Decoding Marketing Attribution in the AI Era with Mada Seghete and Alex Bauer

This episode is brought to you by Boast AI. Each year, the US And Canadian governments give out billions of dollars in R and D tax credits and innovation incentives to fund businesses like yours. But the application process is cumbersome, prone to frustrating audits, and receiving the money can take up to 16 months. Boast AI gets you access to research and development tax credits and innovation funding opportunities without the headache and red tape. Join thousands of North American companies leveraging boast AI software to maximize cashback. Check out www.boast AI.

It's almost impossible to find the right person who does it. All right? You usually find someone that's really good in one area, you fix that area and then you find the next person for the next area. So I think you can't actually say that there's still going to be a change in leadership, but I think what we can come in is if you have the right data, when there is a change, the new person knows how to they won't do exactly like your story, right? Someone came in, they bought the company, they didn't have the data, so they made the wrong decision. So if we actually can come in and always be the right data, even when someone leaves on their own accord or whatever, the organization has the data. So when they make a shift in leadership, that the new leadership has all the information to continue and to make the right decisions moving forward. Instead of being like, oh, I'm looking at the. I don't have the data, so I just assume everything's not working. So I'm going to try everything again. I need some traction. You need some traction.

Hey, what's up innovators, entrepreneurs, visionaries and disruptors. This is your Traction podcast host, Lloyd Lobo.

We're a community of over 100,000 people.

Just like yourself on a mission to help you get the methods, the money and the madness to explode your business growth.

Featuring stories and tactical advice straight from.

Those who've done it before, like Shopify, Twilio, Asana and many more.

Hey Traction Fam. This is your host, Lloyd Lobo. Super excited to have a repeat over and over again. One of my favorite speakers, Mara, co founder, previously of Branch and now something new. She's recently launched Upside and Mara is Badass founder. I think you know her last company. She started from the dorm room, went through multiple pivots and then saw it through a 4 billion valuation managing GTM. Building the culture along the way. And the last time we had Mara on the podcast and at our attraction conference, we talked in depth about the pains of Multi touch attribution and how it is so hard to connect the things that you do online with offline to drive real results and business growth because the teams around you, the board and everyone doesn't understand. So super excited to see Mara solving that problem now with upside. Mara, thanks for joining us. And you have Alex with you, your co founder who used to work with at Branch. So give us the story and how it all came together.

Oh my God, Lloyd, I'm so excited to be here today and always enjoy our conversations and so happy to be here today. Yeah. So as I grew Branch worked there for 10 years, took it from the very early stages to over 100 million in revenue and really was we started as a developer tool and then we started selling into large enterprises. Our deal sizes kept getting bigger and bigger. We hit the 1 million deal, $3 million deal. And as I was running marketing and getting involved with a lot of the go to market, it became harder and harder to understand what works. You got this like really large deals with a lot of people involved, sometimes five, sometimes 10, sometimes even 30 people on a customer who are like trying to make a decision and you influence all of them. Some of them came to my webinar, someone went to a dinner, others were replying to emails from AES and we're like, we think these things are working, but what's actually working. And with Branch, we were solving attribution for mobile companies and we were an attribution company but we couldn't solve our own attribution. And we found this thing where every team was actually like trying to use their own metrics. So we came, I remember this one time we were all in a meeting and every leader of every team, including myself, was coming and saying I drove this much pipeline and I drove 2 million and someone drove 5 million and someone drove 4 million. And Alex, my old co founder added it all together and he said, this does not add up. We only have half the pipeline if you add all these numbers because you're all double counting. And we were all trying to take credit and there was a lot of infighting. So I remember Alex here was on my team. I hired him eight years ago and he was one of the smartest people I know and he was also very data driven. So I went and I was like, Alex, there must be a better way of doing this. And he came up with a different model, so I'll let him talk about it.

