Recursive Marketing

Julia Nimchinski:

And joining us next, Gil Allouche, founder and CEO of Metadata and Tag Bozied. Senior Manager of Paid Social at Docebo. Welcome. How have you been? Gil, you’re you’re in it.

Gil Allouche:

Of course I am. Thanks, Julia, for, for inviting us. We’re excited to be here.

Julia Nimchinski:

Likewise well, Recursive marketing it is.

Gil Allouche:

Yes, that’s something that we actually practice. I know it’s kind of in the early stages, but we practice that, and I’m excited to show your audience what’s possible.

Julia Nimchinski:

Amazing. Very excited for this. And yeah, let’s get into it.

Gil Allouche:

Awesome. Tagg, would you mind, sharing your screen? Amazing. So, first, let me just introduce My friend and customer, Tag Bozid, he’s, I would say one of the most forward-looking marketers I know. He tries technology without fear, and he puts the time, like, his prompts sometimes are very long and very thoughtful, and then he gives a lot of feedback back, which is very appreciative. I’m very appreciative of that.

But Tagg is doing some pretty cool, you can stay on the same, on the first slide, Tagg. But essentially, you’re gonna see some of the cool results, like, actual business results that are coming out from using artificial intelligence. And I think that’s the most important part for practitioners outside of the technology. And myself, I’m the founder and CEO of Metadata.

We’ve been running a set of advertising layer for B2B brands, so companies that spend, you know, between a million and a half to $50, $60 million a year in ad spend. We automate the entire process, and we use AI heavily for it. We built an MCP server with about 180 tools and 7 different agents that can do anything from landing pages to audiences to creatives, and then running and optimizing campaigns.

I also run our own Agentic go-to-market operations, so we can go to the next slide. Maybe I’ll quickly start with, what is a recursion? A recursion is really a definition from software engineering. And the idea is to have a process, a rule, or a function that repeats itself. Or it refers to its own previous steps. So it calls itself with the context of the last round.

Programmers love that because, one, it simplifies it, so instead of creating this, like, long code base with a lot of loops and workflows, it’s really a basic empiric function. that calls itself, and every time it calls itself, it’s calling itself with additional context, so it’s smarter and smarter, and it’s… eventually it reaches some condition where it knows, okay, I’m done here, I performed the task.

And it’s moving on to the next one. So that’s just so you know, like, the general understanding of what is recursion. And in recursive marketing, the idea is that every campaign that you run. it informs the next campaign.

So, for example, you’re running an outbound campaign, and you’re sending a thousand emails, and maybe you A-B tested or multivariate tested a few different subject lines, or emojis, or different senders, so on and so forth. the system will recognize that there is some kind of a lift to a particular variation, and it will call itself again, but with this new context, and it will assess the context and just do a better job outbounding.

Same goes for ads, gifting, or anything else. So now for my next and final slide before Tagg will go into, his actual use case. This is what we’ve built here internally at Metadata. This is something that I practice.

I have the ChatGPT Codex window open right now with exactly these. these motions, and essentially, it’s a… it’s a… it’s an agent that has a bunch of skills, and it goes and sends outbounds every day, gifts. social outreach, direct mail, runs ads, and then looks at the data around it every day, and then informs the next day of what should or should not happen, and it does that across functions, so across different channels.

So, for example, if it learns something about Like, it pulls a gong recording, and it learns something about the aha moment in that gong recording. First of all, it will apply some of that messaging and try to use that in the next outbound, in the next ad, but it will also. update the weight for that particular email to show that, okay, that email led to a good outcome.

So, that’s essentially what I run here at Metadata. I’m using, of course, Metadata MCP for all of the ads. I’m using about 7 other APIs, like Instantly, S64, and Sendoso, so on and so forth, for the… for the gifting, and I’m also running a workshop. If you want to do this across different channels for your own company, then I’m teaching that at no cost.

Tagg, without further ado, I’ll pass it to you, and Maybe you can show how we use AI internally at the table.

Tagg Bozied:

Yeah, thanks Gil. Absolute pleasure to be here today, and talk about the exciting things that I get to work on. I will say that, you know, in the last, what, 6 to 9 months. integrating pretty much the entirety of my workflow into AI.

