48:04Think Media Podcast · Interview
What's Actually Working on YouTube Shorts in 2026
Pat Flynn built a 2.7M-subscriber Shorts channel from scratch — solo, no team, 20 minutes a day — and breaks down every tactic live.
June 23rdThe argument in one line.
Views are the least useful growth metric because a video only builds a business when it earns trust from a narrowly defined audience, and trust scales fastest when data picks the topic while creativity supplies the insight.
Read if. Skip if.
READ IF YOU ARE…
- A creator or course/coaching business owner who already has an offer ladder and wants a data-backed system for choosing which video topics to make next.
- Someone drowning in views-only analytics who wants a better way to judge whether their content is actually building a business.
- A YouTuber or short-form creator open to running content in strict batches and tracking outlier scores against a set of competitor channels.
- Anyone curious about AI-assisted content research tools, including hooking a research tool up to Claude via MCP for topic mining and script templating.
SKIP IF…
- You want a quick viral hack rather than a full research and metrics system that takes weeks of batch-testing before it pays off.
- You don't have an existing offer or email funnel yet — the whole framework assumes content is feeding a ladder of paid products.
The full version, fast.
Kallaway argues views are the least important metric a creator can study, because a video only compounds into a business when it earns trust from a narrowly defined audience. His 'content minutes' framework maps how many minutes of trust a viewer needs before buying at each price point on an offer ladder, which is why short-form and long-form both need to funnel into email. His three real metrics, ranked, are emails collected per video, followers gained per video, then views last. To find topics, he built Sandcastles: a tool that tracks mid-size competitor channels, scores their videos against a rolling 3-month outlier baseline, and runs deep AI analysis on hooks and structure. He sources 70-80% of early topics from that competitor data, then supplies 100% original creative insight on top of it.
Free for members
Chat with this breakdown — free.
Sign in and you get 23 free chat messages on us — ask for the hook, quote a framework, find the exact transcript moment, generate a markdown action plan. Bring your own key when you want unlimited.
Create a free account →Who's talking.
00:00hostJay Clouse
00:12guestKallaway
Where the time goes.
00:00 – 01:03
01 · Intro to Kallaway
Cold open on Kallaway's core philosophy of de-risking effort with data, then Jay introduces him: four failed businesses, no audience, then over a billion views and a 50% viral hit rate.
01:03 – 02:12
02 · Introducing "content minutes"
Kallaway introduces content minutes: map your offer ladder from cheapest to most expensive and estimate how many minutes of trusted content a viewer needs before buying at each tier.
02:12 – 03:45
03 · The modern content stack
Kallaway's triangle framework: one short-form platform and one long-form platform (YouTube), both funneling into email, the actual asset.
03:45 – 05:40
04 · Why you want a narrow ICP
Why a narrowly defined "audience of one" beats broad appeal, and why 10-20K views per post is plenty once you have that narrow focus.
05:40 – 08:34
05 · The 3 metrics that matter
Kallaway ranks the three metrics he actually tracks: conversions (emails) per video, followers gained per video, and views last.
08:34 – 10:12
06 · Sponsor break: 1of10
Mid-roll sponsor segment for 1of10, the YouTube strategy consulting team that helped grow Kallaway's channel.
10:12 – 14:38
07 · Where topics come from: lead magnets and idea sourcing
Kallaway explains he builds lead magnets as reusable inventory rather than reverse-engineering topics from them, and that 70-80% of his early topics come from data on what's worked for other creators.
14:38 – 16:14
08 · Sandcastles demo: building a channel watchlist from scratch
Live walkthrough of building a curated watchlist of niche creators inside Sandcastles, including the "anchor" feature that finds channels closest to one you already admire.
16:14 – 22:18
09 · Why you study mid-size creators, not mega-famous ones
Why mid-size (10K-250K follower) creators are the best research target, how to filter by engagement rate to exclude boosted brand deals, and how the outlier score normalizes performance across channels.
