learn 98% of signal based outbound in under 10 minutes

I think broadly cold outbound is kind of dying. you get some gurus saying yeah, you can do it still with volume, and that is true.

but after testing many different outbound tools, the cold email tools, the data tools, all the GTM tools, and speaking to over 100 companies about their GTM, I think it's very clear that if you blast 100K people 100% cold and someone else does with intent data, the second guy is winning

so in this post I want to go over the GTM engine, GTM algebra, GTM alpha, types of signals, trust, a live example with Claude, interfacing with the data, and the conclusion.

(for anyone that's new: this newsletter is me building a GTM engine in public and trying to push GTM engineering as far as it can go.)

(and if you’re waiting on the reddit pain point free service, bare with me one more day, had a lot of sign ups and just trying to get through them all next couple of days to check it’s all good value for you)

and if you prefer the video version it's here:

first let’s look at what’s going to be taking the signals and put them into campaigns. This diagram looks a little bit confusing, but it's basically just how I think about the GTM engine as a whole. so on the left side you basically have signals and monitors, and this is where you're feeding in all these signals and data from various different platforms, like what people are talking about on GitHub, what people are saying on LinkedIn, all these different things.

and then you're basically bringing this into your central engine, which is basically scoring things, understanding it and really routing it to the next opportunity. this is why on the right side you have channels, where you can basically route them to different outcomes.

a very clear distinction I want to make is that the stuff people are doing with Claude is mostly list building outbound, so they create a list of people that fit an ICP and then they put them into their software and that’s great, I think it makes sense to start with Claude. you can build a list, you can build a good offer.

but ideally I think where things are headed is more towards this closed-loop outbound where it’s on 24/7 optimising the plays/

cool so now, we have an idea of what we’re doing, I want to go over the some ideas on how I think about optimising them under this “GTM algebra” concept (which dw we will go over more later)

the signals that we're going to be picking up today are through Apify. I feel like Apify is a good starting place. there are better tools out there, but it's got a very broad range of signals and data you can choose from, and the MCP is pretty good.

so we're basically going to be using Claude as an interface to build lead lists that go into these two different channels. so you'll be like hey, I need people that have been hiring, it speaks to Apify, it's going to have an actor, and it's going to build that lead list.

so, okay what is GTM algebra and the components of it, I think any motion in GTM can literally be boiled down to an equation. I think it's all equations, and I think you're always trying to figure out what the variable is that gets you ROI

it’s a trial-and-error process…

so let's say you're trying to do signal-based outbound for LinkedIn. so who are the leads you want? let's say they are looking for GTM engineers, like you want to sell GTM engineering tooling.

then from that, what signal data do they have? are they hiring? did they like a post on LinkedIn? did their company just fundraise?

so you start to map out this little equation

leads plus data, plus your offer, like what are you actually selling them? plus your sequence, which order of copy are you going to do? plus your avatar.

then really after that it’s just experimenting and testing things

who's sending matters.

okay, picture made up scenario: Elon Musk sends you a DM on X. it's some slop DM about how much he wants to exchange synergies and get on a call with you. you're not even going to care that it's slop. it's Elon Musk you're probably going to take the call and you don’t care it’s slop

well yeah, he's a big name, you know he has a lot he can offer you, you're familiar with his work. but this is just to say that who the message comes from matters a lot.

Same dm from some random faceless account and you’re just not replying, that’s very obvious

I actually co-founded a record label, and it's the exact same thing when pushing artists. Drake could say one thing and people would be like, oh yeah, cool. but then some nerdy guy in your class could put out the exact same song and people would make it a meme or just dismiss it, even though the "product" is the exact same.

so yeah, a big name is part of it. but the most important thing I think to keep in mind is that it's not only big names, it's relevance to you, who you are and the authority/ trust you have over that person.

so yeah, I would reply to Elon Musk but would some artist on tiktok, probs not a lot of them hate him. but to bring it down a level: say I'm a small MSP, and two people DM me the exact same message. one of them says MSP specialist in their bio and all their content is about MSPs. who am I going to reply to, and who do I want fixing my problems?

right, so you can gain that level of trust without being this massive name. it's the same goal, really, just that knowing who's sending matters, matters and if you’re take away is hey i need to build a personal brand or hey I need to tailor my accounts more to the ICP I am talking to that’s fine

but you need to understand this when building these big company-wide outbound campaigns or you’re ngmi

so leads, data, offer, sequence plus avatar equals outcome. the outcome of all these things is probably like hey, do you want to get on a call? so that could be a good outcome.

basically in this GTM algebra framework you just want to keep changing variables to optimize your outcome. so you might be changing the signals, you might be changing the offer slightly, maybe you're inviting them to a webinar for example, or just straight selling them. you might be changing the sequence, you might be changing the way you set up your LinkedIn profile.

so you're constantly optimizing these kind of things, and you're using Claude to basically do that as an interface. an example might be you do it on HeyReach, do it as LinkedIn, and then optimize it.

here's an example on Reddit, because I think even if something isn't fully automatable, I still think you can apply this GTM algebra framework.

so it's something like same kind of lead, same kind of offer, same kind of sequence, but instead of automating it and having the outcome be send an automated DM, what we could do is just send that lead into Slack for a sales rep to message them, and it's kind of the same thing.

so literally anything a human can do or an agent can do, you can break into this GTM algebra framework.

