Leevi

ecom

[RK_Ecom @rk_ecomm]

Day 1 of $1M with ecom AGAIN
It's been 448 days since I last posted.
A lot has happened since then. But a lot has still remained the same.

I have ran 2 products during this time. In the second half of 2025, I found a semi-winner.
A knee pain product I scaled to $1k/day but due to multiple DMCA-s and sizing issues with customers I had to cut it at the end of 2025.

Went back to the drawing board. I used @peralaleevi_ 's resources about building unique mechanisms and found an unique mechanism that could be implemented in a super competitive niche. Actually reached $3k/day with this product in March, with healthy 30% margins. And it's a consumable product so already had some repeat purchases.

In April, everything collapsed. My winning ad died, I could not replicate the success it had. Absolutely panicked. My CPA kept climbing (doubled by now).

I Scaled down, basically running at breakeven ever since. I've been on the edge of cutting this product for almost 2 months now. But something tells me I could still make this work. The mechanism is decent (every post purchase survey submission confirms it). I have fully custom packaging and 30 active subscribers that keep some cash flowing in, which has stopped me from cutting this product completely. I made like $12k profit in 3 months and now it's slowly burning as I have been struggling to find winning ads.

I feel like I have bit off more than I can chew by entering one of the most competitive niches in the market.

In the past month, I have been struggling with analysis paralysis. Should I keep trying, should I test a new product? This has spread my focus, my output has decreased, I am testing less and stressing more. I constantly keep looking for new products while I could be scripting new ads, putting effort into better ads.

So I decided that I will give myself 1 month to make it work. If not then I put all my focus on a new product.

The plan:
As the whole ad library in my niche is filled with AI animation ads and native statics, I will try to move in the opposite direction. I will put my effort into trying to crack a more unaware VSL style ad using again @peralaleevi_ resources around this (these are actual gold and helped me reach to $3k/day in the first place).

Let's work.

[Leevi @lepa_2]

@nor_ghost game knows game

https://t.co/Umf7FXJ4a7

[Stefan Georgi @StefanGeorgi]

Looking for a badass full-stack engineer to join the StefanBrain team. Willing to pay top dollar. And for anyone who sees this and refers someone we end up hiring, happy to shoot you a $5k referral fee as a way of saying thanks.

Details:
https://t.co/r3C9Sdob8D

[Leevi @lepa_2]

Depends, how is it self-inflicted?

If it's truly a 2 + 2 =4 thing then why would they even look for a new solution? If it's that obvious surely they'd do something about it?

If you truly can't twist the problem mechanism then it's just product mechanisms & identity. Hard to say anything concrete without knowing the market and type of product.

[Leevi @lepa_2]

Short version on the importance of mechanisms (long below):

- You probably won't need a long advertorial explaining every enzyme and receptor that is a part of your UMP.

- Simple does works (especially for low AOV.)

- Your UMP/UMS should have VARIANTS for ad types.

- I still advocate science/studies.

- Have I changed how I use UMP/UMS? Yes.

Long version:

The way I look at mechanisms is ALWAYS nuanced. The product, the problem, and the market dictates what I do. Keep that in mind when reading this, because you should not take things you see on X at face value.

If the problem I’m solving for has had scientific breakthroughs, paradigm shifts, or “overlooked” science, obviously I’m looking for ways to integrate that into the positioning.

But you have to be conscious of the market.

@StefanGeorgi put it well. Worth a read for everyone.

The market is more knowledgeable. They can Google shit and get an answer from Gemini immediately, they can open ChatGPT and ask, etc.

So the probability of you scaling some random rogue enzyme that may or may not be behind problem X is lower now than it was 4 years ago. But IQ is a curve. Some people believe the earth is flat, so some people may as well believe in the rogue enzyme.

But most probably won’t. Hence why scale will be limited.

Coming back to the paradigm shifts, you 100% can scale off of explaining a “new” cause for their problem.

And I think this is where market research is the most valuable.

Not necessarily in finding pain points, since anyone with an above-room-temperature-IQ should know the problem of the demo. If you don’t, just look at someone scaling in your niche. Or watch 5 tiktoks. Not hard. The mechanism is the hard piece to crack.