22:18 – 24:59
10 · Deep video analysis: hook formula, storytelling structure, contrarian angle
Sandcastles' AI breakdown of a single video: its topic, unique angle, contrarian take, hook formula, and storytelling structure, and how Kallaway uses that without directly cloning it.
24:59 – 30:47
11 · Kallaway's research flow
How a dedicated researcher filters competitor videos daily, builds a running doc of candidate ideas, and how the two of them greenlight 8-10 approved ideas per week on a standing call.
30:47 – 36:34
12 · Solving structure and scripting problems using your own channel's data
Once you have your own back catalog, sort your best-performing videos, study their structure, and use AI to extract a reusable script template rather than fully outsourcing the writing.
36:34 – 38:19
13 · Data-enabled creativity: the thesis behind the whole system
The closing thesis: dial each stage of the content pipeline between fully data-driven and fully creative, and a thought experiment on stripping all visuals from a video down to talking-head and captions.
Lines worth screenshotting.
- A trust-based 'content minutes' framework, not views, predicts whether a viewer will buy: a $47 offer might take 20 minutes of watched content to earn trust, a high-ticket program can take hundreds.
- The modern content stack is a triangle: one short-form platform and one long-form platform, YouTube, both funneling into email, which is the actual asset being built.
- Views only convert to revenue if the topic accrued trust with a narrowly defined audience; a broad topic can pull 100,000 views that carry zero buying intent.
- Algorithms have gotten targeted enough that 10,000-20,000 views per Instagram post is plenty to drive meaningful email list growth, more views past that mostly just satisfies ego.
- The three metrics that actually matter, ranked: emails collected per video, followers gained per video, and total views last, since views is the only one visible on competitors' channels.
- A follow is worth 5-10x more than a view because someone only follows out of fear of missing your next video, not from casually watching one.
- Studying massive creators for video ideas is a counter-signal: when someone with millions of followers has an outlier hit, built-in fame explains it more than the idea did.
- The best channels to study for topic ideas are mid-size, roughly 10K-250K followers, because their outlier hits prove the idea worked on its own merit.
- 70-80% of a new channel's early topics should come from data on what's already worked for adjacent creators; only about 20% should be untested original ideas.
- An 'outlier score' compares a video's views to that same channel's rolling 3-month average, normalizing for channel size and growth trajectory so performance compares fairly across channels.
- Topic research should stay tightly within your niche, but hook formats, thumbnail styles, and editing structure can be borrowed from completely unrelated niches.
- The single highest-leverage thing to get right in a video is the topic itself: a great topic works with almost any hook or edit, a bad topic fails no matter how well it's produced.
- A dedicated researcher can filter competitor videos down to 8-10 approved ideas twice a week by asking one question: do we have something non-obvious and contrarian to say about this?
- Each stage of the content pipeline, topic, research, hook, script, record, edit, post, can be independently dialed anywhere from 100% data-driven to 100% manual and creative.
- Fully automating content with zero originality trends toward average, because too many creators pull from the same data and converge on the same ideas.
Chase trust and topics, not views
WHAT TO LEARN
The metric that predicts revenue is how much trust a video earns from a narrow audience, and the fastest way to find topics worth that trust is data first, originality second.
01Intro to Kallaway
- Four failed businesses taught Kallaway the same lesson each time: no audience means no customers, no matter how good the product is.
- Building an audience isn't a nice-to-have marketing channel, it's the leverage that makes every other business decision easier.
02Introducing "content minutes"
- Map your offer ladder from lowest to highest price, then estimate how many minutes of trusted content a viewer needs before they'll buy at each tier.
- A cheap offer might convert after 20 minutes of content; a high-ticket program can require hundreds of minutes of accumulated trust first.
- Long-form video builds trust faster than short-form because a single sitting can deliver 60-200+ minutes of content in one binge.
03The modern content stack
- Run exactly one short-form platform and one long-form platform (YouTube), and route both into an email list, that combination is close to unbeatable.
