We Reverse-Engineered Comfrt and Neurogum's Million-Dollar Affiliate Engines

"I sent 5,000 samples to creators and got five videos back. I made twenty grand and I don't know how to scale it."

A founder said almost exactly that to me recently, and it's the most honest summary of how most brands "do" affiliate and creator marketing that I've heard. You buy a pile of product, you mail it out, you cross your fingers, and you treat the trickle of posts that come back as the program. Then you wonder why it stalls.

Meanwhile a handful of brands are running the same channel as a genuine engine. Comfrt is selling at a scale that doesn't look real until you see the screenshots. Neurogum has built a creator system so self-sustaining they say it almost runs itself. Both are wildly public about how they do it.

So I pulled both engines apart to see what's actually inside, and then I did the bit nobody does, which is work out what a normal 7-figure brand could copy without a cult following or a war chest. Here's the teardown.

The default that doesn't work: spray and pray

Let's be fair to the spray-and-pray model first, because most of us started there.

You find people with reach, you send them free product, maybe you pay per post, and you hope something sticks. There's no click data, no sales data, it's impression-driven and it's basically a lottery ticket. The "strategy" is volume of outreach.

The reason it stalls is that you're recruiting people who already value their own reach, which means they negotiate hard, they post once, they fight you over the caption, and they're gone the second a better offer lands. You're renting an audience for a single exposure. There's no relationship, so there's no second video, and definitely no fiftieth.

Now here's what the brands that actually print money do instead. It's not one trick. It's four mechanics stacked on top of each other.

Mechanic 1: recruit for zero followers, not big ones

This is the part that breaks most people's brains, so sit with it.

The brands winning hardest deliberately recruit creators with as few followers as possible. Not despite the small following. Because of it.

The logic is sharp once you hear it. The algorithm rewards good content, not follower count, so a great video from a nobody can outperform a mediocre one from a name. More importantly, someone who has never made money online and gets their first cheque from you becomes a loyalist in a way a 200k-follower influencer never will. You're not renting their audience. You're building one with them, and they remember who gave them their start.

Comfrt's founder, when asked how many followers he looks for, apparently answered "as little as possible." He's run tens of thousands of complete unknowns through his system. Neurogum's whole creator base is built the same way: cultivate people, don't just rent the famous ones.

Compare that to spray and pray, where you chase the biggest accounts you can afford. You're paying the most for the people least likely to ever care about you.

Mechanic 2: turn recruiting into a challenge, not a contract

So how do you find the good ones out of thousands of unknowns? You don't audition them. You run a challenge and let attrition do the work.

The headline version Comfrt ran: post a thousand videos in a month, and you get a guaranteed payout. Thousands of people enter, dreaming of the payout. To hit it you'd have to post ten-plus times a day, which almost nobody can sustain. By week two, the vast majority have dropped out, and every video they already posted, you got for free.

But the people still standing at day 30 have just done a month of daily reps. They've turned into genuinely good creators by sheer volume, and they made commission along the way. You didn't train them. The challenge did.

I love this because it mirrors how creative testing actually works. You don't find a winning ad by being clever, you find it by making a lot of ads and letting the data pick. Recruit fifty thousand creators to find the five hundred who can really do it. It's the same maths as making a pile of ads to find a handful of winners. Volume in, winners out.

The contrast with spray and pray is the whole point: one mails 5,000 boxes and hopes, the other sets a clear bar and lets people prove themselves against it.

Mechanic 3: the home base, the tiers, and the onboarding

Recruiting gets you bodies. Keeping them and making them good is a community job, and this is the bit the trickle-of-samples brands skip entirely.

Both engines run on the same three parts:

• A home base, usually a Discord, where creators live, talk to each other, and swap what's working. Once creators start comparing notes, the system starts improving itself. A hook that's working spreads creator to creator faster than you could ever brief it.
• A tier system, so there's a ladder to climb. You hit a sales bar to get in, and you can work your way up through levels. People stay because they can see the next rung, and because falling off the bottom is a real thing. Status does a lot of work that money alone can't.
• A real onboarding, not a welcome email. Neurogum makes every creator sit through a recorded orientation, around three hours of it, and a brand manager follows up with specific questions to check they actually absorbed the product. The point is simple: a creator can only add their own spin on top of your hook if they genuinely understand the brand. Thin understanding makes thin content.

