All the signal from my 1,000+ tweets in one knowledge base

Don’t be afraid to test new things in your ad account: Meta doesn’t launch features randomly. The budget sharing feature recently added to ABO is crushing on several product campaigns after switching. Also play with settings: 1-day click vs 7-day click attribution, incremental vs standard attribution, bid and cost caps vs highest volume. There is no best media buying setup, whatever the online preachers claim; every business is different, and the only way to find yours is testing new things regularly. (posted December 2025 · practice might be outdated)

Switching from ABO to CBO, or the other way around, will never be the lever that enables you to scale. The actual levers are outputting great creatives, congruent landing pages, and a good offer. Campaign-setting toggles don’t move the needle; those three do.

Inconsistency on Meta is normal, so base decisions on longer time frames, not daily swings. At low spend a few sales make a huge difference in daily metrics: one day performs, you scale, the next tanks, but the actual difference in conversions might be tiny and statistically meaningless. If 7-day ROAS hits target, daily variance shouldn’t matter. Stop making decisions off two or three bad days, or you’ll cut creatives that were actually working because you panicked at red numbers. The inconsistency isn’t a bug, it’s how the system works.

Your ad account structure is never the reason it’s not working. Stop obsessing over CBO vs ABO, campaign setup, or bidding strategies. The fundamentals decide everything: creative, landing page, positioning, offer. Fix those and you’ll scale.

One CBO per product per geo. 3 products in 3 geos means 9 CBOs. Each creative batch launches in its own adset, and each adset gets a minimum spend. (posted July 2026 · practice might be outdated)

Keep total min spends at max 50% of the CBO budget. If the CBO budget is $300 with 3 adsets at $50 min spend each and you want to add a fourth, raise the CBO budget to $400 and add the new adset with a $50 min spend. (posted March 2026 · practice might be outdated)

One CBO can handle both testing and scaling. Testing: one adset per new concept, each with a minimum spend. Scaling: increase the CBO budget 25 to 30% every day as long as the previous day was above target and the last 3 days combined were above target. (posted March 2026 · practice might be outdated)

There is no one best ad account structure; anyone claiming otherwise is a noob or selling something. This one works at 8-figure scale, even post-Andromeda. One CBO per product category per geo, highest volume, no caps, with testing and scaling in the same campaign (no separate testing campaign). One adset per new concept, each with a minimum spend to force delivery, and the total of all minimum spends stays under 50% of the CBO budget (at $1000/day CBO, minimum spends total $500 or less) so you can still see where Meta allocates freely. New video concepts launch with 3 to 5 hooks; standard statics get 1 headline with 3 to 5 visuals per adset. At the ad level: 2 primary texts, 2 headlines, sometimes 2 landing pages (the 3:2:2:2 method). Each adset runs a minimum of 4 days, usually 7 unless performance is dogshit. (posted October 2025 · practice might be outdated)

Run a low-budget “zombie” campaign with a cost cap and let it rip: it will spend efficiently on a few ads that never made your scaling campaign. I do not run this myself because I force spend to all my creatives, but multiple friends do it and it works for them. (posted August 2025 · practice might be outdated)

Run one CBO for both testing and scaling: dump new creatives in as one adset per concept with 3-5 ads each, set a minimum spend at adset level, and scale when campaign performance beats target. When the campaign runs under target, kill the high spenders; also kill mid and low spenders with bad performance. (posted August 2025 · practice might be outdated)

Split campaigns by product category or by margin, and send traffic to collection pages with the offer in a banner on top. Make sure the products shown in the ads sit at the top of the collection page; apps and UTM rules exist for this. Do not diversify beyond one channel until at least $100k/month: each new channel costs time and resources and pulls focus from what already works. (posted August 2025 · practice might be outdated)

Run everything broad and exclude all purchasers. A further test worth copying: exclude 1-day website visitors too, on the hypothesis that extra spend on them is not incremental, since those people are already as likely to convert as they are going to be and more spend won’t help. (posted September 2025 · practice might be outdated)

If 10-20% of acquisition spend lands on people who already converted, your new-customer CAC inflates; the CPM objection is a clown argument. Excluding a small past-purchaser audience barely moves CPMs, and if it does you have a different problem. Even when CPMs rise, NC CAC stays lower and NC ROAS stays higher with purchasers excluded, and those are the only KPIs that matter when running acquisition. (posted August 2025 · practice might be outdated)

Exclude past purchasers from acquisition campaigns using three layered audiences, because no single tracking method is complete. Method one: a website custom audience on the Purchase event with a 180-day window, pulled from pixel plus Conversions API. Method two: connect Klaviyo to Meta, build a segment of everyone with at least one order, and sync it as a custom audience; this catches iOS users and ad-block traffic the pixel missed. Method three: export all customers from Shopify as CSV, upload as a customer-list audience, and refresh it monthly.

Then exclude all three audiences at the adset level of every acquisition campaign. Pixel data misses people, Klaviyo lacks some emails, and manual uploads go stale, so only the three together stop you advertising a product to someone who already owns it. (posted August 2025 · practice might be outdated)

Make every decision on new-customer ROAS or new-customer CPA, and aim for a 3:1 LTV to CAC ratio. Early on you do not know your CLTV, so hold acquisition at profitable or at least break-even. As the brand matures and CLTV becomes clear, lower the target accordingly.

Multiple 7 figures a month bootstrapped came from knowing when NOT to scale. Ecom growth is not linear: in scaling season everything you launch works and budgets and revenue climb daily; in slow season every ad flops and efficiency tanks, and that cycle repeats forever. Scale hard when the market and algorithms allow it, and when it’s not scaling season bring spend down to protect margins so you can deploy more cash when scaling season returns, with each season raising your baseline. Don’t use slow season as cope: keep working the inputs (testing creatives, offers, launching products) to get back to scaling.

Kill on bad performance plus bad upper funnel metrics, never performance alone. Let a new ad run 3 to 4 days: if both performance and upper funnel KPIs like CPATC, hook rate, and hold rate are bad, kill it; otherwise give it 7 days and reassess. Top spenders get more rope: if frequency and CPMr are low and the campaign performs well, a bad in-platform ROAS or CAC on that adset is not a kill signal. (posted July 2026 · practice might be outdated)

A simplified Meta KPI setup is 13 columns. Amount spent, budget, NC ROAS, ROAS, NC CPP, CPP, NC purchases, purchases, ATC, CPATC, frequency, CPMr, and date created. (posted May 2026 · practice might be outdated)

Launch wide on avatars and concepts, then surf one CBO upward. The launch: 2 avatars with 4 angles each for 8 total angles, 20 concepts in 20 adsets mixing UGC, native ads, AI animation, and statics. Most traffic went to a pre-sell page, with some lower funnel statics going directly to the PDP. Budgets scaled $250 to $450 to $650 to $1,000 to $1,350 to $1,850 to $2,500 in 10 days inside 1 CBO, on an account that ran some ads pre-launch to warm it up and raise the spend limit. (posted March 2026 · practice might be outdated)

Surf scaling runs on an intraday tracking sheet, not the ads manager. The internal spreadsheet takes 4 inputs logged through the day: time, total revenue so far, total Meta spend so far, total Google spend so far. It calculates period performance plus MER automatically, so you see in real time whether the current scaling push is holding efficiency.

Stop spending when you’re far below target efficiency, even if LTV would eventually recoup it. With unit economics dialed in you know exactly how much you can spend without cash flow issues, so in weaker periods scale down: lower spend usually pushes efficiency back up, and running at low efficiency creates problems beyond the payback math.