I really think that us as marketers, the use cases, the applications to the use cases that are specifically for marketing has to be one of the most transcendent and exciting ones. in any industry worldwide, for sure. I work for a company called Docebo. I’ve been at Docebo about, 5 years. Docebo is a learning software company, probably the most capable one in the world, from internal, external, and hybrid use cases.

So… I came to Docebo to build an enterprise demand engine. We built, the core of our business in the first 20, 25 years of our existence on SMBs and mid-markets. About 6 years ago, the company went public and had a charter to start selling to enterprises. That’s what made the opportunity with the company particularly appealing. I will say.

From a work culture standpoint, we are about as AI-forward as it gets, supported fully by leadership to transcend how we’re working as professionals. And modernize what we do to that end of creating more enterprise pipeline and, deals. So, when you do that, and in our… in my time, I believe the average contract value, at Docebo has gone from about $15K when I came 5 years ago, and we’re up to $75K now.

So when you transcend that sort of business operation. And you’re using media to do that. I think that The biggest question that we had is. What exactly, how long do we need to advertise them in order to make an influence? Through our partnership with Metadata, I’ll say that when I came on, I was asked for my wish list, and Metadata was the one thing on my wish list.

I had some prior experiences that were disappointing with Some of their competitors shall remain nameless, demand blah blah blah blah, and 6… And I knew that I wanted to use metadata and the best way to describe the contrast of those experiences is, you know, technology is robust and often impressive and it definitely is out there in this. this marketplace we’re talking about, when you use metadata, it’s clearly made not by a technologist, but by a marketer.

So the intuition of working along the program, working inside of the system is very intuitive, it’s very smart, and with the evolution of the AI tools that they’ve provided in the use case that I’m about to show you. It is very smart, and it thinks like us as marketers, and it makes this recursive idea of using one campaign to inform the next.

It really transcends performance, and most definitely the work it takes in order to draft campaigns at scale. So, what we’re able to do with the integration of metadata into all of our systems is we’ve taken 4 years of ad data. And now I’m able to use and have insight and have clarity into the entirety of the prospect journey.

And I can take, no joke, what would probably be weeks to easily months of work between the strategy and the campaign builds, and using the, the Agentic approach inside of the metadata platform. what you’re about to see takes me about 3 hours. We’ll… there’ll be some hop-skipping and jumping just for the sake of this presentation, but.

The… Intelligence of the platform, it understanding what we as marketers need to do, and where the human should be in the loop, is absolutely profound. So, while you soak in the data that’s in the right side here, I’ll read you the prompt so you guys don’t have to do that, just to give you some context in terms of how to how to cue the thinking and the thought process inside of this Agentic approach.

So for every account… so what I did is I took a CRM export of all of our pipeline from the previous quarter, and simply, for every account with an opportunity in this export, show me the buyer journey in metadata before the app was created.

Pull two time windows from the account journey report in pre-pipeline from April 2022, when we started, our ad program to when the op was created, and also the final 90 days before the op. For each window, return LinkedIn impressions, clicks, website visits, unique visitors, channels, and high-intent pages visited by the prospect. Don’t estimate if an account has no data, market and move on.

Return one table, and this is the important part, plus a deck I can use as a campaign brief. So… With that prompt, I’m… and… and pairing that with our CRM export, I’m able to pull this kind of data.

And this is not only just interesting and great as marketers, but this I also get to share this with our revenue marketing team and our regional marketing managers that are spread out throughout the globe can start their campaign campaign planning process and optimize their budgets to because they can start to understand influence. That’s using four years of journey data. To, give us an idea of what’s happening.

So what you’re seeing here is an example of the deck that gets built. So you not only have the dashboard, but I built a deck that ultimately is used as a campaign brief. that feeds intelligence, okay? So, treat this as the campaign brief for a new campaign, summarize the rules it implies, incite the start… the stat behind each. Don’t build anything yet.

Essentially, what this is wrapping itself into, and before I move on to the next slide, probably the. Most surprising… most surprising stat, or most surprising aha moment that’s with us, is that if you were even to ask around our company, and I’ve done it a lot, how long do you think that we show a prospect? How long do we advertise to them before they become an opportunity?