At the time we were using, I think what's probably pretty standard in most enterprise, B2B teams, which is we were trying to use a deal source model, which is for those who aren't familiar, you try to figure out whatever came first, whoever touched the deal before anyone else. I think some companies even have this sort of rolling timer where one team can claim the deal and then someone else has to wait 30 days before they can claim it. And we were trying to use something similar to this, which is we'll just figure out which team was first. And then we were pretty quickly realizing that some of our best tactics would be something like marketing, hosting, hosting a fancy dinner at a nice restaurant. But the best way to get people to show up to that dinner was to have all the SDRs send personalized invitations. And in that situation, you've got a chicken and eggs problem, you've got the dinner and you've got the invitation. And if the invitation gets credit, then the dinner doesn't. But if you give the dinner credit, then the SDRs don't want to send those invitations anymore. And so we were coming up with these much more complicated rules. If there was a dinner but there was an email from an SDR within a week before, then we're going to split the credit for that deal 60, 40, or something equivalently complicated. And so what we figured out was that if all of these touch points are happening, then they all are having an effect. And the job of the attribution model isn't to say this deal, which is maybe going to take 18 months to close this deal, is entirely because an SDR sent an email once that's not useful. You need a model that is trying to understand all of the complexity and say, these 80 touch points, they all had an effect. They all deserve a credit, but none of them deserve all of the credit. And our breakthrough was realizing that we're already collecting this data. Much like many companies, we were even already giving it a score in some cases because we had a lead qualification model, which say if you get enough of a score, then it's a qualified lead. So why don't we build on that scoring model and use that for attribution instead of trying to say what was first? And so that was the beginning of the model. For a long time, it was just a bunch of hacky Python scripts that lived on my computer and I would run them once a day. And then eventually we had all of the teams using this data, so we hired somebody to take it over. And then once we came to upside, we actually had a real engineer who took all my scripts and rewrote them in an Afternoon. Because I'm not a programmer, but that is how we built the first version of this model.

Oh, that's a wonderful intro. But in those three or four minutes, I saw 15 years of attribution, PTSD in front of my eyes. Right? Because like you guys at branch, you guys drove a lot of growth through events and community. And I host my previous company, we got from 0 to 10 million ARR in 3 and a half years by building a massive community. And a lot of it was events and, and we agreed to just do last touch attribution. So last, who is the person who dipped it in the favor of the close? And that's what we tell the whole.

Story ever, especially in larger deals.

100%. Right. And so then what happened, Mara, is this is, this is a very funny and interesting story. We would put the lead source as the first touch and then the last touch as the key attribution point that got all the credit. And we built this massive community. And not too long ago we sold majority of the company to growth equity. And the new leadership, of course, when you transition out as founders, they come in, right? And now they're looking at all this attribution and they literally assume that the community that we took years building, which actually helped US bootstrap to 10 million ARR with zero funding, with less than 30 people, with no marketing people, it was just the community. And the new leadership actually thought that it drove no value. And it was very hard to prove because there was no data. The positive side of it is I inherited that community then because they thought it brought no value. But that's why solutions like this need to exist. And I'm so excited that you guys are building this.

Yeah. And it's interesting. There's this idea of credit versus attribution. Right. So people sometimes don't like the idea of credit because when you think about credit is like, oh, I just want to get credit for my work. And I think sometimes if you think about, oh, there's a tool that gives people credit, there is like the dark side of that where people just care about the credit that they're getting and they're not really thinking about how can we work together on things that work. But the flip side of that is if something doesn't get credit for the worst, that is building to a team, then you don't actually do it anymore. You might let go of a team, you might let go of a community. And, and I think there is something around like with the right attribution, the credit is given for the thing that actually deserve credit. So it's not about getting credit in general, but about giving the right credit to the people, the campaigns, the thing that deserve it. Because when you do that, those people feel a sense of purpose, that they're doing the right things and they will keep doing it and then you get to the right outcome. So I think it's an interesting balance. You can like use it only for credit. Attribution should come first, but credit is a secondary part and giving it makes a lot of it's important.

Alex, you said an interesting thing. You said, I'm not a developer, but I wrote these Python scripts. So you are a developer Because I'm an engine software engineer and I don't think I can put myself through to writing a bunch of Python scripts to solve a problem if my life depended on it. But you mentioned we talked about credit and we talked about attribution. What is the difference? I think most people don't even understand the difference between credit and attribution, so how can they even think about doing it the right way? I think that's a good starting point when we say credit, like credit to individual versus like attribution to a specific activity kind of thing is what I'm saying. Right.