Ad performance always sits in one of four CAC-by-volume quadrants, each with exactly one correct move. Low CAC plus high volume is the winner: scale aggressively, this is the sweet spot. Low CAC plus low volume is a missed opportunity: you’re being a pussy, scale up. High CAC plus high volume is fixable: you have demand, so reduce spend, fix your funnel, ads, or offer, then scale again. High CAC plus low volume means everything is broken: pause, figure out what’s wrong, restart small. Never allow yourself to sit in that last quadrant.

If you fulfill from China, run worldwide campaigns: they’re still crushing and there’s no downside. One hypothesis on why they scale: advertising across timezones lets Meta spend during each country’s optimal hours instead of peaking at the same times daily in one market. The pattern in practice: Australia and New Zealand perform best in the morning, Japan and South Korea a few hours later, Saudi Arabia and UAE around lunch, then US, Canada, Mexico, and LATAM in the evening. It’s like running ads during peak hours 24/7, more opportunities to spend at optimal times. (posted December 2025 · practice might be outdated)

Once you cross $500K/month on Meta with a working product and funnel, start testing new ad channels. You already have winning creative and proven messaging, so adapting them costs fewer marginal resources than starting from scratch, and you diversify channel risk: if one underperforms, others pick up the slack. The interesting dynamic is that smaller channels are often more efficient but less scalable, so cheaper acquisition there subsidizes a higher target CAC on Meta, letting you scale harder where the volume actually lives. (posted October 2025 · practice might be outdated)

Add email and Google first, around $50k/month; amplify a proven funnel with native later. Taboola only got added after $1M/month, but knowing the impact it made, it should have started way earlier. Native is not for every product: it depends heavily on your product and audience, and is especially recommended when the target audience is 45+. (posted September 2025 · practice might be outdated)

Three underused sales moments: Singles Day (11/11), huge in Asia and growing hard in the West; post-Christmas (Dec 26-31), when people spend gift money and buy what they did not get; and Q5 (Jan 2-15), when new-year-resolution buyers are still hot but CPMs are back to normal. (posted September 2025 · practice might be outdated)

Build your own BM and ad accounts instead of renting: at $1M/month spend, a 2-3% account fee is real money. You also never build a relationship with Meta on rented accounts. It is a hassle up front and worth it in the long run. (posted August 2025 · practice might be outdated)

One click tells Meta a user has category intent, and then it serves them every solution in that category. Clicking a single hair growth serum ad flooded the feed with the same brand’s down-funnel ads (the algo is very good at sequencing), then derma rollers and shampoos, then telehealth brands selling finasteride, then local hair transplant clinics: you compete with your whole category, never only your direct competitors. The brand that wins has an ad live for every part of the funnel; run only TOF and others scoop up the awareness you paid to create. MOF: statics that remove objections, UGC creators showing real people getting results, comparisons against the other solutions in your category, social proof and reviews. BOF: urgency and FOMO, a reason to buy NOW, because if that feeling comes from another brand you lose the conversion even though you created the initial intent. (posted August 2026 · practice might be outdated)

Match the format to the funnel stage. TOF: VSLs, AI animation, podcast clips, long form native statics, and yapper ads. MOF: native UGC and creator videos, unboxings, GRWMs, testimonials, content natively integrated into what creators normally post. BOF: offer and FOMO statics. (posted July 2026 · practice might be outdated)

Avatar is WHO you sell to, angle is HOW you sell to them. For a collagen powder: avatar 1 is women in their 30s concerned about early aging signs, avatar 2 is women 45+ dealing with visible skin aging, different people at different life stages with different priorities. The same outcome gets framed per avatar: “prevent fine lines before they start” for avatar 1, “reduce fine lines and look like you’re 30 again” for avatar 2.

Running ads to multiple awareness stages simultaneously is how you unlock efficient scale. Your market isn’t one homogeneous group; it’s small pockets at different stages, and most marketers only run bottom-of-funnel ads for people who already know the product, which is a tiny slice of the addressable market. Unaware and problem aware get long form video (VSLs, TikTok-native educational content) sent to advertorials, long form sales pages, or quiz funnels. Solution aware gets third party comparison ads (us vs them) sent to comparison funnels like “We reviewed the top 5 in this category, this was the clear winner” or “The best in the category, tested and ranked by an authority figure.” Product aware gets testimonial and social proof ads landing on pages heavy with reviews, before and afters, and trust signals. Most aware gets offer and FOMO ads straight to the product page.

The key is running all of these at the same time. If you only run problem-aware ads you educate the audience while competitors running higher-awareness ads steal the conversions you warmed up, which is exactly what Grüns and IM8 are doing to AG1. If you only run most-aware ads you leave 95% of the potential market on the table.

Messaging is the specific words, phrases, and language you use to communicate your angle. Huel example: the angle is convenience for busy professionals. Messaging option one: get complete nutrition in under 2 minutes. Messaging option two: save 10 hours of meal prep every week. Same product, same angle, two different ways to communicate it.

Creative diversity separates brands that plateau at 6 figures from those scaling to 8 and 9, because your market is dozens of segments, not one homogeneous group. Map avatars first. For Loop Earplugs: the rave girl protecting her hearing without missing the music, the light sleeper with a snoring partner, the open-office worker trying to focus, the parent managing loud kids, the frequent flyer dealing with airplane noise. Each needs earplugs for completely different reasons, so each gets its own angle: sound quality for the rave girl (“Hear every note clearly while protecting your ears”), peace for the sleeper (“Finally sleep through your partner’s snoring”), focus for the office worker (“Block distractions while staying aware of important conversations”).

Then layer awareness stages per avatar: the unaware rave girl gets “the permanent hearing damage one festival can cause”, problem aware gets “why your ears are still ringing 3 days after the show”, the product-aware light sleeper gets “Loop Dream review: do they actually work for side sleepers with sensitive ears?”. Finally match creative formats to how each avatar consumes content: a listicle for the office worker (“5 ways Loop Engage helped me focus in our open office”), us vs foam earplugs for the rave girl, POV for the light sleeper (“POV: Your partner starts snoring at 2 AM but you have Loop”), myth vs fact, day in the life for the parent, outfit of the day styling Loops with festival fits, airport street interviews for the flyer. Multiplied together (avatars, sub-segments, angles, awareness stages, formats) that is thousands of creative combinations capturing segments competitors completely miss.

One angle, many avatars: Grüns sells greens gummies, and the protein-absorption angle alone targets gym bros building muscle, women focused on glute development, and boomers maintaining muscle mass. Each avatar shares the core problem but wants a different outcome, so the messaging changes while the solution stays identical: “Stop wasting your protein powder” for the gym bro, “Finally grow the booty you always wanted” for the woman, “Stay strong and independent longer” for the boomer. Same gummies, same protein benefit, three completely different value propositions. Once you understand this, you never run out of ads to test.

Your market is not one big group: it is dozens of smaller pockets that each need different content and messaging to convert, and diversification across five layers is what separates brands that scale from brands that plateau. Diversify angles by hitting every psychological trigger: pain point (“Tired of acne ruining your confidence?”), aspirational (“Get the glowing skin you see on Instagram”), social proof (“Why 50K+ women switched to this routine”), convenience, urgency, and authority (“Dermatologist-approved formula”). Diversify formats: VSL, listicle, podcast style, authority endorsements, street interviews, us vs them, POV, testimonials, myth vs fact, customer reviews, advertorial style, promos, product demos, unboxings.