- If you have to choose one format first, pick YouTube: short-form is easier to bolt on later once you already have a working long-form formula.
04Why you want a narrow ICP
- Picture one specific person in a chair and make every piece of content valuable to them; a broader, more generic topic gets more views but accrues zero trust.
- Modern algorithms target so precisely that 10,000-20,000 views per post is already enough to drive meaningful email growth, chasing more views past that is mostly ego.
05The 3 metrics that matter
- Track emails collected per video as your top metric, every video needs a CTA into a lead magnet, measured through a tool like ManyChat.
- Track followers gained per video as your second metric; a follow is worth 5-10x more than a view because it signals real fear of missing your next piece of content.
- Only use total views to study what's working for other creators, since it's the one metric you can actually see on someone else's channel, on your own channel it's the least important of the three.
- Not all views are equal: a million-view entertainment video builds far less trust than a 40,000-view education video that's dense with value for a narrow audience.
07Where topics come from: lead magnets and idea sourcing
- Build lead magnets as reusable inventory ahead of time, then attach the closest one to whichever topic tests well, don't reverse-engineer topics from lead magnets.
- The single highest-leverage decision in content is the topic itself: a great topic survives almost any hook, structure, or edit, while a bad topic fails no matter how well it's produced.
- Early on, source 70-80% of your topics from validated data on what's already worked for similar creators, and reserve about 20% for untested original ideas, tracked in batches of 10-30 videos.
08Sandcastles demo: building a channel watchlist from scratch
- Build a tight, hand-picked watchlist of creators in your niche rather than following everyone, most people don't actually know who the good performers are in their own space.
- Anchor a research tool on one creator you already admire and let it surface the closest-matching channels, instead of trying to describe your niche in words.
09Why you study mid-size creators, not mega-famous ones
- Study creators in the 10K-250K follower range, not mega-famous ones, when a huge account has an outlier hit, built-in fame explains it more than the idea does.
- Filter competitor videos to a 2%+ engagement rate before analyzing them, since ultra-high-view, zero-engagement videos are usually boosted brand deals that don't reflect real organic performance.
- Topic research should stay inside your niche, but hook formats, thumbnail styles, and editing structure are fair game to borrow from completely unrelated niches.
10Deep video analysis: hook formula, storytelling structure, contrarian angle
- For any competitor video worth studying, break it down into its topic, its unique angle, the common belief it contradicts, its hook formula, and its storytelling structure.
- Don't clone a competitor's exact example and framing, find a different example or angle you have a genuinely unique take on, and apply their structure to that instead.
11Kallaway's research flow
- Delegate first-pass filtering to one dedicated person who screens competitor videos daily against a trained sense of 'interesting' and 'do we have something non-obvious to say.'
- Run a standing twice-weekly review where each candidate idea gets pitched, gets a contrarian angle proposed, and is either killed or greenlit into an approved bucket.
- A disciplined research process should produce roughly 8-10 fully-approved, data-validated video ideas per week, each vetted for having a genuine, non-derivative point of view.
12Solving structure and scripting problems using your own channel's data
- Once you have 10-20 of your own videos, sort them by performance and read the top scripts to internalize what structure is working, don't copy-paste, let it train your instincts.
- Feed your best-performing scripts into an AI tool and ask it to extract the common storytelling structure into a reusable template, then treat that draft as a 90% first pass you still rewrite by hand.
13Data-enabled creativity: the thesis behind the whole system
- Every stage of content creation, topic, research, hook, script, filming, editing, posting, can be independently dialed anywhere from 100% data-driven to 100% manual and creative.
- Fully automating content with zero originality trends toward average, because too many creators pull from the same data and converge on the same ideas, originality is what a fully automated pipeline can't produce.
Terms worth knowing.
- Content minutes
- Kallaway's framework for how many minutes of a viewer's attention and trust you need to accumulate before they'll buy at a given price point on your offer ladder.
