $6,000 tier is like very impressive with all the things you can do, and it's perfect for any brand scaling and really needing to understand what's causing growth to happen. And it includes a multi-touch attribution tool, a post-purchase survey tool, and incrementality as well, all calibrating each other. So it's like it all talks to each other.

It's all very useful. So that is our current pricing, and it works very well for us right now, where we get a lot of the refugees from our competitors that can't afford their 12,000, 15,000, $20,000 a month contracts that are locked in for a year. We tell them to use Stella. So like we've tried to prioritize just providing the most value regardless of the price.

Okay, but still tell me about the perception of the price Oh yeah. yeah Sorry, I know. I go off on these tangents. Feel free to stop me.

Feel free to just be like, "Brendan, stop talking." So we have had issues with the perception of the price, where the whole industry has been in cahoots of "Let's charge a minimum of $12,000 a month for holdout studies." And when we come in at $2,000 a month originally, or now, like that tier is ab- about 3,000 now people look at that price and assume we're worse. And we've signed some really big companies that in America that I look at.

Like I, there's tons of companies I love to, I, we work with that I love, that I've loved before they re- reached out, and then there's still like companies that we're talking to that like have. would be a dream if we could sign them. And a lot of them, a lot of the really big ones have told us, "We use the other guys, and we're looking for cheaper alternatives, and we found your site, and we're seeing like, we're being charged 25,000 a month, and then you're offering everything they're offering and more for 6,000, so we know it must not be good."

And one specifically booked a demo with us, and he said, "Hey guys, like great to meet you. I've been checking out your website for the last six months, and I have not reached out because I think the price is too good to be true." But like he. And then he mentioned I made a Twitter post that like convinced him to make a, to schedule a demo with us.

And he scheduled a demo, and we told him like, "Dude, if you're skeptical, if you use the other guy, take the data from their platform and put it into Stella for free, like we have a free tier and see how the results compare. And if the results are the same, if not better," better being more narrow confidence band for the statistical analysis Like, like you have it for free. You can keep using it for free. You can like cancel and just use Stella for free.

If you want more bells and whistles and more tools that help you actually scale, like we have like upsells for that, but like there's no reason you should be paying 25K a month for this analysis. I think what it is, and coming from a marketing background, is data science is very complex. It's very important. Marketers know it's important.

They don't understand what it is. So data scientists are able to charge a significant amount of money to make you think to have the perception of its high quality which is what I thought too when I worked at an agency. But now actually knowing how models are built and seeing people's MMMs, like people show me the that they paid like $200,000 for, and I look at it and it's not statistically valid. Like it would not pass the checks that we have at Stella.

But they don't know. Marketers don't know what questions to ask. They don't know the statistics, and I think that's what the whole industry has been based on for a very long time. So we price as fair as we can, right?

We've even been told to raise our prices, and maybe we should, but like that's not our mission. Our mission is to democratize the whole industry and that's what we're doing. But yeah, like the low price sometimes shoots us- is us shooting ourselves in our foot because people get turned away thinking we're too good to be true, which is something that we did not expect to happen. Have you changed something after that on your pricing page Yeah, so that's why we switched from 2,000 to the pricing tier.

So we have this ramp-up approach. We do have a package that's 6,000. It's still significantly cheaper, but it. $6,000 a month is still pretty expensive just relatively for But it's not so cheap not to believe in it, right?

Yeah, exactly. And then on top of that, we started offering full consultancy, 'cause what we've noticed is like even with doing a proper holdout or an or basically using the tool that we've built, it gets so confusing for marketers to know, okay, where do we invest now? We'll tell you how incremental something is, but not a lot of tools will tell you where to invest now. So we also offer hands-on consulting starting at $10,000 a month, so still cheaper than the c- alternative, where our team does everything for you.

So you don't have to lift a finger. You still get access to everything, but we do it for you. We run the models, we validate them, we ensure that they are accurate, and then when we present them to you, we give the interpretation of this is how you use this to actually increase incremental leads or conversions or sales. Which is a, again, higher price, higher ticket, but still much more value than what the competitors offer.

And I think our current clients understand that and are big advocates for us, which is great. I think there's a whole herd of people that are still outside looking in and wondering how we're able to price so low. Okay Yeah. So that's the.

We added tiers to where it's like, if you want to pay us $12,000 a month, if that's what will make you Please pay. We can. We will take $12,000 a month. We'll have a service for that.

Yeah, so that's how we've got counteracted it. Or if you wanna use it for free, we have that too. It's like we have a big range of a la carte buy what you need Actually, the free tier, I think this is something not not so usual in those kind of companies that charge a lot. Usually you only can enter the platform if you pay.