Diversify creators, because people buy from people who look like them: the 45-year-old suburban mom and the 22-year-old college student need different messengers, so span age groups, ethnicities, body types, and lifestyles. Diversify pages: your brand page, authority pages (dermatologist, nutritionist, fitness trainer, vet), magazine-style pages, and creator partnership ads. Diversify landing pages: homepage, collection, product page, advertorials, listicles, VSL pages, quiz funnels. When one audience pocket saturates, twenty others are still printing; that is how brands break plateaus and reach 8 and 9 figures.

Facebook statics take up the whole screen when comments are expanded, so you can almost build a full landing page as an ad. Headline, hero, social proof, and a CTA can all live in one static. It also raises the stakes on getting good comments: they’re among the best forms of social proof because people can click the profiles and see real FB users behind them. (posted August 2026 · practice might be outdated)

The current top spending ad is an authority led VSL, 2:14 long, no fake AI. The block sequence: fear hook, open loop with a big promise, authority intro, failed alternative 1, failed alternative 2, failed alternative 3, hidden root cause, problem mechanism, cost of inaction, product intro, solution mechanism, short term transformation, long term transformation, social proof, price anchoring, risk reversal, scarcity, call to action.

Ugly ads win because the time went into research, not design. The person who made them spent their hours learning how customers think, the lingo they use for their problems and symptoms, what they already tried that failed, what makes them skeptical, and how they describe the dream outcome, then wrote copy that resonates. Good copy, an easy to read font (bigger than you think) on a contrasting background, and 1 accent color is enough to build creatives and pages that convert; design should be purely optimized for ease of digesting the copy and leading to the next click. Until you do at least $10M a year, don’t focus on design.

You won’t unlock new scale with your ads until you become better at storytelling. People scroll with their guard up and skip anything that looks like a normal ad (a view you already paid for), but a story told in a platform-native format bypasses that guard and lets the viewer reach the conclusion your product is the answer themselves instead of resisting your claim. Good stories trigger dopamine (attention and memory) and oxytocin (trust), and because they hold attention the algorithm pushes them into colder audiences and rewards you with lower CPMs, while a hard-selling ad only reaches the small pool already in-market. Those cold viewers then enter your funnel already believing in your solution.

The formats working best right now: AI animation ads (once messaging is dialed in, go ham on variations), yapper ads, and native statics with long form copy. The practice: walk with no phone, music, or podcast, think up story ideas around your current best angle, write them down, then come home and write a storytelling ad and run it, repeatedly, until you get good. (posted July 2026 · practice might be outdated)

Viral formats work for paid ads because the audience is pre-conditioned: format equals entertaining and worth watching. When a format goes viral, the best editors and creators pile in and flood feeds with great content in it, so when your ad uses the same format people hook better and watch longer, not because your ad is better but because they’ve been conditioned to expect a payoff. You’re borrowing the success of the format itself, though you still need direct response principles on top. Formats die once too many brands flood them and the conditioning breaks, so when you see one going viral, test it immediately.

Speed up your video ads by 15 to 30%. Tested for 2 months on both UGC style ads and AI voiceover plus b-roll ads: the sped-up versions outperformed normal speed in the majority of cases. (posted November 2025 · practice might be outdated)

Creative isn’t just the video or static: it’s the video or image plus primary text plus headline plus landing page plus identity page. To feed Andromeda the diversity it wants, run different videos, images, primary texts, headlines, landing pages, and identity pages (brand and third party). But don’t mix these randomly; the combination has to be congruent as a whole. (posted October 2025 · practice might be outdated)

A direct response ad runs thirteen beats in a fixed order. Contrarian hook, call out skepticism, problem intro, agitate pain point, risk reversal, moment of doubt, desired outcome, product demo, unique mechanism, justify value, future pace benefits, offer stack, CTA.

Never think about copy versus creative: think about the ad as a whole. Creative, headline, and primary text work together. The headline has to make sense with the creative, and the primary text completes the picture.

The net new versus iteration ratio should never be static. Launch as much volume as you can without diminishing quality, favoring fewer high-conviction ads (conviction, not production quality). You test net new to figure out what works; when something works, you iterate the shit out of it until it stops working. Some periods run 20% iteration and 80% net new, others 50/50, depending on where you are in the cycle.

The best creative strategists obsess over every detail of the final product; the bad ones think their job ends after the script. Bad ones do surface-level Reddit research, deliver a Google Doc with a script and maybe an Atria reference link, hand it to the editor with minimal context, launch whatever comes back, and blame bad performance on the edit. Good ones watch hours of the content their customer consumes, then obsess over the hook (not just what’s said but how, the framing, the cuts in the first 3 seconds, the exact moment text appears), text placement and readability, pacing and cut count, and whether the b-roll reinforces the message or fills space. Their editor feedback is specific: “cut this section by 2 seconds, show the hook text 2 seconds later, add this exact testimonial here,” not “make it better.” They own performance: when an ad flops they diagnose it and apply the learnings to the next batch instead of pointing at the editor or media buyer. Hire the person who watches the ad 20 times before launch and takes ownership of results, not just the script.

When a native static wins, split it to find whether the image or the copy carries the weight. Make 10 new ads with the winning image plus 10 different copies, and 10 new ads with the winning copy plus 10 different images, each batch in its own ad set. Most of the time it’s the copy, especially long form, so the next move is testing a bunch of different native images against that same copy. When those iterations produce a new winner, run the same process again.

Only studying your winning ads is survivorship bias, the same mistake that almost got Allied pilots killed in WWII. The military mapped bullet holes on returning planes and planned to armor those spots until statistician Wald pointed out the returning planes showed where a plane can take damage and survive; the fatal zones were the engines and cockpit, the areas with no holes, because planes hit there never made it back. Same with ads: run 50, find 3 winners, and you study the 3 while ignoring the lessons in the 47 that died in testing. Go through the losers and diagnose why: weak hooks, wrong awareness stage, unclear value prop, format that doesn’t fit the audience, bad execution. Knowing what not to do eliminates bad bets before they eat budget.

Build your swipe file from brands that are actively printing, not from theory. The list worth studying: Norse Organics, Gruns, Kind Patches, Happy Mammoth, Primal Queen, Loop, Hike Footwear, Spartan, Hollow Socks, Hi-Smile, Blissy, Feals, Grounding Well, Armra, Petlab Co., Heights, Primal Herbs, Everyday Dose, Ryze, City Beauty, and Ka’Chava. (posted September 2025 · practice might be outdated)

Test concepts, not creatives, and stop caring which variation eats the spend. The setup: a CBO with min spend (same holds for ABO), each adset holding 3 to 5 intentionally similar variations of one core concept, videos with the same body and different hooks, statics with the same headline and different visuals. The hypothesis is “will this concept work”, not “will this specific creative work”, so one creative absorbing all the budget still answers the question. If the concept works, you’re onto something. If it doesn’t, move to the next concept. Only after finding a winning concept do you isolate variables to extract more learnings.

Run a weekly creative review where the team dissects every ad, winners first: for each one, break down angle, hook, promise, avatar, awareness level, pacing, desire, and format, then do the same for losers, which matters more because you have more of them. The goal is understanding why something worked, not celebrating: you are literally paying Meta to test hypotheses and extract learnings. Iterate the winners systematically: for video, same video with new hooks, same script with different creators, winning hook on an old winning video, same script with new B-roll; for statics, same headline with new visuals or same visual with new headlines.

