The thesis's baptism by fire: $3 CPC — how a signal becomes data in allbound.

I'll be honest — I wanted to test myself.

Outbound, when you're running a solo business, gets fragile fast: one burned domain or number, and the whole business shakes. The textbook fix is to add paid and content — but that's exactly the catch. Adding channels usually means adding cost and noise: higher CAC, and one more brand saying what every other brand says, only louder. In a market where AI collapsed the cost of making content, sameness is the default and attention is the scarce thing.

So the real challenge wasn't “add a channel.” It was: how do you open the whole allbound surface — paid, content, inbound, landing pages — without inflating CAC, and without sounding like the feed everyone's already tired of?

Twenty years ago, when Aaron Ross was building the outbound playbook at Salesforce that became Predictable Revenue, inbound was the long game — a slow, sometimes tiresome organic motion of classic content marketing (which, for the record, I love). The thing is, today you can theoretically compress months, even years, of that work into a couple of weeks with paid ads.

You'll hear people say Meta ads have gotten expensive and don't scale without burning through piles of money. The way I see it, those are the same people who complain about outbound — because in both motions, they bet on volume.

The secret was never volume. It's calibration, perception, and a sniper's aim.

This is where the Ouroboros closes the circle: the same dataset that feeds the BDR and the call briefings now runs every motion of my GTM — or, as the new kids call it, “allbound.”

The thesis stopped being a theory and became a tested framework: one source of buyer truth and conversation intelligence, feeding outbound, paid, content, offers, and inbound alike.

This is the one flagship that isn't an n8n workflow. No webhook, no state machine. The machine here is a process, and I'm a step inside it — which is the point.

The raw material never changes: what the buyer actually said. The whole build is about turning that into every ad, every post, every page, without bleeding out the thing that made it true.

Every angle starts as someone else's sentence.

Before anything creative, there's the research, and it's less glamorous than it sounds. I query my own conversations in SQL — the WhatsApp threads, the call notes, the objection that came up twice — and pull the raw language straight out. Then I go where the market talks when I'm not in the room: the communities my ICP lives in, mined the same way. Not paraphrased, not “insights.” The actual words, typos and all.

One source is what buyers said to me. The other is what they say about the problem when no one's selling. Between them, there's no room to guess.

Raw verbatim is chaos: hundreds of quotes and no shape.

So I built a signal taxonomy. Every quote gets a type — operational pain, financial pain, an objection, an explicit desire, the desire people won't say out loud — and then each one gets ranked, because each type carries its own downstream field: what it means for the offer, for the SDR, and, the one that matters here, for content.

So now, instead of an invented persona/avatar, I have a real one: the buyer's actual pains and desires, in the order they actually drive them.

Here's the part everyone gets wrong — and I got wrong too, at first.

The data doesn't just tell you what hurts. It tells you the order: what matters most to this buyer, then next, then next. And the counterintuitive thing I kept hitting is that the number-one pain is usually the worst thing to lead with.

Take money. For my ICP, making more of it is often the deepest driver — and precisely because it's the deepest, every competitor is already shouting it. Say it the way they say it and the defenses go up on contact; the ad slides past, unseen. Attention is jagged that way: loudest exactly where it's most defended.

The opposite failure is just as common. Saving time is real, and it ranks — but on its own it rarely carries enough urgency to make someone buy today.

So the angle is never the top pain, and never a random one. It's a calibration: the signal high enough on the priority order to move someone, sitting in a lane the competitors left quiet. That's the sniper's aim. Find it, and the rest is reverse engineering — the angle sets the copy, the copy sets the promise, the promise resolves back into the offer.

The agents do real work here: they process the data, digest it, draft. But the taste and the decisions can't be delegated without the output getting worse — and, downstream, the performance with it.

This is the part I'd defend hardest. As scientific as marketing gets — and with datasets, psychology, game theory, all these artifacts, it gets pretty scientific — I still think the human mind, and taste and intuition especially, is the one piece that can't be automated yet. You see, a good chunk of what we call thinking comes from exposure and lived experience. The way someone sees a problem isn't built only from the context in front of them.

So I have the agents generate and brainstorm with me off the data — a couple of iterations with a friend who keeps handing you new ways to see the thing — and that makes nailing the right angle far easier.

From there a content pipeline takes over: a handful of agents running my own frameworks and thesis — Figma over MCP for the creative, the pipeline for the words — so every ad, every post, every landing page, every email stays coherent. The same true thing, said once, everywhere.

Then the ads run, and the report is a wall of numbers.

I have Claude read it over MCP and hand me a digest — what moved, what stalled, what's worth another dollar tomorrow. It's the same loop as everywhere else in the business: measure, read, adjust, and feed what I learn back into the next angle. The tool makes the read fast. The call on what it means is still a call.

None of this means anything if you can't trust the number under the click.

Browser pixels have quietly stopped working — cookies gone, iOS locked down, blockers everywhere — so a good share of the conversions a normal ad account reports are really educated guesses. I moved the measurement server-side instead, through Cloudflare Zaraz: a custom lead event fires the moment someone actually opens the WhatsApp conversation, sent straight to Meta's Conversions API, with Microsoft Clarity watching session behavior and Umami keeping a privacy-first count alongside. So when I say $3 CPC, it's counted on an event I own — a real person starting a real conversation — not a browser pixel guessing on my behalf.

As much as taste can't be automated, part of it is still intuition, even with data backing it up. That's the reason we test things.

I ran an A/B on the creative, and I was sure how it would go: the version with a man in it would win. I had reasons — I could have argued them to your face. It lost. Not by a little, and not ambiguously; the data pointed the other way and kept pointing.

Taste sets the hypothesis; it doesn't get to be the judge. The reason I measure conversions server-side, down to a real event, is precisely so that when my intuition is wrong I find out in days instead of quarters. The taste is the edge. The measurement is what keeps it honest.

One number carries this page, and I want to be precise about it.

$3 CPC on Meta, held for a month — not a spike I screenshotted on a good day, a cost I sustained. In a market where everyone complains that Meta ads only get more expensive, the signal-first angle bought attention cheaply and kept buying it. That's the proof the thesis crosses channels: the same buyer language that runs the BDR, priced in the open market and holding.

What I won't do is dress it up with volume I can't stand behind. The CPC lives in the ad manager, not a database I can export for you — so I'll show it as what it is: documented, sustained, and mine to defend on a call.