So I think this is a really nice view that you want to democratize this Yeah, we also think like what we do is just so complex and maybe other, like your listeners like have SaaS tools that are so complex. I also think when we're building the tool, like me and my co-founder, and then we have a team of developers on the Stella side that help us as well, we are in the platform every day and we build a tool and we build modules. And every additional module we build based on like we get feedback from somebody that like, "You guys should add this," so we add it.

Every time we add something, we get more support tickets of our current users confused. And it's not even like they're just interested to learn more, they like stop what they're doing and they can't move forward with our tool without understanding what this one scorecard says. Is this good or bad? And it's you gotta chill, dude.

Like we just made up that number. Like that number is like we just completely made it up. Kidding, kidding. But anyways, we'll like.

One, one thing like we introduced like comes some p-val, right? A p- a p-value, and it was tripping up a lot of people. But we try to be as transparent as possible with the data science and surface all the validity metrics. Anyways, we've had to play a delicate balance between simplifying the UI but still allowing marketers to have all the information they need to interpret the data, like whether they screenshot it and put it into Claude or GPT.

But we're so close to the product that we understand how to navigate around it very easily but the end user doesn't. So we try to keep it as easy as possible. But the idea with the free plan is they can get their feet wet, they can see if it works. It's still the same models we use in the paid plan, so the model, like the accuracy of the data science is still there, but we know there's friction.

There's no data ingestion. There's no - Like it's much easier to upgrade, right? And that's what we're selling is the convenience, right? The free tier has everything you need to run proper holdouts.

You're just gonna have to pull the data manually. You're gonna have to format it. You're gonna have to upload it. You're gonna have to check it yourself, right?

Our paid tiers reduce that friction, so it becomes easier and helps you become better. So that's kind of the idea b-behind the free tier, that it gets people into the door using our platform, and then they start noticing like friction that they can pay to reduce. So then it upsells them over time. I also really like that certain companies with complex products started to introducing this role of implementation engineers.

I heard about it couple of times. This is not a customer success. This is something more, more technical, that these people are technical, but they help clients to set it up and to understand the data or to whatever they do. So I think this is also something that is very helping for the success metrics and for churn and everything Yeah.

Yeah, we're seeing that too. I guess with my previous point was like sometimes people get into the platform have no idea what to start with, and in my head I see it, I'm just like, "How do you, how are you not understanding this?" And I think as like product owners or the CEO of a company, th- when you're building a SaaS, like it's hard to disconnect yourself from the product and try to simplify as much as possible. But having those implementation specialists is like perfect.

We're a small startup, right? So I'm basically the implementation specialist. We like hop on calls with a client, make sure they're set up for success. For every client, even if they're not paying for additional support, which we offer, the first 30 days we're pretty hands-on, 'cause obviously we we want them to know how to use the product so then they keep going.

We open up Slack channels, we have op- open communication. We'll go into their platforms and help them run everything, make sure everything's connected properly. But we still see this learning curve. But we have some brands that have been with us for over two years, so like basically the whole time we've been a company, and they are like wizards at the platform.

It's very cool to see that we've built a platform that they just completely understand and they feel comfortable in. But then, we just signed clients like earlier this week, last week, and they're starting their journey and they're trying to get acclimated and see if Stella is even the right tool for them. But yeah, I think customer success, there's just so many parts of any business, especially a bootstrap business, that like there's customer success, there's like implementation and obviously churn with like a SaaS product, a subscription SaaS product.

Like trying to mitigate churn is a big part of that. And ultimately our philosophy is like, if the people are not getting value out of the product, they're going to churn, and that kind of transcends anything. So what we talk about in our weekly syncs and our standup calls is like the value we're providing to clients to make sure they're using the tools. And not just measuring incrementality, but increasing incremental revenue or incremental leads, 'cause that's ultimately what we're after.

We're not after just measuring your marketing, we're after helping you actually grow and scale a business. Sounds amazing. As the last part, I want to ask you, of course, about AI. How, measurement is changing with AI.

Do you see. Because if Stella is around three years ago, you started in the early days of AI development, I would say, and how do you see it's changing now? Yeah, so I think there's a few things in the industry. I saw a tweet from Sean Frank, the CEO of Ridge, that said, "I would hate to be a SaaS founder right now because of AI."