Log every test in a spreadsheet: status, batch number, angle, hook, core desire, persona, format, testing hypothesis, and a four-way result grade (high spend + good CPA, poor spend + good CPA, poor spend + poor CPA, good spend + poor CPA). Do this every week for a year and you understand your audience so deeply you never run out of winning concepts. That is the difference between a 6-figure and a 7-figure-per-month brand.

Test every ad for 7 days, and make every ad with intent. Randomly spamming creatives guarantees you burn money on testing. Put real thought behind every ad you launch and the testing phase itself can run profitable.

Two iterations of a winning long form native consistently perform. Turn the winning long form into a scrolling text video with voiceover, using the winning image as the background. Or turn it into an AI animation story, which takes effort to do well but scales. (posted July 2026 · practice might be outdated)

Milk winners with iterations that isolate one variable while keeping everything else constant; that’s how you find what’s actually driving performance. For video ads: new hooks (as small as a text overlay change or as big as a fully new spoken and visual hook), new leads, the exact same script rerecorded with different creators, swapped b-roll (especially on AI voiceover plus b-roll ads), sped up or slowed down versions, shorter cuts down to the core message or longer cuts with more proof and mechanism explanation, and different testimonials. For statics: mostly swap headlines and visuals, winning visual with new headlines or winning headline with new visuals, and on “native” statics with no headline, change the person or main object with Gemini or Higgsfield, turning a winning young black woman into an old black woman, young asian woman, white woman, and so on. (posted January 2026 · practice might be outdated)

Consistently launching high-quality new ads is the only long-term answer to creative fatigue. Short term, you can prolong a winner’s lifespan by pairing it with a congruent landing page.

One proven concept can fuel months of profitable campaigns if you iterate systematically instead of chasing the next shiny creative idea. For video: new overlay text hooks and new visual hooks in the first 3 seconds, the same script sent to a new creator, new b-roll under the same voiceover, added testimonials or social proof, a greenscreen version with a creator reacting to the video, a changed location (car, living room, dog park), adjusted length, the same message in a different format (podcast, UGC, AI voiceover plus b-roll), and winning hooks turned into static ads. For statics: keep the visual and swap headlines, keep the headline and swap visuals, turn the image into a GIF, add motion, add urgency or social proof elements, and turn winning headlines into video hooks and scripts.

Never turn off your top spender: your whole campaign performance will probably die with it. Build an iteration of it instead. Read the ad’s comments, see what people are actually saying, and use that to improve the next version.

Everyone chases doctors, dermatologists, and vets, but licensed professionals are not the only way to add authority. A good authority is simply someone your customer already trusts on that specific problem, and they’re far easier to sign and cheaper to work with. Beauty: makeup artists, hairstylists, barbers, Sephora employees. Health: ex pro athletes (way cheaper than active pros), coaches, military. Food: chefs, bakers, butchers, bartenders. Home: plumbers, electricians, carpenters, interior designers. Tech: YouTubers, photographers, DJs, gamers. Use them in your creative and funnels and watch performance go up.

Only film your own ads if you look like your audience. A 23 year old founder selling a lifestyle change product to boomer women has no business on camera: a good creator who actually looks like the audience will almost always outperform the founder, and that easily makes up for the extra cost of not filming yourself. Always match the creator to the audience you’re selling to.

Run 40 to 50% of all ad spend through third party pages: any page that is not your brand page. That means creator pages, owned third party pages like magazines and blogs, and authority style pages. Taking this seriously increased top-of-funnel performance significantly.

CPMs are usually lower on third party pages, but not because they’re non-brand pages: the ads look more organic, which drives higher engagement, which lowers CPMs. The variable that matters is engagement, not brand page versus third party page; you can hit the same lower CPMs on a brand page if the ad is engaging enough. Third party pages just make the organic feel easier to achieve.

50% of ad spend runs through third-party pages because it looks less like an ad. Four page types that work: authority pages (dermatologists, doctors, athletes, trainers, nutritionists), creator pages (whitelisting partnerships, or “real person” pages that look like your target avatar), niche community pages (“Moms of Toddlers”, “Plant-Based Living”, “Home Gym Warriors”), and magazine or blog pages (“The Health Guide”, “Beauty Review”, “Wellness Insider”). Running everything from your brand page alone leaves money on the table. (posted October 2025 · practice might be outdated)

Full AI animation ads are top spenders in multiple campaigns right now, beating winners that ran for months, highly produced expensive videos, and long multi-creator VSLs. It’s less defensible since competitors can replicate AI ads far more easily than real creator content, but the data doesn’t lie. The real defensibility is volume: making these concepts actually look good takes time, and most brands won’t put in the effort to produce them at scale. If you haven’t tried the format, test it; @frankyecom’s method leveled up the output quality significantly. (posted March 2026 · practice might be outdated)

Obvious “AI slop” in trending organic formats is now outperforming most ads in the account. It beats real-looking AI UGC on performance and is easier to make. (posted February 2026 · practice might be outdated)

Our top spender for over two months on one product is a static made with ChatGPT in 2-3 minutes, simple cartoon style. The creation took minutes; coming up with the concept and writing the perfect prompt took a few hours. The thinking is the work, not the rendering. (posted September 2025 · practice might be outdated)

AI collapsed the cost of creative, so raw volume no longer saves you: when a bad video ad cost $3-5k, people put thought into each one, and now everyone can generate 50 a day. The feed is flooded with AI slop, consumers are going numb to it, and it mirrors what cheap stock photos did to websites: everything looked the same and the brands investing in custom content stood out. Volume still matters for out-testing competitors, but the winning move is using AI to raise output while holding or raising your creative standard. When everyone can make 50 ads a day, be the person making 5 great ones.

VEO3 ads are ripping right now, but only when you exaggerate that they are AI instead of hiding it. Podcast-style VEO3 ads are crushing, and visual hooks made with VEO3 are printing. (posted August 2025 · practice might be outdated)

A PDP is a lander, but a lander is not a PDP. Same way a cow is an animal but an animal is not a cow: the product page is only one kind of landing page. Send traffic first to a pre-sell page, an advertorial or a listicle for example, and from there on to the PDP.

Adding a 1-question quiz before an advertorial, then dynamically swapping the headline and hero image based on the answer, improved performance. Example: ask “How old are you?”; a 25 to 35 answer loads “Why women in their 30s are seeing results with this supplement” with a hero image of a woman in her 30s, while 45 to 55 loads the over-45 headline with a woman in her 50s. The rest of the page stays identical, only the entry point is tailored. Two hypotheses for why it works: micro-commitment (answering one question invests them in what comes next) and personalization (the advertorial speaks directly to their answer).

The “remove all friction” advice is wrong: adding useful friction made performance improve every time. VSLs got extended from 60 to 90 seconds to 3 to 10 minutes, static ads got 2000 to 4000 word primary texts instead of 3 sentences, and funnels got more steps before the offer page. It works because showing an offer page to someone not ready to buy makes no sense; each piece of friction educates, qualifies, or builds trust, so by the time they see the offer they actually want it, and retention likely improves too because buyers know what to expect and why consistent use delivers the outcome. The friction has to be useful, guiding people toward understanding why they need the product, not random annoying steps. A $19 impulse buy doesn’t need a complex funnel and a 10 minute video, but for any product requiring trust, education, or commitment, useful friction makes you more money.