so okay once we find the varrible that works, what is that, that is alpha

GTM alpha is what meta, or more what competitive advantage, do you have over people. so for all the different signals and all the different data, I think what you're basically building is a small bit of alpha.

what you could do is you could run maybe 20 different campaigns, and you could say hey, on these 20 different campaigns these signals with this offer or this ICP tend to do better, and that is GTM alpha. so you're going to just keep optimizing and keep learning, keep experimenting to get more alpha.

alpha doesn't have to just be this. I think alpha could be a new channel, like how saturated is the channel.

so for example, alpha could be that there aren't that many people outbounding on X. so that could be the signal based alpha that you want to strive towards.

don't worry, it's not all going to be this complex, but this is just the framework and the terminology that I'm using, just so we have good context.

we actually did a blog post on this, but there's literally so many different types of signals. I think we did a blog and there was like 27. there's a few other people that have done blogs on this.

I think generally you want to think of signals as first party, second party, third party. is it first party, is it something where they've interacted with my CRM? is it second party, like have they engaged with a partnership event, blah blah blah? third party, like has someone said something on social media?

this could be a whole other webinar within itself because there's so many different types of signals, so I'm not going to go over it too much, but just know there is a shit ton.

one more thing before I go over the live example is this 2025 LinkedIn DM report by Matthew Lakajev. I don't know if he's using signals, he seems quite keen on human connection, but one thing that he said is it's all about building trust, and the more messages he sent and the more conversations he had over time, the more likely someone is to close.

so that is something to keep in mind. signal based outbound does work, but the limitation is if you have the right opener and you have a longer conversation with someone and build a bit more rapport, that's usually going to work better, at least from the data that he showed.

what I have with Claude right now is the HeyReach MCP, the Apify MCP, and then I have a few others, but I'm not going to go with them for now, to not confuse things. what I'm going to basically do is I'm going to say hey, scrape 100 people hiring GTM engineers using Apify.

the reason why I'm asking for people hiring GTM engineers is because if we think about this from a step back and we say hey, I'm trying to sell a GTM engineering tool, let's take my tool for example, around signal infrastructure, what we want to do then is figure out okay, if they're hiring a GTM engineer they probably have some sort of GTM engineering problems, and they're trying to fix them or they're trying to grow. so that is a very basic signal.

and then what it's basically going to do from there is it's going to bring me back some of those people. it's going to launch an actor.

then from there, I probably don't want to speak to every single person within the GTM engineering space, so I'm probably going to filter it down.

something to keep in mind is it's very good to layer signals, you don't want just one signal. so from that, can we look at now who's been active on LinkedIn within the last few months? because I saw someone else do it, I can't remember his name sorry, but he said the reply rate from that was much better, because they're actually active. you're going to outbound to them on LinkedIn and they're going to be active.

the next thing we're going to do is say hey, can you create a HeyReach campaign based on this data set, and you can do that straight through the Claude UI. so what I usually do is I go into ChatGPT and I write a lot of copy myself.

I think a big limitation so far of Claude and a lot of these tools is you definitely get really bad copy, so I wouldn't let that happen.

I think where a lot of the SDRs fall short is they try and automate it from start to finish, and they have pretty generic signals, and then you're writing the copy with them as well, so that's pretty generic. and then oftentimes their data is very outdated, so you're sending them wrong DMs. at least that's been my experience.

anytime I see a very sloppish DM on LinkedIn, it's usually from an AI SDR tool. I think the actual trust you're ruining is that when I see one of those DMs and they've got my name wrong, they haven't got my company right, they're just sending a sloppish, like hey, I don't want to work with this brand. so I think definitely be careful.

here's a very quick example of a campaign that I'm setting up in HeyReach. this is just one I set up on Monday, so a 40 connection rate, which is pretty below average to us but still a good example

one other meta I'm seeing people do is you can actually have Sales Navigator, or something like HeyReach actually has this built in, where you can check if it's an open profile or not. if it's an open profile you can send a lot of the messages, and you can do that as a signal based outbound instead of the connection rate.

so you can start to really scale up the number of connections. you can do that across multiple accounts, multiple people, and it scales up a lot.

once we've actually set up the signal based outbound, how do we actually interface with it, what's the best way to do so? here's a very quick example with fake data.

an example of something you can do is you can start to set up a Slack bot. so every time someone replies to a lead we send it into Slack, every time someone views the website or makes a post about something that's relevant, send it into Slack. so these are just ways you can interface with the data better.

some other things that we've been experimenting on for signal based outbound, especially with Scale Intelligence, which is my company for context: basically what you can do is you can start to scrape loads of signals and build a market understanding of like hey, here are all the signals within my market, and here are the industry standards of how people tend to do better, based on data we're scraping from Reddit, different benchmarks, etc.

and then you can start to have this bird's eye view of how well each campaign is doing, and by that you can start to promote and demote signal based campaigns that are doing better or worse.

so basically, in conclusion, three things I would take away:

there are levels to this. a lot of people have written off AI SDRs, and honestly I've written them off a little bit too. but the problem was never automation, it was generic signals, bad data and slop copy. with the signal infrastructure out there now you can get very, very granular.

thank you.

roman

(PS: If you want help finding your equation, or want systems like this built, I'm the founder of Scale Intelligence. We build GTM engines on a fractional GTM engineering basis, or you can use our signal infra to build your own. Apply to work with us here.)