And that's always scary to read. So there's a couple things with AI, right? We - AI helps us move quicker. There was a period in time where we gave all of our develop- we still do our, all of our developers have the max plan on Claude and everything el- whatever other AI tools they want, we'll give them And we were like I don't know if like vibe coding 'cause they're real developers, but we were like u-utilizing Claude a lot and for PR.

So we're pushing PRs to GitHub for my co-founder, Vinny, to manually review. And when he was manually reviewing even though the UI looked great, the back end looked atrocious for a lot of PRs. And my co-founder would spend a week or two weeks like updating the PR manually before pushing live, while at the same time, our developers just kept pushing PRs and just hitting with tons of PRs. And what we do, like the UI might look simple, but there's a lot of back-end infrastructure that goes into what we do.

Obviously, data science and storing and models and runtime. Like there's so many things that we collect on the back end that the user doesn't see. So and that's with any SaaS tool. So anyways we had a big conversation with our developers to stop using AI as much because we're like the.

It looks like it works, but when we look in the back end, it is terrible. So we know that there's a lot of people vibe coding tools that that are not really concerned about the back end and they will break. And yeah, we've had the ability to vibe code all of Stella. It's taken us three years to build what we've built.

So if you think about it that way, like this is all we think about. If we wanted to vibe code and just push it really far, we can. That's just not the right way of building any proper SaaS that's going to last forever. Stella has no downtime.

Like we never really break which is cool. Our models run flawlessly. We've built everything as perfect as it can be. Also, my co-founder is a perfectionist, which sometimes we argue a lot about things, but hon-honestly, he's right about certain things.

Anyways, so that's like the building aspect of AI. When we built Stella, we had a lot of AI in Stella not to build models or do an not to do the statistical analysis, but to interpret it. So when you get a lift report it says Google Ads drives, you know, a $70 cost per incremental lead, right? Like what does that mean?

And then based on this data, like can I trust this data? Like what's the validity metrics like MAPE, like mean absolute percentage error, or R squared or stat sig. Like, how can I trust this data? Like what confounding variables were used?

So we have an AI layer within Stella that can interpret and explain to you as a marketing leader, what does this data mean and how do I use it? So that's what we've kept the AI within Stella for. We used to brand Stella, as most startups did, as, an AI marketing measurement platform, when really it's not AI. All of our model - like none of our models are AI.

The interpretation layer is AI But now everyone's talking about MCPs. I was trying to think of a funny joke with MCPs, 'cause like MCPs is now like the buzzword that everyone's just building MCPs and they think they're really smart for building an MCP, when really it's just a collection of APIs. But um, we have another update in Stella's AI where we h- have connected internally an MCP so the AI can control our own model. So the UI within Stella can all be controlled with the AI, which makes it very easy.

You can say, "I need to run a holdout study. This is the situation I'm in." So the AI can help you structure that, and then it can build it within Stella for you, and then you can toggle over to that tab and see the results, 'cause it's connected to your data and it's connected to all of our tools within Stella. So that's a layer.

So it's still an interpretation layer, but now it's selecting your test and control and date range, like all that stuff can be done by the AI very easily now. And then we've released the MCP publicly, so you can pull it into your own Claude or GPT to use Stella. Yeah, without MCP, I think nowadays it's almost impossible to survive because everyone wants M- MCP. Even we in our agency, we built our own MCP even though we don't provide any products Exactly.

So our value is we think St- we think of Stella as the brain. We don't think of it as necessarily like a marketing - Like, it is a marketing measurement tool. But like I said before, I think any business, D2C, lead gen, B2B, whatever it is, there's three pillars to understand what's causing growth for your business. It's multi-touch attribution, it's post-purchase surveys, the qualitative data, and then it's causal data.

Stella has tools for all of them, and if you're running holdout studies and you're calibrating the other tools and you're running MMMs and you have the post-purchase surveys running 'cause we start tying back each post-purchase, post-purchase survey with an actual lead or an actual order that came in. And then we have your multi-touch attribution data that is like storing every touch point someone has. When you query that into Claude through our MCP, when you are talking to your data, Claude has full context.

So a lot of times when you're like talking to Claude about "Where should I reallocate budget right now?" It might start doing the original mistake that I made at the, when, at the beginning of this podcast with investing in high ROAS or high C- or low CPL channels and divesting. For example, like I've seen it say, "We need to cut YouTube immediately. We're spending 1,000 bucks a day on YouTube and it's driving like a 0.

2 ROAS." But when it has context of our post-purchase survey and it sees that it drove 10% of the revenue that we've driven in the last 30 days, that flips the s- the script a little bit. Oh, based on that that's a 2X return based on qualitative data. And then if we're big enough to run a holdout study or a media mix model, it can look at that data as well to be like, "It looks like TV or CTV is, or YouTube is driving revenue even though the multi-touch attribution data is not saying it is."