The high-performing supplement funnel runs static ad to advertorial to quiz to offer page to checkout. Three brands running this exact structure live and worth reverse-engineering step by step: ColonBroom, Bioma Health, and HerBodhi. Study each stage in sequence, not just the offer page.

Build the entire journey as one narrative, not just matching pieces. For a skincare brand: get a dermatologist (or an AI dermatologist creator) in your ad, run it from a Facebook page carrying that dermatologist’s name and photo, write the primary text and headline in that same dermatologist’s voice, send traffic to an advertorial on the same angle written like it lives on the dermatologist’s blog, then land on a product page that feels designed for that exact problem. Every touchpoint reinforces the same narrative, so the customer never feels sold to; they feel like they are following a natural path to the perfect solution for their situation. Most brands optimize individual elements. The real leverage is optimizing the entire journey as one connected experience.

A landing page is any page someone lands on after clicking an ad: product page, home page, collection page, advertorial, listicle, long-form pre-sell page, or VSL page. When operators talk about landing pages as a lever, they mostly mean the last four: advertorials, listicles, pre-sell pages, and VSL pages.

The current winning landing page runs 834 words. The hero section carries a headline in the shape “do you have [symptom]? Here’s how to fix it in just [amount of time] and finally [desired outcome], no [failed alternative] needed”, a before/after hero image, 3 checkmarks removing objections, and a soft CTA that scrolls down to the next section.

Below the fold the sequence is: agitate the problem, cost of inaction, a symptom checklist so visitors diagnose themselves, false belief kill, unique problem mechanism, product intro, unique solution mechanism, how it works in 3 steps, time to result (what to expect by day X), us vs them, a UGC wall for authority and peers, reviews, a 3 step qualification quiz, scarcity, and a CTA to the offer page.

In saturated markets, dedicated us-vs-them comparison landing pages are a proven scaling lever. Emma Sleep built a $1 billion per year brand doing this and runs 26 localized versions of its comparison page. The format’s whole job is positioning you against the alternatives a solution-aware buyer is already weighing, in their own language and market.

Name the product with the keywords buyers actually search, not an invented brand name. “Anti-Aging Vitamin C Serum for Dark Spots & Wrinkles” beats “SuperGlow Pro”: benefits plus what the product actually is plus primary use case. Then run two distinct short descriptions on the page. Under the title and reviews goes a factual 1 to 2 line summary of what the product does: “Clinically proven vitamin C serum that fades dark spots and reduces fine lines in 4 weeks.” In the outcome section next to the lifestyle image goes the aspirational, outcome-focused line: “Wake up to radiant, youthful skin that makes you feel confident without makeup.”

A high-converting product page runs thirteen sections in a fixed order. The announcement bar carries the offer (“Back to Work Sale: Buy 1, Get 1 FREE + Free Shipping”). The media gallery works as a horizontal funnel: HQ product image, product in use with benefits overlay, before and after transformation, what’s in the package or ingredients, how it works in 3 to 4 steps, us vs them comparison chart, review screenshot, then lifestyle shots if suitable. Next comes a clear keyword-rich product title with star rating and count (4.8/5 based on 123 reviews), a 1 to 2 line description, and 3 to 6 USP bullet points. The purchase block stacks bundle pricing with a strike-through price around 30% higher, a bold add-to-cart button, guarantee plus shipping estimate, micro-trust signals (“Ships in 24 hours”, secure payment icons), and an optional free gift unlock at 2+ units.

Below the fold: UGC-style product and testimonial videos, an FAQ handling ingredients, usage, sizing, shipping and “will this work for me”, press logos, a desired-outcome section (large lifestyle image, outcome-driven headline, 2 to 3 lines of benefit text), testimonials, a how-it-works timeline showing realistic results at week 1, week 4 and week 8, another us vs them, an “X-Day Risk-Free Guarantee” callout, and the review section.

In super aware markets, third-party badges break the claim tie. Every competitor claims the best quality, reviews, and formula, so customers can’t tell who is telling the truth; an award or certification backs up your claims with borrowed trust. Most are not hard to obtain: some just require X amount of votes (easy with a big email list), others you can simply buy. Examples: beauty has Dermatest, Vogue Beauty Awards, Allure Best of Beauty; health has NSF Certified and Men’s Health Awards; tech has Red Dot and the CES Innovation Award, plus local consumer awards like “Voted Product of the Year” running in 40 countries. Use them in ads, LPs, and PDPs and watch performance improve.

Stop only showing reviews from your own site. Add screenshots of your Amazon reviews, Facebook comments, and Reddit threads that mention you to product pages, pre-sell pages, and ads. Consumers are getting smarter and know on-site reviews can be manipulated; they trust third-party reviews (or ones that look third party) far more.

Consumers know polished testimonials can be fabricated, which is exactly why raw reviews work: they feel real because they are, or look like they are. Screenshot your ad comments and Trustpilot reviews and place them throughout your pre-sell and product pages. Doing this lifted conversion rates significantly.

Revenue per session tools can’t see your cost of traffic, so run LP tests directly in Meta. Targeting and CPMs are determined by the post-click experience as much as by the creative, so on tools like Intelligems the CPM gets averaged and split across both pages while one page might actually drive cheaper traffic: a page can look worse on revenue per session yet win on ROAS, and these tools can raise your overall CPM during tests on top of that.

The Meta method: grab your winning ads, spin up a separate ABO testing campaign, one adset sending those ads to page A (the control duplicated to a fresh URL so no outside traffic pollutes it), another adset sending the same ads to page B, equal daily budgets, extra variants as extra adsets. Ignore in-platform attribution: compare revenue generated per page against spend for a ROAS per page, then run significance on abtestguide.com. The lazy version for variant pages built off winners: duplicate a winning adset in a live campaign, swap in a new hero section, and see if it picks up spend, less accurate but it drives spend efficiently. (posted July 2026 · practice might be outdated)

Test pre-sell pages (advertorials, listicles, long form sales pages) inside the CBO by duping ads and swapping the URL; test new PDPs with an Intelligems A/B test. To start, build landing pages based on your top performing ads so the page matches the angle of the ad sending the traffic. (posted February 2026 · practice might be outdated)

Testing lots of hero section variations (headline plus sub plus image) lifted revenue per session 30% on its own. The other win: delaying the product intro section on top-of-funnel landing pages.

Run landing page A/B tests inside Meta instead of a dedicated tool: same adset, duplicate the ad, point each copy at a different landing page. After running LP tests on Intelligems in the past, this switch works very well and is way easier to test. (posted October 2025 · practice might be outdated)

Optimizing RPS in isolation is fundamentally flawed. Heavy CRO work and offer testing raised AOV and RPS significantly, and then CAC rose in the same period, frequency climbed because fewer people can buy at higher price points, and Meta appeared to start optimizing for bigger spenders, shrinking the addressable audience and making scale harder. Most CRO agencies won’t tell you this: CAC rises alongside RPS and the market shrinks, so judge the whole system, never one metric. (posted April 2026 · practice might be outdated)

The easiest way to improve your ads, product pages, and landing pages is asking people who just bought, while the decision is still fresh. Ask: what almost stopped you from buying today, what questions or doubts did you have on the site, what was the main reason you decided to buy, what problem are you hoping this solves, who are you buying for, and how did you compare us to other options. Make them open-ended, never multiple choice: the gold is in how customers describe things in their own words, which becomes the copy for your next winning ad. Tools to set it up: KNO Commerce, ZigPoll, Triple Whale Post Purchase. Boost response rates with store credit or a chance to win a full refund, and block weekly time to read responses and actually implement what you learn. (posted December 2025 · practice might be outdated)

Store-wide CVR says nothing: it mixes warm traffic, cold traffic, and repeat purchasers, so it skews wildly. CVR also trades off against AOV; push AOV up and CVR mostly falls. Judge revenue per session instead, measured at the page level or on specific funnels.