I think the distinction is in how you're using the data. Because attribution, when it's used correctly, is almost like a diagnostic metric. It helps you understand how the machine is working, the machine being your go to market organization. It helps you understand where things are running smoothly or where there might be issues that could use some attention. And one byproduct of attribution is giving credit to perhaps the person or the team or the activity that received the attribution. But where things can sometimes go wrong is if the teams start claiming credit as the primary purpose for the attribution, rather than looking at it as a piece of information that helps them do their jobs better. Because then you get into these credit claiming wars where if we go back to that example of the SDR and the IP dinner that marketing was hosting from earlier, you'd have the SDRs not being willing to send email invitations to the marketing dinner because if they do that, they'd have to share credit, they're not going to get as much credit. Whereas if you look at attribution holistically, then you understand that this is like a compound play. It's the invitation plus the event that makes a result. And if you take any one of those pieces individually, then everyone's going to end up Aurora. So Attribution becomes a useful input, but credit isn't the purpose of it anymore.

This is fantastic. I think the whole world should be jumping on it. I wanted to ask, like, why now? But I have personally faced this problem across so many startups and companies, not just startups. Over 15 years and a couple of boards I sit on, it's the same thing. So why now is a very interesting question. I don't think there's anything that does the job out there, but would love to understand how did you. As you. You stepped out of Branch, right. And Branch has done well. Why do this? Over a hundred other startup ideas you could have had, Mara, because you're a great entrepreneur.

I just felt a lot of passion around this. I think my personal story around, like Branch and like really understanding myself as a leader and knowing what to work on. And I think that even this idea of like understanding if the stuff my team is doing is actually working well, I think a lot of marketers and I've talked to a lot, so I do come of it from a marketing perspective where a lot of marketers work on things and they just believe it works and they have a really hard time understanding of the stuff that they're working on is actually bringing value to organization. And they also have a hard time getting like, you know, a recognition on those things from the organization because sometimes the stuff that they're doing, it's hard to like, tie it to. Okay, this actually helped drive pipeline. Everyone wants more from marketing, but then they're like, oh, but I don't know if this is actually working. So I think I felt this needs this truly on my own. So I think as we build this internal tool and I saw both us improving how we're doing, go to market and seeing how my team was more valued by the rest of the organization because there was more visibility into the things they were doing and how those were impacting pipeline and revenue. I think that really had a big transformational impact on me. And I also spent a lot of time asking others how they were doing it and realized there wasn't anything in the market. So I feel like even though I kind of was passionate about doing a branch, this is solving something that was the bane of my existence for so many years. And I saw such a shift once. We solved it once. And that was just the beginnings. I think even now looking at how we were solving at that branch, we evolved so much from that. I think for me it was just such a personal. I want to help every team out there who is earlier in the buyer, who are earlier in the buyer journey, the SDR team, the teams that are bringing in pipeline, I want them to really know what's working and give them the purpose of knowing that they're working on things that actually make a difference. And I think the why now is also, I think technology has shifted so we can build better models.

What's really interesting, Mara, about what you're solving though, I know perhaps the CMO role or the head of marketing role is one of the most short lived roles in an organization. And you can benchmark this, right?

No, no, it's true. I know this. Every CMO conference I go to, the CMO is out of all the leadership roles is shorter tenured.

It's shorter tenured and like the CEO, especially an engineering driven CEO. Now I've been fortunate because I've done gtm, but I'm an engineer so I understand that pain. But most CEOs think like marketing does a bunch of stuff like the south park, right? Like you, you put some stuff, there's this underwear gnome action that happens and money comes out. They don't know what exactly marketing is doing and a lot of it is because of the offline activity that is driving human to human connection. Nobody wants to switch their cmo. Okay. Nobody wants to fire their VP marketing. They just want to understand what results they're driving. And so by as a function of driving longevity for the marketing leaders in a company, you're also helping those companies grow faster and not lose all this time. Because onboarding a new cmo, you take six months, seven months to get up to speed and restart the whole process. They bring their own tooling. So this is like a great service you guys are doing for companies in general, I think.

And it's more than even just that. Even if you do switch, let's say you're letting go of the CMO. I think there's a lot of complexity around why CMOs have a short tenure then. I actually think a big part of that is many other roles only have one function versus a marketing leader needs to manage events and brand and they need to be like analytics driven. It's almost impossible to find the right person who does it. All right? You usually find someone that's really good in one area, you fix that area and then you find the next person for the next area. So I think you can't actually say that there's still going to be a change in leadership. But I think what we can comment is if you have the right data when there is a change the new person knows how to they won't do exactly like your story, right? Someone came in, they bought the company, they didn't have the data, so they made the wrong decision. So if we actually can come in and always be the right data, even when someone leaves on their own accord or whatever, the organization has the data. So when they make a shift in leadership, that new leadership has all the information to continue and to make the right decisions moving forward. Instead of being like, oh, I'm looking at the, I don't have the data, so I just assume everything's not working, so I'm going to try everything again.

