The Purple Ocean Engine

Everything I promised on stage. Nothing held back.

You scanned a QR code in Budapest — or someone who did sent you this link. Either way: this is the full system from the talk, written down so you can run it.

The advertorial I opened the talk with has done over €18,000,000 in revenue to date — and it took about two hours to build, using AI. Not because of a magic prompt, but because of the system behind it. This page is that system.

On stage I had 25 minutes. That was enough to show you what the Purple Ocean Engine is and why it works. It was not enough for the hands-on parts: the research workflow, the exact prompts, the tool decisions. That's what this page is for. If you were in the room, the three steps will feel familiar — but every section here goes one level deeper than the stage version, and everything you need to execute is included.

One promise before we start, because it's the same promise the whole system is built on: every number on this page comes from published research. Every study is cited with its PMID in the appendix, and you can verify each one on PubMed in ten seconds. The system preaches "no source, no sentence" — so this page practices it.

Throughout, we'll use the same worked example from the talk: a (fictional) cushioned work shoe for women who spend their entire workday on their feet — nurses, kindergarten teachers, hairdressers, cashiers. The product is invented. Every study behind it is real.

Every vertical is a red ocean — and AI made it worse

Supplements, beauty, health, food — whatever vertical you're in, whatever GEO: countless competitors, rising customer acquisition costs, dropping AOVs. Proven demand, brutal competition. That's the red ocean, and in 2026 there is no escaping it by "finding a niche" the old way.

Then AI accelerated everything. For years, the players with the highest creative output won. That edge is gone: everyone can pump out a new campaign in 30 seconds now. Output multiplied across the whole market — and quality collapsed. If your performance is dropping even though you're producing more than ever — that's this, not your media buying.

The root cause: AI doesn't give you the best content

Here's the technicality you need to understand, because the entire system is built on it. A language model was never designed to give you the best content. It's trained to predict the most probable next word — and "probable" means: what it has seen most often. So by default, AI gives you:

• the most probable,
• the most average,
• the most seen-before.

So when people brag "with AI I produce ten times more, ten times faster," they're producing ten times more of the most average thing on the internet — the same five angles, the same voice, written by them AND their competition, from the same model. And "most seen-before" also means: most seen before by your audience. That is the real cost of slop — not that it's bad writing, but that it makes everyone sound identical.

Red, blue… and purple

You know red ocean and blue ocean. Red: proven demand, brutal competition — you fight everyone on price and CPMs. Blue: no competition, which usually means no market — and a fortune spent educating people about a problem they don't know they have.

There's a third color. Purple: a specific, underserved micro-segment sitting right inside the red ocean. A blue ocean hiding inside your red ocean. (And no — "purple ocean" isn't my invention either; it comes from strategy people who realised blue oceans never stay blue. Everything gets copied, and in the AI era it gets copied instantly.)

Purple inherits the good half of each:

• From red you inherit proven demand. You never have to educate anyone that the problem exists. She already knows her back hurts; she's already been buying solutions.
• From blue you inherit no direct competition. Nobody speaks to these exact people directly, so there's no CPM war for their attention.

Why now? Because it never paid — until AI collapsed the cost

Nobody targeted these micro-audiences before because it simply didn't pay. A properly researched advertorial for a micro-segment used to take a team days, sometimes weeks — and you couldn't even be sure it would work until the end. You can't justify that for a segment of eighty thousand people. AI took that cost to near zero. What was economically impossible is suddenly profitable — lower CPMs because nobody talks to these exact audiences, lower CACs, higher AOVs. (With Meta's Andromeda it got even better; more on that in Step 2.)

That's the entire window the Purple Ocean Engine exploits. And it's open now.

The Purple Ocean Engine — three steps

If this system converts deeply skeptical Germans at scale — and it does, we run it in the arguably toughest market in the world — it will convert your traffic too. Whatever GEO, whatever vertical.

I've broken it down into three steps you can steal from me today:

Six copy-paste prompts are embedded along the way — research (1, 2, 3), purple-ocean discovery (4), voice-of-customer clustering (5) and the full editorial rewrite (6).

1.1Your prospect is burned — and her bullshit-o-meter is on

Whatever your vertical: your prospect is not a fresh, hopeful buyer anymore. False and exaggerated promises, bad products, and solutions that simply didn't work for her have made her deeply skeptical. She has bought product after product that promised to fix her problem — and every one let her down. By now she has a built-in bullshit-o-meter, and it is very, very sensitive.