Test increasing your 1 piece base price while keeping bundle prices the same. Bundles look like a better deal as the discount versus 1 piece widens, more people pick a bundle because a single unit gets less attractive, and AOV goes up so you capture more LTV on day 1. It also attracts better customers: retention data shows people who bought more on day 1 retain better, because repeat purchase products take time to show results or become a habit, and someone who buys 1 piece and runs out before seeing results rarely comes back. There are downsides, but this move improved unit economics by a lot.

Urgency and scarcity convert because loss aversion is hardwired, but only when they’re believable and honored. Countdown timers that reset on every visit, stock warnings stuck at “3 left”, and sales that “end tonight” at the same price tomorrow get seen through, and once customers catch you lying they never trust you again. Believable evergreen versions: seasonal offer rotation (same core offer reframed as New Years Offer, Valentine’s Special, Easter Promo, Summer Sale, each with a deadline that actually ends), production-based (“limited batch, next batch ships in X weeks”), inventory-based (real stock shown for the exact variant picked, “only 3 left in size M”), ingredient constraints (“this ingredient restricts production to X units per month”), and social proof urgency (“bestseller, already sold out 4x this year”). Honor every claim consistently and customers learn to trust it; trust is what drives action.

Stack six campaign types across BFCM so people have multiple reasons to buy throughout the whole period. The announcement (“tomorrow our biggest sale of the year starts”) builds anticipation so cards are ready. The flash sale runs a 6 to 8 hour window with an extra discount or bonus, urgency inside the sale. The refund giveaway makes every Xth order free because people love to gamble. Scarcity emails (“XYZ is already sold out” or “almost sold out”) and urgency emails (“sale ends in X hours”) keep the clock ticking. The extension (“due to overwhelming demand we’re extending through Cyber Monday”) captures everyone who missed the first wave. This is how you capture all the demand you’ve built over the last 11 months.

Stick to one already proven offer during Black Friday. It is not the period to experiment: you had 11 months to try new offers.

Send at least 3 emails per week through all of November and December, then ramp hard on BFCM weekend. Friday gets 2 to 3 emails plus 1 SMS, Saturday 1 to 2 emails, Sunday another 2 to 3 emails plus 1 SMS. Throughout the period, run 24 to 48 hour flash sales to push engaged subscribers who haven’t bought over the edge, and close the season with a refund giveaway: most people are broke by then, so a chance to win their money back combined with a good offer drives extra conversions.

Test offers with 3-day weekend promos, Friday to Sunday, in the months before Q4. Three days because most people buy on day one (excitement) and the last day (FOMO), so you capture peak buying behavior without discounting too long. Rotate a different offer each weekend: Buy X Get Y Free, X% off everything, X% off plus free shipping, spend X get a free gift, tiered spend-more-save-more. Track total revenue, conversion rate, AOV, profit margins, and CAC for each. The goal is not maximizing revenue, it is collecting data: by the time BFCM hits, you already know exactly which offer performs best.

There is no ideal email count, only the right emails at the right moments. The essentials for a post-purchase flow: social proof (you made the right decision), why we started (a personal founder message), a check-in, FAQ answers, a review request, and a winback after X days. Focus on the doubts people have, motivate consistent product use, and manage expectations so customers return. Place each email where the customer journey says people typically have questions, see results, or need a reminder.

Pick a product with an edge, fat margins, and an accessible price. Look for something you already know a lot about or are passionate about, since that gives you an edge. Demand at least 70% margin on CM1, and stay under a $100 selling price when starting out.

Sell the product before you own it: spend ~$500 on professional product renders, build a simple funnel, run $2-5k in ads, then refund. The funnel is either straight to a product page or a pre-sell page into a product page with Shopify checkout. Test static ads built from the renders plus VSLs cut from stock footage and ripped content. Email every buyer an apology with two options: immediate refund or first in line at launch. Demand validated, and you never ordered 1,000 units of something nobody wants.

A brilliant ecom product solves a real problem, is consumable or expandable, has a big TAM, ships small and cheap, and carries 70%+ margins. Consumable or expandable means customers reorder regularly or you can build a product line around it. Anything under 70% margin makes scaling ads profitably very hard, and heavy or bulky products let shipping eat what margin remains.

Scaling past $500k a month in 4 months was inputs, not infrastructure. What got done: increased creative output, onboarded new creators every week, hired an ad creative agency and an email agency, read reviews, ad comments, and customer service tickets manually to actually understand the customer, made ads that drive real net new TOF visitors, tested prices and offers (prices AND shipping costs went up), and negotiated better manufacturer and 3PL pricing off the increased volume.

What did not happen: no new marketing channels, no ad account structure debates, no custom dashboards, no new product launches (too early), no raised money or debt.

Copying big brands means copying a strategy built on advantages you don’t have. Almost all of them are funded with completely different unit economics, so they can sustain far higher CACs without running out of cash, and years of running plus celebrity endorsements bought them trust and brand awareness you lack. The best brands to take inspiration from are bootstrapped and scaling quickly.

Early retention orders are the license to scale at break-even. One month into a new brand, the first repeat orders arrived before post-purchase flows were even set up, and that conviction justified lowering the target NC ROAS and scaling at break-even, while waiting 3 months for real LTV data before dropping efficiency targets further. Offer testing is one of the biggest levers: a new offer already outperformed the original. Next moves: 10x the angles, formats, and LPs already working, add the missing middle and bottom of funnel ads, set up proper Klaviyo flows, launch Google Ads.

After a first 8-figure year, the 4x plan is five levers, and creative is the highest leverage one. Creative quality and volume come first, with output raised only if quality holds, since more mediocre ads are pointless. Retention gets a dedicated hire to own it completely, because better retention raises the affordable target CAC and frees cash to reinvest. Market expansion replicates the proven model in 5 new markets with fully localized ads, funnels, and customer experience. B2B expands into local retail across every active country because it adds profit margin on top of DTC. Product work focuses on improving hero products rather than launching new ones, betting that better products lift retention more than another launch cycle.

Going local halfway caps your reach. English-only ads reach only a small part of any non-native-English country and perform worse, so launch ads in the local language. The same holds for the rest of the funnel and store: local currency, local payment methods, local delivery.

You don’t need localization from day one: scale into multiple countries with unlocalized ads and funnels first, then localize the winners. There’s eventually a plateau where localization becomes necessary, and full localization is the only way to truly dominate a market (this brand is fully localized in 8 countries). But the path there is campaigns targeting multiple countries at once to test where actual market pull exists, then localizing only where it shows.

The sequence for cracking a new market: buy data cheap, find the angle, rebuild everything around it, then restore your price. A month after launching a new market, CAC was high, conversion was shit, and money was burning. First move: change the offer and drop pricing, since a price drop usually decreases CAC (at the cost of margin) and the extra volume produces the data you cannot optimize without. Winning ads from the main market did not transfer, so it took fresh research and 25+ new concepts tested for this specific market before one showed potential.