And we’re an experienced B2B learning SaaS company, right? And most people say between 12 and 18 months, and what we’ve been able to prove with this is that it’s 3.7 years.

3.7 years, that’s how long they were on, and if, if for the listeners that are out there, if you’ve done any, research or done any reading of the LinkedIn B2B Institute, you know, the, 95-5 rule, how only 5% of your target audience is actually in-market buyers when you’re advertising to them. What we have found with 4 years of ad data is that is certainly, a truth. And we need to strategically adjust accordingly.

Essentially, to simplify everything that comes through in that campaign brief. is what you see on this slide. This was, as somebody who’s been in the day-to-day for the last four and a half, five years with our company, looking at all of our, all of our brand awareness spend.

Well, understand that that first LinkedIn impression, the length of time that you’re seeing there, our strategy was kind of always on to everybody. Once I received our new VP, I well, not new anymore, but a great follow on LinkedIn is our VP of Revenue Marketing, Sylvia Valencia, who comes from Refine Labs prior to Docebo.

And a big strategic shift that she chartered to me was, we want to focus on maximizing our reach and understanding our frequency. to our audience. And… with a buying committee like ours, at enterprise-sized companies, that’s talking about, like, 15 to 20 people.

How this recursive idea, and how this Agentic GTM approach, and being able to yield this data, what we found is, is being on to our practitioners and to our core audience is certainly important for that length of time, most definitely because we’ve seen positive results.

But where this data gets interesting, and it feeds campaign strategy and this recursive idea, is the green at the bottom is when the buying committee gets involved. That’s when the rest of everyone who’s gonna be an influencer or a stakeholder in the deals that we’re trying to… trying to close, that’s when they’re meeting.

That’s when we want to… that’s when we want to use our media at that inflection point to maximize our influence in the memory structures that we create. So, to simplify it, how we adjust it and what this data is feeding us is start broad and cheaply.

Use… we use CTV ads and a combination of LinkedIn and Reddit at a brand awareness scope to drive so that our CPMs stay low, our reach and penetration stays high, and then we wait and watch for that signal. And using this, using Gil’s agents, that metadata, we’re able to. appropriate and align those signals and create audiences without too heavy of a lift, quite honestly.

And then once we know that signal, then we want to surround, loudly, for lack of a better term. We want to maximize our frequency. To that buying committee. So that while they’re meeting and while they’re talking, we have the extended influence and the extended knowledge that we’re delivering through ad impressions to the practitioners.

And then when the buying committee is coming in and hyper-focused on this, on the specifics of the use case of the deal that we’re creating, that’s when we want our impressions. That’s when we want to take our money and our investment to maximize, not that whole time. So there’s not only a performance improvement that comes from this, but a cost savings as well.

So… This is an example of the workflow start to finish. from… from… from the Agentic approach. Notice the gate checks. The gate checks are the human in the loop where everything that you see in purple goes out and works. Very intelligently.

Particularly, the more integrated that your system is, and the better you inform the platform of what your strategy and outcome goals are, clarity’s important. of always do, always don’t. Setting up clear guardrails for that. The areas in purple here work in… unbelievably well. And then where you understand where it’s really smart.

Is when the human in the loop comes in and it asks you questions, literally gives you answer boxes to confirm what it’s analyzing, what it’s recommending from a build or an optimization standpoint, and you confirming in plain language whether or not that’s what you want to do. 12 prompts is a lot.

If you look at the length, I cut them down into simple quotes inside of one sentences, because they are a lot, but I will say… I will say, if you have clean prompts, and you have a clean, clear idea with clarity and conviction of what you’re creating. The system will work unbelievably intelligently through these purple areas, and ask you great questions at each one of these gates. Right here, you see an example.

This is a dashboard. This is a visualized dashboard of the information that I’m getting, so it goes out, it understands what just happened, it understands what I want to do next, and then now it builds an example of, this is what we’re gonna go with.

So between putting the budget in front of you, understanding how your trigger segments are gonna work, and actually seeing what’s gonna go out and be shown in the marketplace. All gets put in a single… a single view in front of you, so you have a clean understanding.