I learned to convert skeptics in the hardest room on earth: Germany. You cannot just tell a German something about your product and expect them to throw money at you. You have to deeply convince us. You have to prove it — and prove it well, or we simply won't believe you. Converting ice-cold traffic on first click is already hard; doing it with Germans is the final boss. Master that, and the discipline transfers to any GEO.

This is why it's not enough to sell "the old way." To convert these prospects you have to break an old belief system they hold, and install a new one. And you do this best with logic, science and truth.

Truth beats scam. 24/7.

Watch what many marketers do when things get tough: they drift to the dark side. Here a fake doctor, there an invented study, everywhere claims screamed a little louder. And aside from the ethical problem — it doesn't even work better. Your prospect's bullshit-o-meter smells it from miles away. Louder just proves you're one more liar.

That's why I defend one sentence like a religion: truth beats scam, 24/7. It doesn't just feel better — with AI, it now also works better and faster. And it does something brutal to your competition: when you show up with a real mechanism and a real study, every competitor running a vague claim suddenly looks like a scammer next to you. You don't attack them. You don't even mention them. You raise the standard of proof, and everyone who can't meet it disqualifies themselves.

1.2The chain every advertorial runs on

Every advertorial that works is the same machine under the hood — one specific chain:

• PROBLEM — open on her pain. Concrete, specific, her actual day. Not "back pain" — "by the end of every shift, her lower back is done."
• FAILED SOLUTIONS — everything she's already tried, and why it let her down. Name them. She needs to see her own history on the page.
• ROOT CAUSE — the real, physical reason nothing worked. The thing none of the failed solutions ever touched.
• UNIQUE MECHANISM — how your solution addresses that root cause, logically, where the others physically couldn't.
• OFFER — only now: why you, why now, what her life looks like after.

Knowing the chain is table stakes — every decent copywriter knows some version of it, and there are countless winning advertorials out there you can swipe for structure. The edge is understanding that the whole chain exists to produce one moment: the belief shift, right at the center, where ROOT CAUSE hands over to MECHANISM. If you get the unique mechanism right, that's 80% of the work. That moment is what we build now.

1.3The worked example: three numbers, one belief shift

Meet your hardest prospect. She's 41, a nurse. Ten-hour shifts, almost all of it on her feet. By the end of every shift her lower back is screaming. She's tried painkillers, physio on her days off, an expensive new mattress. Nothing really moved it. So she carries one firm belief:

"My back is just wrecked from work. Nothing helps — that's the job."

That belief is the wall between you and the sale. You can't climb it by shouting your claim louder — she's heard "fix your back pain" a hundred times. Break the wall instead. A belief shift has exactly two halves:

• BREAK the old belief (failed solutions + root cause): show her why everything she tried couldn't work — not because she's unlucky, but because of a root cause none of them ever touched. Her belief moves from "nothing works" to "everything I tried couldn't work."
• INSTALL the new belief (mechanism): show her, logically, why your solution works where the others physically couldn't.

Here are both halves live, running on published research and nothing else. Our fictional product: orthopedic work shoes against back pain. Note that the mechanism doesn't revolve around back pain in general — it revolves around back pain from standing too long. That specificity is everything.

THE PAIN, quantified — the Unique Mechanism of Problem. A systematic review of laboratory studies found that the average person develops clinically relevant low-back symptoms after about 71 minutes of continuous standing (Coenen et al. 2017). People prone to standing-induced back pain get there even faster — around 42 minutes (Khoshroo et al. 2023). A cashier hits that before her first break. And the root cause is almost absurdly small: after about an hour of uninterrupted standing, your body sinks into a hollow-back of about 4.4 degrees — and this tiny postural shift overloads the small joints of the spine and creates measurable back pain (Sorensen et al. 2015). Not a weak back. Not bad luck. A few degrees of posture, hour after hour.

THE BREAK. Now watch what that root cause does to everything she tried: the painkiller numbs the signal — but she's still standing at those extra degrees. The physio session on Sunday can't fix the eight hours she stands on Monday. The mattress? She's not in pain lying down — she's in pain standing. None of them ever touched the root cause. Old belief broken: it's not her fault; everything she tried was aiming at the wrong target.

THE INSTALL — the Unique Mechanism of Solution. If the strain builds through the hard ground under her feet all day, change what's under her feet. In controlled studies, a cushioned surface cut perceived standing back pain by about 47% in exactly the people who develop it — same person, same standing time, only the ground changed (Winberg et al. 2022; supported by Aghazadeh et al. 2015). Our fictional shoe is that cushioning, carried with her all day — working during the shift, where painkillers, physio and mattress only act after it.