Alex, through your time building this, you probably saw two or three shifts in the startup landscape in Silicon Valley, right? You saw one shift which was growth at all costs, and then came the period where we want to see profitability growth. And you were probably building this right at that time to prove out the attribution for Branch, right? How do you think about this whole landscape here and the two big shifts? We've probably seen growth at all costs and then profitability.

I think that whether this was a blessing or a curse, we probably made those shifts just a little bit earlier at Branch than the rest of the market, which means we were building this efficiency focused attribution culture right when the rest of the market was still pouring money on the fire. So it meant that we had this model ready to go so that by the time we're leaving Branch, everyone else in the industry is just starting to pay attention to this. I've heard probably more times than I can remember at this point over the last year. Everyone's saying growth at all costs is the way of the past. Now we have to be efficient and efficiency is what you need data in order to do. If the factory that you're building, you have to be able to measure things, which is different from just spending whatever it takes. So I'd say that this was probably the quote unquote right way to run a business all the way along. But when you have unlimited cash, it's a lot easier to pretend that you don't need to pay attention to the fundamentals. But now that the swimming swinging back the other direction, everybody's looking for an attribution model or a measurement approach that will let them run more of an efficiency focused factory. And it happens that with some of these advances like mata mentioned, with AI and LLMs that can now process a lot of this data at a scale that wasn't really possible five years ago, it's at exactly the right moment to build something that really does answer this question. It's been a problem for as long as people have had marketing teams, but maybe now finally we're in a position that we can bring something new that hasn't been seen before.

And the corporate mandate, like especially in enterprises and I don't know which segment of the market you guys are targeting, but the more and more CFOs I've talked to, it is figure out how to reduce spend or reduce the increase the bottom line by 20%. Like 20, 30%. Right. And that doesn't always mean firing people. And even if you willy nilly fire people without data, then you're doing your business a disservice. Right. This is arming them with the right data to spend more wisely so they spend on things that actually move the needle. How are you guys thinking about it? How does the upside model work?

We think of it in a few different stages. So one of the challenges we've seen with a lot of the companies we've worked with so far is once you build out a go to market organization that has multiple teams, you've got sales, you've got marketing, you've got partnerships. The customer success team is in there working with customers. Once you've built that out over the period of several years, you probably have a quite fragmented tool stack. And when you have this fragmented tool stack, it means each team has their tool of choice which has one tiny little slice of the data. And nothing allows you to look at the whole story as the customer has experienced it. Which means any kind of analytics you try to run is just going to be impossible. So the first thing we've had to do in every case is find a way to knit that data back together again so that you actually have a data set that can be analyzed. But then the next part is it's not even enough just to collect the data in one place. There's often a lot of ways that data is like Swiss cheese, it's full of holes. So it's not just collecting the data, it's also finding some ways to heal the data. And one of the examples that I like to tell just because everyone can identify with it is salespeople know they should add contact roles to their opportunities. It never ever happens. So if you don't know who's involved with the deal, you can't really analyze the activity on that deal. So one of the first things we always have to do with every company is what we call contact role healing, which is running an Algorithm against the data to detect the people who were involved in that opportunity but were just never added to it. There's probably a half dozen other little healing steps that we take and we're finding more of them as we go. But you have to do that healing. But then once you've collected and healed, then you have this data set that you can actually do some really cool things with. And that's what we've seen a lot of our design partners starting to dig into.

Yeah. And I think it's interesting, the other important thing there, which I think most organization, including asset, branch mistakes that people make is they want to have a model and use it for everything. So even when we originally built this at branch, we build this multi touch influence model and we're like everything gets a weight and then we split the pipeline between all the different weights and everything gets like a weighted pipeline. That all adds up to a hundred percent. And that's a really good model. When you ask what are the things that are influencing me getting more deals, you can see, you look at the multi touch, everything gets like a part of the credit. That's not the right model to answer the question of what's helping me get more people into my database, what's helping me get more leads or what's helping me activate pipeline or deals or even what's the ROI of the different things? What's the cost of a deal? So I think as we step back, we realize that you need different models for different questions. So we have now an activation like model, a source model, an influence model and a cost model. And those are like different. And they are all a version of multi touch. But they look at things very differently. And you can't use just one for every question.