This right here, in order for me to take All of the… all of the intel that came. from the prior campaign, devise a campaign strategy, and then go out and build it. This is about in the two and a half to three hour mark, I now get to see, in one consolidated view, a confirmation of, this is what I want to be seen, this is who I want to see it, and this is how much I want to spend on it. Definitely next level.

Definitely next level to get to this point. And the creative tools that go into this, I will say. the… what you’re seeing in terms of the creative and those real DeChebo ads. It has been a process. I highly recommend you work with your creative team to have very strict rules, a very clean understanding of exactly what you want it to look like. That was a process for me.

I will say that I’ve learned a lot just from being able to get to this point. But the cleaner your brand system is, the more informed you make. the more informed that you put in your infrastructure of what your brand is, who you are, what you want to do, and what… over what time and what you’re willing to spend to do it. This is not that far away.

It was shocking to me I did have to put some time in to get it down to 3 hours, but to be able to get here has absolutely changed the way that I work. So, in summary, what I’ve learned from this, deals come from presence, not clicks. It was an aha moment for me and pretty much everybody in our organization about how long that we’re on, and also. the… only 20% of our pipeline ever engage with our ads.

That’s very interesting, yet they’re still coming, yet they all go through the same process of coming to our website. There’s a… there’s a site surge by a multitude of members from each of our… each of our prospects. that become in our… that become in our pipeline. They all show that similar behavior, regardless of how they… of how they respond to our ads.

So, the preemptive impressionability, the frequency of which you are on to stay front of mind, is certainly important. So we went from optimizing for clicks, being on to all personas all of the time.

And having the same exposure on every account, to optimizing for account warmth. timely impressions to the buying committee when they’re actually in market, and they have the mind space to, to absorb what we want them to know, and find the interest cheaply, then surround it. I’m gonna circle back, just real quick before I hand it back to you, Gil, to understand right down here, compare segments by outcome.

This is what… and I’m in round one, right? I’m… this is all brand new, what I’m showing you. This, just this workflow. just got deployed recently in the last few weeks. How Recursive Marketing and how this is going to inform us is what you’re seeing in that bottom right, the comparing segments by outcome.

Us being able to understand that the opportunities that we win, we show a median impressions of 861 versus the ones we lose at 251. And the median impression is the final 90 days. You know, we have three and a half times the amount of impressions in those final 90 days. Using this system.

And using this intelligence to inform the delivery of our campaigns is going to be able to us use what’s happening on… and what’s going on in that loss column, and move those up. To move more wins… to move more wins into that win column. So with that, Gil, it’s been a pleasure being your partner. Absolute dream state for me as a professional and the technology that you’ve built. Congratulations.

I tell everybody when I talk about you, you guys are the best ad tech company in the world. I can’t believe that we’re here. And in this age of AI, somehow you seem to have even transcended what is thought possible right now. So I’ll I’ll I’ll leave it to you there.

Gil Allouche:

Tagg, thank you so much, also for the kind words, for the partnership, as well as for being at the cutting edge. You know, I’ve been running Metadata for about 10 years. What I was dreaming of building when I was a marketer with an engineering degree, what I was dreaming about building was impossible. But as the years went by, and especially in the last 3 years.

Truly, like, you are able to do 100x marketing, and if you have, you know, a smart individual like yourself. Then you can kind of run the show on your own without too many, moving, moving pieces, so… I wanted to make sure that we have a few minutes for questions, or any comments that Julia or anyone else wants to make.

Julia Nimchinski:

This is amazing. Thank you so much, Gil and Tagg. My first question is, is a context question. So, obviously, it’s the dream state to get to your point, Tagg, by leveraging metadata, but how much context is enough? in order to see that type of productivity.

Tagg Bozied:

Yeah, good question. You know what I haven’t done and taken the time to do is, what if I only gave it one year’s of ad data, right? We’ve been a partner for four, four and a half years, so that context that I’m able to give is very, very rich. What I will say is that. there’s nothing that we’ve seen on a quarter-to-quarter basis. I bet by 3 quarters.

By 3 quarters, an understanding of running, okay, what have… what is the look-back window for us to understand? The Spotlight tool inside of Metadata’s platform gives you kind of a three-dimensional understanding of that user journey. You understand all of your marketing touches.