And now watch what happens, purely from turning study facts into your mechanism. She finally understands where her back pain really comes from (the 4.4° arch). She understands why no other solution could ever help her — the root cause was never fixed. And your product becomes the only logical solution: she now knows only cushioning under her feet, during the shift, will ever fix her standing pain. And guess who has an absolute no-brainer offer ready for her on this same page where she just understood all of this. (You.)

Her years of disappointment now make sense, and every failed fix becomes proof for the new mechanism. That's the most beautiful judo move in direct response — her skepticism starts working for you.

Notice: the honest version is not weaker. It's sharper — because it survives the one reader who checks, and because precision itself signals credibility to everyone who doesn't.

1.4How to find YOUR mechanism — with Claude and the actual science

This is the part I could only tease on stage. It's simpler than you think — four steps, one discipline.

The connector workflow (what I showed on stage)

Claude supports connectors — integrations that let it search external databases directly instead of answering from memory. Several free scientific connectors give Claude live access to research databases like PubMed (the U.S. National Library of Medicine's index of medical literature: free, public, no login — pubmed.ncbi.nlm.nih.gov). In Claude, open Settings → Connectors, browse the directory for a PubMed / scientific-literature connector, and enable it.

Then the workflow is exactly what you saw on stage:

1. Give Claude your customer avatar — for us: "people with back pain from standing too long."
2. Ask it to search the literature for the root cause of that symptom and to build your Unique Mechanism of Problem and Unique Mechanism of Solution — citing every study it uses with its PMID (Prompt 1 below).
3. Verify each PMID yourself. Open every citation on PubMed and check that the abstract actually says what the mechanism claims (Prompts 2 and 3). Claude found the studies behind our three numbers within minutes — but the numbers only went on stage after each abstract was read.

No connector available, or you're working outside medicine? Do the same thing manually: search PubMed or Google Scholar with 3–5 search strings built from symptom + population + suspected mechanism (e.g. "prolonged standing low back pain", "lumbar lordosis standing pain development", "anti-fatigue mat low back pain"), filter for meta-analyses and systematic reviews first, hunt the abstracts for concrete numbers — minutes, degrees, percentages. Small precise numbers ("4.4 degrees") persuade better than big vague ones ("up to 90%!").

Run Prompt 3 on our worked example and it flags exactly the boundaries in the honesty box above — acute vs. chronic, mats vs. insoles, perceived vs. objective. That's the whole discipline, automated: find the real mechanism in your vertical, and you never need a fake study again.

1.5Build the mothership

Now assemble the advertorial itself: the full belief-shift chain, powered by your study-backed mechanism — on a slightly broader angle. For our example: not "back pain" (too red), and not "nurses with back pain" (not yet — that comes in Step 2), but "back pain from working on your feet all day." Deliberately broad enough that the nurse, the teacher AND the hairdresser can land on it and convert — because your mechanism is about standing, and it's strong enough to convert even very skeptical readers.

We call this the mothership advertorial. It's your only significant upfront build — a day or two, with AI doing the assembly on top of your research. Editorial look, on its own domain (never the shop — more on that in Step 3). Why "mothership"? Because in Step 2, a whole swarm launches from it.

2.1One mechanism. Hundreds of purple oceans.

You already built your advertorial and your unique mechanism. Now give both to Claude and ask it for a deep research: which hyper-specific customer avatars are affected by back pain from standing too much? AI will surface purple ocean after purple ocean — more than you could ever probe at once:

Each one: a segment nobody writes for — because until now it was too much work and too risky to find out if it converts.

Level 2: ground it in real customer voice

The deep research gives you the map. If you want her actual words for each territory — the raw material your future headlines are made of — mine the voice of your market: product reviews (yours and competitors', especially 2–4 star), Reddit threads where your prospect complains in her own words, comments under competitor ads, one open post-purchase question. Paste it all raw: typos, rants, emojis. The mess is the signal. Then cluster it:

The verbatim quotes this returns — "my back is done by hour 10," "9 hours behind the chair" — are your future hooks, headlines and UGC scripts. Language that stops the scroll because she recognizes her own sentence.

2.2Probe: ultra-specific ads for micro-audiences

Now — do NOT write seven ultra-specific advertorials. You'd be guessing which micro-angle wins, and guessing is expensive, even with AI. Instead, you fire ultra-specific static ads at the micro-audiences — the nurse, the teacher, the hairdresser — and every single one points back at that same slightly broader mothership advertorial with your killer mechanism.

View the ads as your cheap little probes — each one testing a purple ocean for a few dollars a day. You're not committing to anyone yet. You're asking the market a question.