Once traction appeared, the entire funnel got rebuilt around that angle: new landing pages, rebuilt PDP, changed product positioning, everything aligned to reinforce one message, plus new ad formats iterating on the winning angle. Finally, prices went back to the original level because margins were not sustainable lower, and sales held up since the messaging was dialed in. Result: first-order profitable and able to scale, with retention analysis next to see if target CAC can rise over time to scale harder.

Once ad spend is high, an Amazon listing captures demand your ads already paid for but your Shopify store can’t convert. The halo effect: someone sees your Meta ad, clicks, doesn’t trust an unknown site, leaves, remembers the brand, searches it on Google later, sees the Amazon listing, trusts Amazon, buys. Without the listing that sale never happens. A massive group only shops on Amazon for the trust, fast shipping, and easy returns, and no funnel will change that; Amazon reviews also carry more weight than on-site reviews and can be leveraged back on your store and in ads. You don’t need to go all-in, just an optimized listing. Easy win.

Some buyers are channel dependent: they only purchase on Amazon or Walmart, and your Shopify store will never convert them. In the UK we see a clear correlation between our ad spend and our Amazon revenue, and a high chance those sales would not exist without the Amazon setup. Running Shopify alone means leaving that demand on the table.

Hire an email agency or freelancer, but pay a flat monthly retainer, never a percentage of attributed revenue. “We only get paid when you make money” sounds good and is a terrible deal: their fee grows just because you spend more on ads and your list grows, not because they do better work, and attributed email revenue is very easy to game.

Recruit editors the way you acquire customers: run Facebook ads targeting LATAM and send applicants to a long Typeform that qualifies on portfolio, turnaround time, and rates. The long form filters out the unserious. Backup channels: Facebook communities for video editors, and Upwork.

Split every hire into flywheels and fuel, and sequence them deliberately. Flywheels are great at strategy AND execution: they ship ideas end to end and build systems that keep generating value long-term (growth leads, creative strategists, operations managers, retention marketers); never hire someone who can only strategize. Fuel hires are great executors whose output stops when they stop working (customer service agents, video editors, designers, store managers). Sequence: hire fuel first while you are the flywheel (editors, designers, customer service, a store manager), then add your first flywheel hire, then more fuel to support them, since more editors means more output from your creative strategist.

Source flywheels through LinkedIn Recruiter with a $10-30/day boost (more budget won’t work better) and by poaching people in similar roles at other brands and agencies; source fuel on LinkedIn, X, Upwork, and niche Facebook groups. When candidates apply, don’t start vetting immediately: first make them very excited (growth rate, vision, revenue numbers, reviews, influencers you work with), then evaluate. Process for flywheels: intro interview, case study plus presentation, 2-week paid trial, hire. For fuel: portfolio check, a small practical case (“here’s the product page, make 3 static ad concepts”), one interview, hire. Never hire out of desperation (it fails 19 of 20 times), fire quickly, and expect flywheels to show strategic impact inside the first month: one B-player can turn your A-players into B-players.

The exact team behind our €1M+/month brand, with no big agencies involved: growth runs on a Head of Growth, a growth intern, a creator intern, a freelance Google Ads specialist, a freelance email marketer, and an ads uploader. Creative is the biggest production arm: Head of Creative, 3 freelance creative strategists, 1 freelance videographer, 5 video editors, and 3 designers. Operations is just an Ops, Supply and Ecom Manager plus an ops VA. Customer support is the largest single department at a manager plus 8 reps, because happy customers mean repeat purchases. Plus founders.

Brand manager is the most overrated position in a DTC team: you do not have an actual brand until at least $50M. Caring about brand guidelines and brand consistency actively hurts performance while you are trying to scale. Scale first, earn customers who love the product, then worry about brand.

Hire in this order: customer service manager, video ads editor, finance/data VA, static ad designer, creative strategist. The customer service manager should double on store management. Three of the five roles exist purely to feed the creative machine, and nobody on the list is a media buyer or brand manager.

The best creative strategists usually don’t come from marketing backgrounds. They come from engineering, journalism, and sales: the job is mostly research, breaking down problems systematically, understanding feedback loops, and understanding people. An engineer thinks in tests and feedback loops, a journalist does proper research and tells a story, a salesperson handles objections and knows what actually moves a person. Frameworks, direct response fundamentals, and formats are teachable; empathy and aptitude for this kind of thinking are not, so hire for those.

Train your first creative strategists yourself, then poach the rest, mainly from agencies. Use LinkedIn to find and reach out to them: agencies have already paid for their reps across dozens of accounts.

A creative strategist is the brain and manager behind your ads: they turn customer research into the angles, formats, and messaging that convert, analyze winning ads to build iterations that scale further, and run the creative meetings, timelines, and resource allocation. They also write scripts and ad copy and manage the designers and video editors. Their real job is making sure you are not throwing random shit at the wall and that new creatives ship consistently. A good one is worth their weight in gold.

A daily KPI tracker in Google Sheets beats any SaaS tool because tools are rigid. Sheets let you manipulate data instantly and run custom analysis and calculations on the spot, and being immersed in the numbers every day gives you a feel for the business that passively glancing at a dashboard never will. Hire a data and automations expert from Upwork to automate the data input so the manual entry doesn’t become the bottleneck.

ROAS doesn’t tell you if you’re profitable; your unit economics do, so derive target CAC first and work backwards. Track daily: AOV, landed cost (COGS plus fulfillment), contribution margin 2 (AOV minus landed cost, what’s left to acquire a customer), and target CAC based on CM2, LTV, and cash flow. Starting out, set target CAC below CM2 so you’re profitable on first order; once you understand your economics you can raise it to make scaling easier. Example: $100 AOV, $25 landed cost, $75 CM2, $25 desired profit gives a $50 target CAC, which means you need 2+ ROAS. Simplified, but this is how to set metrics from actual unit economics instead of arbitrary ROAS targets.

Acquiring at $60 CAC on a $40 AOV does not lose you $20 upfront; after COGS, shipping, handling, fulfillment, and payment processing you are realistically losing about $30 per first order. The bigger issue is cash flow: you cannot just “outspend competitors” without significant outside financing, an already established base of retention orders generating cash, or deep pockets to sustain months of negative cash flow. With strong retention AND the cash to fund it, this genuinely is how you scale supplements and consumables, but you need bulletproof unit economics and cash flow modeling before attempting it. Most brands that try run out of cash before the LTV materializes.

Set up ad naming conventions before you need them; at hundreds of launches a month you’ll regret skipping it. When every ad is named “Video 1” or “Static 2” you can’t analyze which angles, hooks, or formats are working at scale, can’t track which creative strategists, editors, and designers make the winners (so you can give them more work and pay them better), can’t answer “which ads did Emily make in the last 2 weeks” across 800+ running ads, and team communication breaks down. Set a system, stick to it religiously.