You understand all of your sales outreach, sales outreach activities, and you understand all of the buyers’ interactions with those emails, interactions with those ad impressions, how they come to the website. I think what’s more important is that when you stitch together those three dimensions inside of that system, that the statistical significance comes pretty quick. So. what I just showed you was from Q3.

I… the quarter before, I did it for Q2, and the results were almost exactly the same. I mean, that… the first one that I did was 312 accounts, I think this one said it was 516. Right? They said that the they said the exact same thing. So I think it’s just a matter of integrating the system fully. and letting economies of scale take over create the statistical significance.

And while I have four and a half years of ad data to be able to understand a full look-back window, that’s just informing us about how soon to be on, not necessarily exactly what we want to do, because it’s really about influencing for us in the B2B space, right? It’s about influencing the right people at the right time, and it doesn’t have to be that far in advance of a look-back window.

You don’t have to have all of that data in order to inform strategy, in order to do that.

Julia Nimchinski:

Amazing. Dagg, just to objectify it, it’s a really, rare opportunity and amazing transparency from Gil to bring, you know, Metadata’s customer on. You mentioned you tested other vendors in the, you know, ed tech space. Just to not name names, but what were the main problems compared to metadata?

Tagg Bozied:

The campaign management was clunky. from… my first 3 days in using the… going back 4 years ago, before AI was, like, a thing in our workflows, multivariant testing and scaling campaign administration. I’ve been running our global campaigns in its entirety for our demand engine, single-handedly. For four and a half years.

The… in the… the classic… in the classic view of, of metadata software, the first thing they show you is how many hours that you saved. And so I think that what sticks out first for a practitioner is how much time you’re saving, how efficiently you’re able to work from a campaign build and administrative standpoint.

And what they… and what they have done, most certainly, I have to mention that, you know, everything that was from a campaign build perspective, what I was also able to do about 12 to 15 months ago was take all of our campaigns and assign Slack channels that deliver high-intent visits.

It alerts each one of our account owners when our AE… an owned account by our AEs and our BDRs. when a high-intent page view happens from one of their accounts, I’m able to use the metadata platform to inform Our sales team in real time, when high intent activities are happening.

So, you know, being able to save all the time in campaign administration, focus on what matters for sales, not just… not just administering, not just handling the logistics, but actually being able to be responsive and deliver intelligence to the sellers. Is certainly something that sti- that sticks out.

Julia Nimchinski:

Amazing. And last question, from the audience. People are complimenting your insights, and they’re interested to learn more if you’re testing creative. In new channels like CTV, look back window is key to really measure uplift on areas like deal value and pipeline acceleration grade data set.

Tagg Bozied:

Yeah, you can reach me on LinkedIn. I’ll also say one thing that we didn’t get a chance to show. I had this idea, Gil, I have to tell a story about at Customer Advisory Board, when, they had a customer panel, and they said, what do you want to see next? They went around and asked us what we wanted to see next. And during lunch, they had a hackathon with their staff, and they went and built their top 3 ideas.

How that’s manifested is an account timeline, where you can literally understand every part of the journey of any of the accounts that you’re targeting, and you can start to see what those inflection points are, what that inflection point is from a years-long look-back window. So… those insights from a channel perspective, are available in the Metadata.io platform.

I do write, not frequently, but I do write, I do have some stuff on LinkedIn that shows, from a channel perspective, the influence And the impact that campaigns have had and everything that that I write on LinkedIn pretty much comes from comes straight from the intelligence that I get from metadata.

Julia Nimchinski:

Amazing. We’ll share your profile. And, what’s the best next step, Gil?

Gil Allouche:

I would say if you are a practicing marketer and you want to embed AI, you should reach out to your peers, so folks like Tagg who are carrying the torch, I think are good people to just understand what’s possible and how to do it. And if you’re looking to get hands-on and you want to join the workshop, then you’re more than welcome to. It’s an Agentic, it’s in a workshop.metadata.com.

We do it for free, just to make sure that marketers can become more modern with artificial intelligence.

Julia Nimchinski:

Thank you so much. Huge pleasure.

Tagg Bozied:

usually.

Gil Allouche:

Tag.