2.3Read the data — one angle will print

One of those angles is going to win, because your mothership still converts for all of them. Read the probes in this order: CTR (did her language stop the scroll?), cost per advertorial read (did the click hold?), CAC/ROAS per probe (did it convert downstream?). A probe that wins on all three is a validated purple ocean. And the winning ad tells you more than "nurses won" — it tells you exactly which pain phrasing stopped her scroll. Keep it; it feeds everything in Step 3.

Let's say the nurse ads won. Now comes the genius move.

2.4The one-hour rewrite

A few months ago, tapping a micro-audience like this was absolute madness — the workload made no economic sense. Now you take the broad mothership and let AI rewrite and completely revamp it solely for your nurse — not just the headline: the root-cause story, the mechanism framing, every example, every scene, first line to last. Give it customer research for the nurse-prospect and it rewrites the whole thing. Before AI: a week of work you'd never invest in one micro-niche. Now: about an hour.

The moment you ship it, the nurse sees a funnel so specific, talking so narrowly about her actual life, that she has no choice but to see you as THE expert for her problem. The most niched-down product in the world for her specific pain. Perceived relevance — and with it conversion — jumps.

Then read the output against your citations file before shipping — rule 2 gets checked by you, not trusted. (Prompt 3 works on finished copy too: paste the rewrite and your abstracts, and let it flag any claim that drifted.)

3.1Pile onto the validated angle

The angle is VALIDATED now. You know it converts. So you double down on it — every asset type, all aimed at the same target:

• A fresh ad set of statics — all on the nurse angle, each built from the pain phrasing the probe surfaced.
• A video sales letter — scripted from the exact language of the winning probe and your mechanism.
• UGC creators — briefed on the same angle: "POV: hour ten of your shift."
• Social proof, cartoon ads, advertorial variations — whatever formats work in your account, now with zero angle-risk.

Most people test creatives hoping to stumble onto an angle. You already found the angle — your creatives just multiply it. Every asset is aimed at a target you've already hit once, which means you take almost no risk investing time and money into these purple-ocean audiences.

3.2What happens to your numbers

Once you're the only one actually speaking directly to your purple ocean, the numbers start moving in every direction at once:

You're the only one in her feed who obviously understands her job. Relevance is the cheapest performance lever almost nobody in your market is pulling, because almost everyone is pointing AI at "more," not "deeper."

The two boring foundations (that make the data trustworthy)

• Every advertorial lives on its own editorial domain — never the shop. Credibility: an editorial page on a shop domain reads as what it is, an ad; on its own domain it reads as content. Tracking: a dedicated domain gives you a clean funnel stage you can measure probe by probe.
• Keep cross-domain tracking clean. When traffic hops ad → advertorial domain → shop, your click IDs and UTM parameters must survive every hop into checkout — otherwise "read the data" reads noise and the engine optimizes on fiction. Set it up once, verify it in a real browser before scaling spend, re-verify whenever you add domains.

3.3Rinse & repeat — the moat

Then you do it again. And again. One purple ocean after another, each with its own probe-validated advertorial you can rewrite and ship within an hour. Run this for a few months and you've claimed five, six, seven micro-niches — with hyper-specific dedicated funnels — before your competition notices even one.

And when they finally do? It takes them weeks to notice WHAT you're doing, and months to understand the SYSTEM behind it. From the outside they see one ad. They can't see the research, the probes, the citations file, the rewrite loop. By the time they've reverse-engineered one funnel, you're four niches further and the established name in every one of them.

That's a structural moat, not a campaign edge.

The tool stack — honest version

No affiliate links, no hype. The system is tool-agnostic almost everywhere; here's what each job needs and what we actually use.

Total stack cost at the start: essentially your AI subscription. The moat isn't in the tools — it's in the discipline of the research and the loop.

Three things to take home

And if the talk felt persuasive — notice what happened to you. Problem, failed solutions, root cause, mechanism, proof. You watched the chain run on yourself. That's the Purple Ocean Engine.

Every study behind every number

"No source, no sentence." Here is the full evidence chain. Every PMID opens on PubMed — verify anything in ten seconds.

A. Standing → acute low-back pain (the dose)

B. The root cause (posture, not bad luck)

C. Why the standard solutions miss

Note the logic discipline: painkillers, physio and mattresses "missing the root cause" is a mechanism argument (they act after the exposure, not during it) — there are no head-to-head RCTs to cite for those, so we argue it as logic, not as "studies prove they fail." Only the insole claim has direct trial evidence.

D. The mechanism solution (cushioned ground)

E. The boundaries (what the evidence does NOT say)