A hundred percent. And that one size doesn't fit all. And we've seen this before, right? The last touch, the first touch, the middle touch. That's all very interesting. Now your first customer set like that you're targeting.

Right.

Because it feels like larger companies need this all the more because one the wastage is higher. Right. My, my brother helps very large brands like the PNGs of the world with advertising and attribution is very hard because there's a lot more offline there. But at the same time they come with a mandate of you need to drive 20% cost savings across the organization. But the other side of it is probably startups are more easier to sell to because they're engineers. So where do you start in this market?

I think we're starting where we know, right? So like with an ICP that we know well, B2B companies with large deal sizes because the larger your deal size, the more complicated it is. But we're not going after very large companies because there's so much complexity there. We're a young company. I think our earlier ICP is somewhere between 50 million to 3,400 million. They're not, they are still, they're mid, mid sized companies, not quite startups. I think from a startup perspective for our model for us to come in and bring value, you need to have tried things, right? We help you get attribution for the things that you've tried and we help you run tests and understand what's working. But you have had to have tried things and we see that when you are like at 2 or $3 million in revenue, you're just trying a lot. And also I think we were just that we run these revenue efficiency events and there was a debate on our panel where someone said if you're like before series B, as a founder, it's way more important for you to drive growth. You're still trying to find product market fit. You're trying to get people. So growth is still more important than efficiency. And efficiency becomes more important when you've hit a certain amount of like growth and product market fit. Now you have to make sure you're also efficient.

What's really interesting is years ago, before Boast, before the previous company I did a startup called Automatically, this is 201314 which was leveraging companies existing data to respond like real humans. The biggest mistake we did was we deployed on Zendesk because at the time Salesforce and Oracle were like, you need to do this long security review. The issue was those star Zendesk at the time had the smallest customers and the data wasn't available. They're just trying things. And then you were spitting out gibberish and it was just a, it was a, it was a good idea that worked well on a Twitter turing test to respond automatically to a real human. But on, on live customers that were small, it didn't work. But coming to that point, you need to try a lot of things, right? And smaller companies are trying and eventually try. They try, they accumulate a lot of data and then they get to a point of needing you. But when they start to need you, they need healing, right? They need to be healed. If you go to a company like Bose, which my previous company does over 20 million ARR, there's a lot of healing that needs to be happening, right? In a process like that, how do you do the healing? Let's say, like it's a sales process and you just do last touch attribution and you have the lead source. How do you then after try to heal that and reconcile all that? It's very interesting.

We don't start from the lead source that the company is already touching. In most cases, we don't look at whatever existing attribution model they've pieced together over the last few years. We don't even really incorporate that signal. Instead, we go to the raw data. So the event campaigns, the emails that have gone in and out, the meetings that have been scheduled, and we pull in these touch points before they've been processed through any prior attribution model. And we run our model on it and then we can compare it against the results of a model that they might already be running to say, okay, you had this deal in three, let's see what our model says versus what your model says. And in almost every case, we're able to show a much more nuanced picture of what led to this deal closing than whatever was in their existing system. But we're not really including that because it's lossy by definition. They've lost the granularity of what led them to make that decision. And because they weren't running a weighted multi touch, they probably didn't make quite the right determination. It might have been good enough for what they were using it for. But we can come up with something that's better.

I think one thing that's also interesting here that kind of plays into the why now? LLMs can allow us to do things that were never possible before. So what we found with some of our design partners looking through their data is there's a lot of data that's missing, but it's hidden somewhere in an email. Like in some cases, the CEO reached out to someone and then an email was forwarded and then eventually gets to an AE and the email from the AE gets tracked and if you look bit down enough, you can find the original email and that's somewhere hidden in a forward of a forward. So with LLMs you can actually take that out and put it in its own touch point. So we're doing, we have this concept of phantom or inferred touch point that are touch points that we see because of the data, and that's already in there, but it's not actually pulled out as a touch point. So that's something where in some way we are healing, but we're doing a different type of healing. We're healing their actual view of what actually influenced the deal.