So when you research, here’s a few approaches to help you:

1. You should look at how people talk about a mechanism, how do people comment on the videos talking about a mechanism, etc.

How do they explain the mechanism? Do they use visual demonstration? Do they use metaphors? Do they link it to a known problem? Are the commenters skeptical? Are they asking for something? Proof? Explanation? Are they agreeing?

2. You should look at pharmaceutical trials of solving the problem you solve.

What’s the success rate of the top treatments? Why is it high? If it’s low, why? Is there a demographic the main pharma approaches fail for? Why do they fail for this demographic? What are scientists trying to crack? What are the proposed mechanisms?

From these you may find that solving problem X through mechanism Y is very probable.

Then you build your mechanism around that.

And the way I’d progress is –

1. Make a “core” positioning doc. The scientific UMP, and the studies that support your UMS.

2. Collect verbatim and ways to showcase your mechanism. If you can demo visually, your life will be easier.

3. Make variations of your mechanism.

What I mean by variations is that you shouldn’t use the same verbatim etc. everywhere.

You want to keep parts of it consistent on ads and landers.

But when you go and do a yapper for example, you really need to dumb it down.

It doesn’t mean you “soften” it, generally it means you cut corners so when a creator films a yapper, it’s short BUT sparks curiosity.

It has to be “scientific” enough that people won’t call BS on it and not click to your lander.

And then on the landing page you can go a step further.

Like say you say “peeing too much can cause bloating” and flash some harvard stuff on minerals.

Then your lander explains that yeah a lack of minerals can cause digestive issues since your digestive enzymes can’t activate.

And maybe throw in something about menopausal women not being able to retain minerals as effectively as younger women.

So the solution is digestive enzymes pre-paired with minerals.

Now if you do a VSL with a nutritionist you may want to go a bit deeper into the science.

You may also want to test it with regular creators (scientific explanations WORK on organic as well…)

So to summarize my point, you want to keep one core UMP/UMS, just arrive at that conclusion through different means.

Maybe you make up a person for native ads with unique life experiences.

And you use this life experience to arrive at your UMP/UMS through a way that fits the person writing it.

People love stories so may as well create them to serve your mechanism 🤷‍♂️

But coming back to Stefan’s point on market sophistication, the barrier is higher. You can’t have a BS mechanism.

If your mechanism is genuinely real, it will survive scale better and have better longevity.

There’s been cases where I’ve been pushing a UMP based on my research, and later the science has “caught up to it.”

Doesn’t mean it was BS in the beginning, there was real research behind it, then it just got proven in a massive study.

So all-in-all, mechanisms are nuanced. It depends so much on your product and the problem it solves.

As long as you avoid the obvious mistakes (inability to make a simple variant of your mechanism for ads, having a rogue enzyme with little backing, saying X is not the mechanism behind problem Y when pharma studies show a fucking 90% success rate when treating X, etc.) you’ll ball.

[ceoxsndre @nor_ghost]

@peralaleevi_ https://t.co/rxko3Sn5Cy

whats ur thought on this?

[Leevi @lepa_2]

Lot of talk around IM8 recently and there's a point that everyone should consider if they try to mimic them.

Quarterly sub ≠ more revenue.

It's revenue pulled forward.

You can argue all day about increased LTV etc with quarterlys but it doesn't matter until your own cohorts prove it.

Now for you it might make sense to pull revenue forward but it always comes at a cost. Volume loss, higher nCAC, etc.

EOD it's just data. But if you're aovmaxxing just for the sake of D1 profitability & being able to spend more just know that it can turn into a slippery slope and bite back.

[Leevi @lepa_2]

@lolo_ecomz Could be worse, have seen quite a bit of products with holy ingredients from the bible...

[Leevi @lepa_2]

100%. Pure "saving" (aka cutting out software we had in the past) is maybe $200k to $300k annual.

I haven't cut headcount, I don't see a need to as long as every role brings more in than it costs and doesn't increase ops complexity to the level where attention to headcount takes away attention from higher-leverage actions.

And purely from an acquisition standpoint I refuse to give out stuff like video editing on CORE assets (more leverage in having well-edited ads VS churning volume with lower quality), writing core copy (AI can't crack a positioning the market hasn't seen yet as well as humans can, atleast from my experience.)