So we have three forms of measurement that can all be used within Stella's brain that you can pull into however you're managing ad accounts or leading your team. So I think that's like the main value of Stella is like it's combining everything to give you the mo- more informed decisions to actually grow the company. Yeah, and after if you pull it to Claude that has all the context about your marketing campaigns I think it's just multiplies the effect Yeah. Actually, someone, just a funny antidote, I know that was your last question or anecdote.

One of our clients pulled in from the MCP basically, like, all of the sales that are last touch Google or whatever the platform. I don't forget what the actual platform was. He asked to query, and by query, just prompting, right? There's no SQL anymore.

He asked all the conversions that came in from Google or from multi-touch attribution, how many had a post-purchase survey that was different from Google. And he was able to pull all that from Stella, which was awesome to see. It was just a cool use case we haven't thought about. Obviously we built it to allow for that.

But he told us this is what he's using it for, and we're like, "Wow, that's really." We didn't think that you could do. we never thought of that, but that was a really cool use case where you can like. All of these would say Google drove the sale, but post-purchase survey saying word of mouth, YouTube uh, Yeah LinkedIn, organic, podcast, PR, like whatever it might be, and then Google takes the credit.

And then you could do that with any other platform. So it's really cool to see what our users are doing with Stella. Yeah. But actually, I just thought about this post-purchase service.

I sometimes hear the objections that users just push a random button or choose And they do. Yeah They do. So there, there are going to be like some. that's why you can't trust post-purchase survey data alone.

That's why it's like multi-touch attribution, post-purchase survey. Because you might be able to see like their UTM was attached to a TikTok ad, but then when they click it, you can see. they might click the first option, but there are also ways to randomize how the options show. But maybe they clicked the first option, which was meta ads, and you can see their UTM was only TikTok, so that's.

you don't always trust it. And then that's why you do lift analysis too. So that's why you need all three together, because there's pros and cons of each one. Even incrementality isn't flawless.

One of the biggest things that we talk through with clients is incrementality is a sta-statistical analysis. But if I give you a number as you're used to seeing in like a dashboard, like IRO as incremental return on ad spend is four, that usually comes with a range above and below, plus or minus two maybe. So that means the true incremental return could be anywhere between two and six, when really marketers are focused on that middle number, which might not actually be the truth, right?

So there's pros and cons. Incrementality gives these confidence ranges. Multi-touch attribution deprioritizes view-based channels like CTV, linear audio, and then post-purchase surveys can be full of bad data if people are just not clicking properly. The assumption is hopefully they are filling out things they're filling out things properly or more people are filling things out honestly than they are just clicking through to get through.

Okay. Can you somehow measure the performance of AI answers? Like for example, if ChatGPT recommends your product Yeah, that's very interesting. I know some platforms are doing this right now, but I mean in the bigger picture Yeah.

So with Stella's framework, right? We can pick up the multi-touch attribution if someone was referred. We can look up referrer ID, so referred from GPT or referred from Claude, so we can see that. The post-purchase survey, we can also see that too, if they do honestly say that they heard about us from Claude or GPT, right?

On the causality side, we can run what's called a causal impact analysis. So causal impact analysis looks at before and after a change was made. Usually you might run one if a PR article went live or an influencer posted about your brand. You kinda model out everything before that.

The. When they post, it's called the intervention period, and then you look at what happened after that. Was there a lift above what you would have predicted to continue? So with GPT, what you can do is maybe pull a report from Amplitude or from um, GA4 to see all of the sessions that are coming in from GPT and see if they start trending up, and then you run a causal analysis.

So before and after that first session from GPT or whenever you think that intervention period should be. And you can see is there a statistically significant lift from from GPT going live compared to maybe sales. You're translating it from GPT sessions to sales. So there's ways to do it, but again, that's a flaw of causality.

Some things can't be measured properly that way but it can with post-purchase surveys and multi-touch. All three kinda tell the full picture this sounds really difficult for me. And I know why a lot of marketing, m- marketers struggle with this, even though I know statistics a little bit, at least I understand how it works. But yeah Yeah.

I hope I I hope the listeners are following along. I. again, and this is the part of it, and I think this lesson that we can all take as founders is sometimes we're too close to the product. Sometimes there's too many acronyms.