Successful Q4 starts with spending now: increase budgets to fill top-of-funnel before CPMs spike. Ensure enough stock, because scaling hard without inventory destroys your Q1; set up backup suppliers and shipping options; secure cash flow and get credit lines approved early. Dial in customer support and scale headcount if needed, finalize BFCM offers with retention offers and higher-AOV bundles included, remake your best-performing content with BFCM and holiday messaging, and write the Q4 email campaigns now, planned at higher send frequency through peak periods. (posted September 2025 · practice might be outdated)

Not getting banned from Shopify Payments is not that hard. Ship the product, don’t promise what you can’t deliver, offer good customer service, and if you run subscriptions let customers cancel without some crazy flow. A clean processor relationship makes life easier: no payment processing fires to fight, full focus on growth. (posted May 2026 · practice might be outdated)

A signed-for £500 delivery can still become a “product not received” chargeback the bank sides with, and the real damage is your chargeback rate, not the lost order. Most payment processors hold your funds or terminate your account if the chargeback rate hits 1%, exactly what you cannot afford going into Q4 when transaction volumes spike. The fix is fraud prevention that tracks buyer behavior across thousands of merchants: Chargeflow flags customers with a history of bogus chargebacks before they even complete checkout, so you keep your goods and skip the endless documentation grind. (posted September 2025 · practice might be outdated)

On Alibaba, only approach Gold Suppliers with high ratings, trade assurance, and a portfolio of similar products; message 10-15, order samples, and never pick on price alone. Filter for responsive communication, good English, and willingness to do small MOQs; or pay a good sourcing agent and lower the chance of getting screwed. For quality, order samples from multiple manufacturers, use the product yourself for weeks, then hand samples to people in your target demographic and ask pointed questions: does it solve the claimed problem, how does it compare to alternatives they have used, would they buy at this price. Genuine excitement means you have something; a polite “it’s nice” means mediocre.

Claude Opus with Projects and Skills beats everything else for copywriting. Better than Claude Code even with a whole memory base attached, better than ChatGPT 5.6. The setup: one project per product, fed with as much high quality context as possible, plus a skill for every format being written, trained on winning examples and direct response foundations. (posted July 2026 · practice might be outdated)

A five-tool pipeline covers the whole AI video ad workflow. Claude for scriptwriting, ChatGPT for prompting images, Gemini for image generation, Kling 3.0 via Higgsfield for image to video, ElevenLabs for voiceover, then assemble everything in an editing tool. For certain formats Grok Imagine gave better image to video results than Kling. (posted March 2026 · practice might be outdated)

Gemini for image generation, Claude for writing copy, ChatGPT for when Claude hits its usage limit. (posted September 2025 · practice might be outdated)

A deep research prompt plus a format to mimic gets AI 40% of the way to a good ad, and you’ll burn a lot of money testing that 40%. Getting AI to write an ad that actually performs requires steering it with detailed feedback, which is only possible if you deeply understand your audience and what resonates with them, human psychology, and which viral formats work and how to adapt them. That judgment is built only by writing manually over and over, the hard path. Bad judgment + AI = scaling garbage; good judgment + AI = scaling wins.

One ChatGPT Deep Research prompt can produce a ranked competitor map: fill in [PRODUCT] and [PRICE RANGE], turn Deep Research on, and run. The prompt scopes to real D2C brands selling on their own site in US, UK, AU, CA (Amazon or Etsy sellers only if they have a real brand site, generic dropship clones excluded) and hunts through core and advanced search queries, “powered by Shopify” footprints, Google Shopping and image search, Meta Ads Library, TikTok Creative Center, Pinterest Ads, marketplaces, review platforms, “best of” roundups, and PDP similar-item widgets. It keeps 12 to 20 brands showing at least three traction signals (own storefront with clear identity, ad activity in the last 90 days, meaningful reviews or engaged social following, aligned pricing), then captures each brand’s URL, hero SKUs, avatar-plus-promise positioning, marketing angles, and traction clues.

Each competitor gets scored 0 to 5 on brand clarity, offer strength, acquisition readiness, proof and trust, differentiation, and assortment/LTV potential, then ranked in a sortable table. The output closes with a synthesis: white space (underserved avatars, unmet needs, ignored benefits), positioning moves (a stronger promise or different avatar than the market norm), and offer improvements (guarantees, bundles, bonuses, urgency, subscriptions). (posted October 2025 · practice might be outdated)

Tools do not solve skill problems. “My creators don’t perform, so use Billo or Insense” is bad advice: the real gap is knowing how to source creators, sell them on your brand, write clear briefs, negotiate rates, and onboard them until they are excited about your product, which is boring, slow work most people skip. Same with “my landing pages aren’t converting, which tool should I use”: a Replo subscription will not fix a page when you do not understand customer psychology, which objections to handle, or how to structure a value proposition; with good copy you could technically sell any product from a Google doc. Comfrt scaled to $500M with thousands of creators not through some new tool but through years of creator relationships, aligned incentives in deal structures, in-house creator training, and a product and mission creators wanted to talk about. If you don’t build the skill, the tool is useless.

Edit your own videos, write your own copy, build your own store, set up your own email flows, run your own ads. It is the only way to actually get good and understand how the work gets done, and you cannot manage someone later with zero clue: you cannot evaluate their output, tell if they are good, or give useful feedback. Having done it yourself teaches what is hard versus easy (no unreasonable demands), what good work looks like (spot mediocrity), where bottlenecks happen (actually solve problems), and what to prioritize (stop optimizing the wrong things). Learn the fundamentals, then hire and delegate effectively; skipping the learning phase and immediately hiring people to do everything is how you build a fragile business.

Don’t start dropshipping with only $2K to your name; you’ll most likely lose it. The better path: learn one skill ecom brands actually need (media buying, creative strategy, email marketing, building landing pages, CRO), then get a job at a brand or agency first. That buys real experience seeing what actually works, a salary while you learn, time to save capital for your own store, and a network of people who can help later. Once you have 6 to 12 months of experience and $10K+ saved, then consider starting your own thing.

The most successful people picked one industry and stayed in it for an unreasonably long time. They know every player, trend, and opportunity before it becomes obvious, hold supplier and partner relationships that take years to build, and move first when something new appears, while industry hoppers restart from zero every 2-3 years. Pick a lane and stay longer than feels comfortable; ambition plus hard work over enough years in one industry makes success near-inevitable. Do not confuse one industry with one skill: be a generalist with a few skills you are exceptionally good at.

There is always more to do than you can do, and the tasks you enjoy most usually don’t move the needle. Keep two lists in your notes, a to do list and an ideas list, and score every item 1 to 10 on ICE: impact on the North Star KPI (revenue or profit), confidence it works, ease of shipping, then work top down on 1 to 3 high leverage tasks a day. Sample scores: new logo and store redesign 2 x 3 x 5 = 30, moving to a 3PL for faster shipping 8 x 7 x 2 = 112, improving the abandoned cart flow 4 x 5 x 5 = 100, hiring a new creative strategist 10 x 8 x 4 = 420, making new ads on a new winning angle 9 x 7 x 8 = 504. Big bets like launching a new product skip the system and get decided on strategy.

You will never decide with 100% certainty; 40% is often the ceiling, so treat every test as buying information from the market, not as a win or loss. Do your research, make your best educated guess on the angle, and execute it as well as you possibly can. If it works, iterate and scale. If it fails, also good: you now have answers and better data. Every angle you test either makes you money or teaches you something about your market, and both outcomes move you closer to what actually works.

Peter Thiel’s decision model: divide your reasoning into parts, find the single most important factor, and if no one reason justifies the decision on its own, do not do it. Pros-and-cons lists fail because they weight every factor equally, and “extra benefits” distract from the actual goal: an extra feature when evaluating SaaS, “also boosts brand awareness” when you need conversions, an influencer’s 2M followers when you need converting traffic. Ask what the one reason is that makes the decision make sense; if you cannot name one compelling reason, you are not thinking hard enough about what matters. Most strategic decisions fail because people try to solve five problems with one solution.