The view I hold is that anything where taste, pattern recognition, and intuition is of leverage, I'd rather have someone with taste and intuition to, rather than AI.

AI can see patterns from data but if the pattern you're seeking is say why does X work, you can only be so analytical about it. Feel like AI has a bit too much confirmation bias and lacks ability to make connections beoynd the surface level.

[Leevi @lepa_2]

Since posting this we've built a shit ton of internal tools with AI.

The revenue impact is at least 8 figures plus cost reduction. All of it from thinking on first principles.

And all of it is "simple."

It works because I'm in the ops every single day and I see exactly where to squeeze performance. There's a million things that we could build, but only a handful have a real leverage.

Point still stands on positioning etc. AI gives you an edge only if the fundamentals are there.

Even if you mimics say big pharma's AI methods in drug discovery for your mechanisms, you still gotta have that intuition to know which way to position your product.

[Leevi @lepa_2]

If you're doing under $300k per month with ecom, you should not be in the trenches with the new AI stuff.

THIS IS NOT your lever yet.

You don't need to build agents for media buying, making static ads, etc. It's just a shiny object.

There is nothing (except you being distracted) stopping you from making one in-feed VSL that gets you to a 7-fig monthly run rate.

... which would make it easier for you to HIRE people better than you at AI to help you build the systems.

AI is 100% a lever when you have the basics down. It's been a lever for years for copy, research, etc.

But it doesn't matter how cool it is to one-shot an advertorial with Claude if your product positioning and copy are shit.

Don't get distracted. You won't be left behind because you didn't adopt AI. You'll be left behind because others got the fundamentals down better than you and then started running laps around you with AI.

[Leevi @lepa_2]

@JuggMotion Hit and miss, sometimes lower CPM sometimes higher, CVR swings, NVP% swings etc etc

[Leevi @lepa_2]

Probably should ask them why they want to run a different mechanism. Is there a hunch it'd perform better? Are they seeing issues with the current one?

EOD if they're incentivized based on performance you would assume they want to use winning elements.

Every good CS knows you can't recycle a winning ad over and over again. But mechanisms don't fatique like ads do. And to your point it's operational hell to have too much going on (can hurt the brand + you need to have variants of funnels to match the mechanism).

[Leevi @lepa_2]

@holdrate I mean I do this too, it's a part of testing

[Leevi @lepa_2]

@holdrate That's how I run it, it's a meticulous system :)

[Leevi @lepa_2]

Depends on the level of the CS.

If they’re good I don’t see a reason to limit their creativity. As long as what they come up with is (a) true and (b) aligned with the core idea, that’s enough. Obviously you should intervene if whatever they’re doing isn’t working. But the core positioning + studies should be enough.

Now if they’re more junior, it helps you and them if they can use a proven-to-work way of presenting the mechanism. And it’s up to your training and their skill level to be able to distinguish between the mechanism being surface level on the ad vs a deeper dive.

+ if there’s a winning “way” of presenting the mechanism I don’t really see a need for the CS to swap it out. Obviously if you have a hunch something could work better, try it. Regardless of the seniority level. But if something is proven to work, you can just reuse it across formats.

[Leevi @lepa_2]

@joeysamiir @conortrains If there's a clear demo difference then yeah, otherwise I don't personally do this too much

[Leevi @lepa_2]

One reason I'm so bullish on quizzes.

Have personally written very very broad videos with individual ads spending $500k+, $1m+ each, all pointing to a quiz that basically covers every "sub avatar."

I can see the argument that if you're a new player stepping into a field with big brands (that actually have good product), you'd want to make your ads super niche to stand out. But to be fair, people overplay the importance.

As long as you have a mechanism to differentiate and create enough intent for them to go through quizzes etc, you'll ball.

[Conor Sunderland @conortrains]

Question for big scalers on here:

Are avatars just bullshit?

Don't most avatars have way more in common than they have not in common? And if so, wouldn't that mean a banger ad for the general avatar can sell to all the subs, making that more effective than niching your ads?

[Leevi @lepa_2]

@conortrains Depends a lot on niche/product, have pushed skincare quizzes at 70% - 80% of total spend.

The data is great, but for me they convert well too. Natives, VSLs, yappers, etc all of it (also works for bottom of the funnel).