This is really good. I'm getting excited because I Mara, seriously, I faced so many pains as a bootstrap company and then when we were majority acquired and had to bring a bigger leadership team or went from pirates to Navy, it was so hard to explain everything. It just felt like I'm being talked to, like I'm a dumb person who never understood any of this and somehow we did magic and willed it into 10 + million in ARR. So this is very calming. What are some problems you guys see in the traditional attribution models and how is upside different? We talked about some of this, right? Like the fragmentation of revenue related data, that is marketing, that is sales. It's sitting in different tools and systems and email and spreadsheets that you're starting to fix with LLMs, creating that opportunity. But what other areas do you see opportunity in?

One of my favorite quotes and also I could see this on the homepage of our website Therapy for your Go to Market organization. I'm going to give you credit for that one day. But one of my favorite quotes is all models are wrong and some of them are useful anyway. I think it always depends what kind of question you're trying to answer. Because if we rewind to our life at Branch, where we're trying to tell you what drives app downloads, you don't need a very complicated model to understand that like an app download is an impulse action. The time to conversion is maybe 30 seconds. It doesn't require a very complex model to understand the cause and effect there. But when you zoom out and try to attribute a sales deal that takes 18 months, that's where you need something of lot more complex. So the example you gave earlier of startups, they typically don't need anything that complex. So maybe last touch is good enough. But you can put all of these models on like you can find so many reasons to do like a two by two matrix, but in this case it would be scope versus sophistication. And some of these models are very low sophistication. But that means they're easy to understand. People know what happened even if they agree that it's wrong. The challenge is where you try to come up with something that is really sophisticated but still intelligible, still something people believe. Because then you start running into this challenge of it's a black box. And people inherently don't trust black boxes usually. So a lot of these systems that are out there, they're either very low on the scope of what they're trying to do. Maybe they're just measuring marketing advertising or they are only measuring the SDR team, or they are getting so complex that it's this magic black box that spits out things that nobody really understands and there isn't much in the middle. Like some of these multi touch models that have existed for years, they still fall low on that sophistication scope because you'll be taking every touch point. You're just saying, well, the email and the VIP dinner, they are worth the same amount. At least people can agree on that. They know it's wrong, but it's understandable. We don't think that's the right answer. But we have to avoid falling into this trap of making something that's so complicated nobody really understands it anymore. So we're not just trying to bring the data into one place, we're trying to understand what is it that will help our customers make a decision based on the data. Because then we can figure out the right model to show them the answer to the question that they're asking. Not just vanity metrics that might look good for claiming credit but don't really inform the decisions they need to make for the business.

No, makes makes complete sense. And it's a lot of the times what I find and in general when you have non data driven conversations, especially with a bunch of emotional like everything, every business is part emotion and part data. But for whatever it's worth, marketing and sales involve a lot of emotion, right? Because a buyer buys for an emotional benefit. Typically the functional benefit is there, but the emotional benefit brings them over the line. And when you have this sort of dynamic in sales and marketing, you start to have a very convoluted conversation around correlation versus really causation. What's causing this? Right. It's like, no, I went to golf with this guy and that's why he closed. Are you serious? He closed a million dollar deal because you went to golf with him. Now that's not quite it.

Right.

So how do you solve this correlation causation? I'm just anything you guys are saying brings back 15 years of my career, right? Like in my mind, all the pains across companies, Enterprise, B2B, SMB, all of that. So I'm just giving you like true examples that I've had to justify in boardrooms. Right?

And I'm going to let Alex answer the causation versus correlation because we actually. But I just put on emotion. I think actually one of the biggest problems that we're Facing in organizations is there are people who feel very insecure. So there's a fear, there's a lot of fear that if we come in with our system, they will not get credit and their work because they didn't know. There's fear that their work won't show up as valuable. One of our biggest challenges is overcoming that fear. But I think the causation versus correlation is you got to the crux of the problem.

And it's a question that I think will probably continue to come up until the end of time. Because even when you ask, there's this move in some places to go towards what they call self reported attribution, which is like a survey. You ask somebody what was it that caused you to buy? And in some cases the person probably doesn't even remember.

We had this. Dan was like, you went to this really fancy CMO conference. Would you like buy this tool because of it? I'm like, no. But you, like, you made us have a meeting with them. I'm like, yeah, I mean you have to have a meeting with them. So it's really funny that I wasn't aware, but it probably did really influence me.