Then there's the cultural glue on top. The founder is in the Discord pumping people up. There are regular group calls. Top performers get pulled in to tell everyone else what's possible. The best programs make creators feel special on purpose: the top few get flown somewhere, get on a billboard, get the nicest box, while the rest can see exactly what they'd need to do to earn it. Make a few people visibly special and the rest push to join them.

Spray and pray has none of this. No home, no ladder, no orientation, no glue. Just a box on a doorstep and silence.

Mechanic 4: pay per sale, and pay across platforms

The money model is where the cash-flow magic hides, and it's worth copying carefully.

These programs pay per sale, not per post. That's a real cash-flow advantage people underrate. You collect the customer's money first, then pay the commission days later once the sale is validated. Unlike paid ads, where you pay upfront to maybe get a sale, you're paying out of revenue you've already banked.

The second move is the clever one. They don't just pay commission on the platform where the post lives. They take the best creator content, run it as paid media on other channels, and still pay the creator a cut of those sales too. The creator's content becomes the fuel for the paid engine, and they get rewarded for it, so they keep feeding you. It's about as fair as a deal gets, and fairness is exactly why the good ones stay.

The honest caveat: this needs margin. Stacking a commission on top of your ad costs only works if the unit economics can carry it. If you can give, say, a tenth of the sale to creators and another chunk to Meta and still come out ahead, you're in business. If you can't, fix the margin before you build the program.

Now scale it down: what a 7-figure brand actually runs

Here's where I depart from the case studies, because "post a thousand videos" and "fly the top ten to an event" are not where a 7-figure brand starts. The principles travel down. The numbers don't. Here's how I'd shrink it.

Recruit from people already in your orbit. You almost certainly have a small seeding list of people who already post about you and never get nurtured. Start there before you cold-outreach a soul. The expensive part has already happened.

Run a 50-video challenge, not a 1,000-video one. Same mechanic, sane scale. "Post 50 videos over the next two months and hit a small sales bar, and I'll guarantee you a payout." It self-selects the committed, gets you a stack of content cheaply, and surfaces your two or three real creators without you having to guess.

Pay per sale, on terms you can afford. Set a commission your margin can genuinely carry, and remember it only applies to first-time orders, so the true blended cost across all your sales is lower than the headline rate looks. The repeat purchases you don't pay commission on are where the economics actually work.

Build a tiny home base. One Slack or Discord channel. You don't need a thousand people in it, you need ten who talk to each other and a founder who actually shows up. Ten engaged creators comparing notes will out-produce a hundred who never hear from you.

Make the onboarding harder than feels comfortable. A short brand video and a couple of real questions. Make it just demanding enough that the half-interested drop out before they waste your product. The friction is the filter.

Then slot paid Meta in behind it. This is where it stops being "an affiliate program" and becomes a flywheel. Take the organic content that's working, the stuff that already proved itself in real feeds, and run it as paid creative. You get ads with genuine social proof baked in, and you pay the creator a slice of those sales so they keep making more. Your best-performing organic becomes your paid pipeline. That hand-off, organic proving the concept and paid scaling it, is the whole game.

What actually makes it work

Strip all four mechanics back and the same truth sits underneath: this is a relationship business wearing a marketing costume. The brands winning aren't chasing the famous mercenaries who jump to the highest bidder. They're building their own believers from scratch, treating them well, and paying them fairly across every place their content drives a sale.

The reason most brands won't copy it isn't that it's secret. Both of these founders told the whole room exactly how they do it. It's that it's a grind: real recruiting, real onboarding, real time in the Discord every day, real fairness on the payouts. Most people hear that and quietly decide it's not for them.

Which is the opportunity, isn't it. If the playbook is public and the only thing protecting it is that hardly anyone will actually run it, the question isn't whether it works. It's whether you'd rather mail 5,000 boxes and hope, or build ten people who'd run through a wall for you.