Sometimes I'm assuming you know the things that I already know that has taken me a long time, and I don't go back to explain what I just said, 'cause I just assume you know it, and not just you, but any. obviously the listeners too. But I think that's a good just lesson to take a step back, take a breath, and realize we have to keep it more high level when we're explaining it to people, when we're, making marketing material to promote, when we're making ads, that people don't always understand the advanced terminology that any industry may or may not have.

Obviously, most marketers understand that, but I always think you might get lost, like how I think maybe I just got with you're just over-explaining the methodology and people are just checked out because you're just saying way too many big words. Yeah. What you just said, this is the golden rules of marketing. Your audience is not you, so this is the main point.

Yeah, exactly Brendan, thank you very much. Do you want to share something else maybe that I haven't asked and you want to add it? I think it's y- I guess on the incrementality side, another thing with AI is it's also de- democratizing the space, where you can just ask ChatGPT or Claude, "Run a incrementality study on this data set." I think they're.

Obviously I'm gonna be against it, right? 'Cause I make money from you not doing that. But just to be honest with everybody any model will output a confident answer even ChatGPT. Even last night I was talking to Claude about something unrelated, and I read it, and I was like, "That doesn't seem right."

I'm like, "Can you check your work?" And then Claude was like, "Oh yeah, dude, I was completely wrong. Don't do that." And I was like, "A second ago, you were so confident."

But I've been seeing a lot of people saying why would I pay for Stella when I just have Claude or GPT?" And it's like any model will output something confident. It'll tell you a confident answer, but how do you know if it's true or not or validated? And if you're moving hundreds of thousands of dollars, which our clients are, based on the result, like you probably wanna be sure about it.

Or you probably wanna have a scapegoat that it was a third-party measurement provider like Stella, rather than just saying, "I, I guess I just did it in Claude or GPT." So I'm seeing that a lot. I also see a lot of people try to do this analysis in Google Sheets, which I think is wrong. And the idea is "Brendan, you sell incrementality.

You're biased." That's why we put out the free plan. You don't have to use. You don't - Use our tokens.

Like you can use the AI within Stella on the free plan to run the analysis. Use our tokens. Don't pay for your own tokens. And then you don't need it to do it in Google Sheets because it's free.

You can do proper holdouts or proper media mix models from Stella for completely free. So I think like I, I encounter a lot of marketers that are like, "No, we're doing holdout studies." And then when I see their methodology it's a lot of Google Sheets test minus control, which is just not the way of measuring lift. And what's worse are there's some agencies doing that, that sell it to their clients as like a $20,000 incrementality package.

And I obviously don't post about that publicly or call them out, but I definitely don't agree with it because the client, the end client doesn't know what to ask. They don't know what they're being deceived, and they're paying a lot of money, so they're assuming it's done right, but it's really not. So I think there's like a low barrier of entry, but to do it right, you have to use a tool like Stella or even use our competitors. I'd rather you just use a real tool than try to do it yourself, because there's so much data science and nuance that goes into all of this that really matters if you're trying to grow your company.

So that's maybe the thing I want to leave the podcast with. Yeah. Thank you. I agree.

Even though now I think AI is better with numbers than before, because before it was just terrible. But yeah, I also don't really believe that, especially with everything that is about data analysis and, with ML, that you can trust it 100%. Even not with numbers with everything. Yeah.

Yeah. I love like, it's a funny thing when you're like, "Hey, Claude where should I invest or where should I scale?" And then it gives you a recommendation, and then say, "Are you sure about that?" And then it says, "Nah.

Yeah, I was wrong." It's like, you were so confident a second ago, but that's my favorite thing to do now. And then I'll screenshot it and send it to my team and be like, "Look at what Claude just said." But yeah.

And I guess the last thing which I think is im- important, I make a lot of content online and I post different things on different platforms. On TikTok, I do post marketing things that are valuable for founders and marketers. However, I also post very silly things on TikTok, just things that I think of off the cuff, things that I probably don't want my boss to see, but luckily I am my own boss. If I had investors, they would tell me that I probably shouldn't be posting what I'm posting.

I just wanna make people laugh and make myself laugh. On LinkedIn, I'm much more buttoned up where I'm posting real value. So if you're a serious business owner and marketer, follow me on LinkedIn where I'm gonna post real value and I'm not gonna But still you post memes and everything. Yeah, I s- It's hard to get away from it.

That's just who I am. I'm a little silly. Yeah so basically that's usually my calls to action is LinkedIn if you want serious content and TikTok if you just wanna have some fun. Great.

Thank you. Thank you for saying that. This was Rebuilding SaaS Marketing. If this episode gave you idea or spark, share it with a founder at the same stage of growth.