And there are all these brand awareness campaign effects and then a statistic that comes up all the time is of 100 companies who could possibly buy your software, only five of them are actually going to be willing to consider it at any given time. So there are a lot of influences here that aren't just as straightforward as causation versus correlation. But that doesn't mean you should stop trying to look for incrementality because you don't want to just continue doing something because the attribution model says that it got a lot of credit. But if you hadn't done that thing, the deal would have closed anyway. So part of what we're doing, it's challenging to run a fully predictive model in most B2B data sets because there is not a high volume of data if you're only dealing with a few dozen deals a month. That's very different from some of these mass market, like you mentioned, the kinds of models you might run at Macy's or Procter and Gamble, these really large B2C companies where you have a huge data set where you can actually run a a proper regression. That's much more challenging in a smaller B2B environment. We're hoping eventually we can improve that by doing industry wide regression models. But in the meantime, what we're investigating is can we build a kind of hybrid model where we are still using proper statistically rigorous regression methods to inform how things are weighted and what deserves how much credit. But then we can use some of these multi touch attribution methods to drill in a level further than you might be able to get with a traditional media mix model so that you get the best of both worlds. It's still statistically backed, but also it's actionable. It's something that you can make decisions on rather than this much more fuzzy top level data that I think has scared people away from using something like a media mix model in the past.

What's really interesting as I hear you guys talk is right now because I've tried everything and in the past, right? And the spend on hiring marketing ops, web ops, GTM ops is very high, right? This thing can run into lots of money like for even for a say series B startup, right? Like 20, $30,000 when you factor in the tooling and the headcount cost. I wonder if you're really selling a tool or you're selling to a headcount. Because I think as we look into the future with the prevalence of AI and LLMs and agentic, customers want an outcome. Customers want Super Mario. Most people think the mushroom is what they're selling, but it's Super Mario, right? And customers want an outcome. Then are you, is it better for them to use a tool slapped together with a bunch of people who try to figure it out or is it just better to give them the outcome and sell service as a software? Because that is the interesting part here is upside selling to a tool or is it like selling to a headcount rather an outcome?

I think the biggest challenge we've encountered so far is that people want consulting and insights and advice. They just want, they don't want just data like we, when we work with our design partners and we like give them presentations of what we found, they're like more insights and don't just want to see the data. Tell me what to do with this. So I think there's. We will have a hybrid model from day one where you're going to get a certain amount of consulting hours and we're going to go and set up the model for you and show you dashboards that answer your questions. But we're also going to give you someone in addition and then you can probably purchase more hours because yeah, it's a hundred percent. People don't just want the data.

But Mara, this is super interesting, right? So you've like now taken this fragmented data that is sitting across I'M geeking out on this because I've tried everything and you gotta trust me. You stitch everything together in one place. You build a localized LLM for them based on their data set and they can query it before even coming to your consultant and say, what happened with this at this time? I kid you not. I was on the board meeting of a public company this morning. I don't know why, but like I sit on a board of a public company and the board chair takes the financials, especially like this sort of whole GTM thing. He uploads it to GPT and he asks why is my revenue going down? Because it came through so last minute, right? And I think it's funny, these things are starting to happen is if you send the board pack or the GTM or the growth last minute, they'll just upload it and ask questions. So I think there's a real opportunity here, twofold, right? It almost like as a function of making what you're doing real time, the attribution real time, you can drive collaborative conversation around that data that now happens after the fact. And that collaborative conversation is happening between multiple different stakeholders like the board, maybe the cmo, maybe the CRO, maybe the CEO. And as a function of that, now you go from having marketing data as a waterfall method to being more agile, collaborative, like a figma, like jira, what Jira did for development. It's a real network effects business under the hood if you think about it long term. Right? Because that's what's needed. Maybe that's not the problem you're solving, but as a function of making it real time and bringing it to the front, that's the opportunity you're driving. I don't know, I'm just geeking out and I've gone like off podcast track now into geeking out mode.

We thought about different ways to bring. I call this the chat with your data paradigm. Because I've seen it come up in so many tools recently and it's limited by the AI is infinite interns problem that kind of exists out there right now. AI can do a lot of really impressive things, but it's still, at least in my experience, at the level of what you would trust a good intern with. They can do things, they can do things at scale, but you don't want them to do sophisticated things without supervision. But it's already in the last year gone from like undergrad intern to MBA intern. So it's going to keep going up and pretty soon you'll get genuinely impressive. We've already run some tests on how well AI, if it's really coached and managed all the way through, can understand the trajectory of a deal. And it does a pretty good job of describing what was influential at each stage. So imagine if something that used to take a group of leaders sitting in a room for half an hour. If you can do that for every deal with AI overnight, constantly, that's where I think we'll get a lot of scale out of this. And that's what, like you said, will allow somebody to just ask questions of their data set in real time as long as they trust that the results they're getting are not hallucinations. But we've also got this kind of split what we would say into exploratory reports versus recurring dashboards. There are some questions that a team just wants to monitor on a day to day basis. That's the kind of thing that alerts via email or maybe even pushed into Slack or a dashboard that you sign into. That's a good place to go and monitor ongoing business fundamentals. But then there's also this other problem that you said about the headcount problem. It's the we have something we have a question about, but we don't know it yet. We want somebody to go figure it out. Right now, AI isn't there yet. You still need a trained data analyst to go do this for you. And that's where the services component comes in. But as AI continues to improve, I think that may be a place where we may not need as much human attention. It could be that AI can do that question exploration process and then maybe somebody sanity checks it to make sure it hasn't lied to you in the meantime. But even that eventually I think AI can make a lot of this just accessible in a way that it wasn't before.

I think we're a few months away from upgrading from MBA intern to actually a highly paid sort of rev ops, GTM ops person. The pace of this, the way it's moving, so it's super, super exciting. What's next on the horizon? You learned a ton. Things are moving at a rapid pace. Mara said there isn't just one model and a lot of excitement here. Where do you see, what do you see happening in the next six months to a year in your space and in the business?

I mean, we're working right now with quite a few design partners and more of a consultative basis. And we're just actually we learned a lot. We had our first version of the base of the data model and we Learned a lot and now we make a lot of those changes and bringing in things like using LLMs in our factoring of what how much a touch point should get depending on the quality, bringing in shadow touch points and building all of that. And then we want to bring this in, bring this to market in the more scalable way. But if you are in our ACP and you're struggling with the questions of what's driving my revenue, I have realized deals and I don't know what I should do more. We still have a couple of spots for more design partners, so just shoot us an out.

Awesome. I think Mara is one of the best, most thoughtful leaders I've ever met. And Mara, what I've loved about you and learned from you personally as you, as I watched you build a branch and I think the first time you came to our conference it was in a dingy nightclub that we had together all the way through. All the way through, all the way through. Scaled up conference is you truly embody this right? Falling in love with the customer and help them become successful beyond the product or service you truly like. You care to create an impact beyond just selling a line item. And I think people that work with you, you'll impact them in many ways beyond just making that one part of their business better. So I highly encourage anyone here who's listening. Talk to Mara. One, because she's built a successful company, right. Two, she's built a massive community. How big was your last community that you built at branch, the mobile dev.

Community, how you counted? It's probably just the people who came in person to our events is probably like something like 20, 30,000. If you think of the global like online community, it was a few hundred, few hundred thousand.

So there's beyond that. I'm trying to sell you a tool perspective but I build a successful multi billion dollar company and beyond that a community perspective that could be helpful to your business. So truly engage with Mara and Alex. They know their stuff and beyond that they care.

Thank you.

Thank you so much Mara for joining. We're going to record this or publish this probably the week after next or next week, as soon as possible. I was super excited. I was like geeking out with you. So this is great.

This is great. Thank you so much for having us.

Awesome.

This episode is brought to you by Boast AI. Each year the US and Canadian governments give out billions of dollars in R and D tax credits and innovation incentives to fund businesses like yours. But the application process is cumbersome, prone to frustrating audits and receiving the money can take up to 16 months. Boast AI gets you access to research and development, tax credits and innovation funding opportunities without the headache and red tape. Join thousands of North American Companies leveraging boast AI software to maximize cashback. Check out www.boast.AI.

Thank you for listening and we hope.

You enjoyed this week's episode of the Traction Podcast. If you enjoyed the show, please leave us a five star review and you.

Can find more information and all the.

Resources mentioned in today's episode at. Boast AI. That's b o a s t a I forward slash blog.