- Modern content stack
- A three-part content system: one short-form platform and one long-form platform, typically YouTube, with both driving viewers into an email list.
- Outlier score
- A Sandcastles metric showing how many times higher a video's views are compared to that channel's own rolling 3-month average, used to compare performance fairly across channels of different sizes.
- Sandcastles
- An AI research tool built by Kallaway that tracks competitor channels, scores and analyzes their videos, and connects to Claude via MCP so creators can find and validate video topics with data.
- ICP (audience of one)
- The specific type of person a piece of content is made for; Kallaway argues creators should picture one person in a chair rather than trying to appeal broadly.
- 1of10
- A YouTube strategy consulting firm and title/thumbnail research tool that has worked with Kallaway's channel and sponsors this episode.
Things they pointed at.
Lines you could clip.
00:47
“Views are not the right metric to study. Views are a metric to study.”
07:56
“A follow, when someone goes to follow, it's like five to 10x more powerful. 'Cause they're like, that was so good, I can't miss it again.”
14:14
“If you're a derivative, you're going to lose. So you have to be original, but you have to fish in the right ponds.”
21:18
“The outlier score, which is this green badge here, you can see like this one is 8.2X. This means that the views of this video were 8.2X higher than the average views in the rolling three months before this video was posted.”
36:19
“I don't think you can fully build an AI factory pumping it out with zero creativity and cut through. I just think there's too much supply and it's going to trend towards average.”
Where the conversation goes.
00:00 – 01:03sparseIntro and guest origin story
01:03 – 02:12denseContent minutes trust framework
02:12 – 03:45denseModern content stack (short-form + long-form into email)
03:45 – 05:40steadyNarrow ICP theory
05:40 – 08:34denseThe 3 metrics that matter
08:34 – 10:12sparse1of10 sponsor segment
10:12 – 14:38steadyViews trap, lead magnets, and topic sourcing
14:38 – 16:14denseSandcastles demo: channel watchlist
16:14 – 22:18denseStudying mid-size creators and the outlier score filter
22:18 – 24:59denseDeep video analysis walkthrough
24:59 – 30:47denseKallaway's research team workflow
30:47 – 36:34steadyUsing your own channel's data for structure and scripting
36:34 – 38:19steadyData-enabled creativity thesis and the visuals tradeoff
Word for word.
Read-along
Don't just watch it. Burn it in.
See every word as it's spoken — crank it to 2× and still catch all of it. The same dual-channel trick behind Amazon's Kindle + Audible.
metaphoranalogystory
I'm trying to de -risk my effort slash increase my hit rate. So I was like, how do I use data, but not make myself a clone that's copy pasting everyone else's stuff? And that's basically why I built Sandcastles.
That's Callaway. He started four businesses that were all killed by the same problem. No one cared.
He had no audience, which meant he had no customers, so he made it his mission to crack audience growth. And now he has over a billion views, a 50 % viral hit rate, and he created a tool called Sandcastles that helps other creators do the same. In this episode, you'll learn the exact system he uses to find and validate viral video topics before he ever hits record.
My framework for content is, I call it the modern content stack. The metrics you need to obsess over so that your content grows your business. Views are not the right metric to study.
Views are a metric to study. And I would say the third most important when I look at the stack. And if you stick around long enough, Callaway will show you exactly how you can build your own content research system from scratch.
Thank you to one of 10 for sponsoring this video. There's kind of a fallacy between like views equals money. It's really like views does not equal money.
You want density around one ICP as many times as you can possibly get it. In a given week. So like my framework on this is called content minutes So essentially like you could imagine your offer ladder or whatever your offer stack at certain price points So I don't know what you're like running right now But I know you have the lab which is probably one of your higher ticket things and then as you go down you've got like Course bundles individual courses.
Maybe you have something at 47 So essentially like imagine you took that ladder from low to high and you Overlaid an amount of minutes that someone would have to watch your content and trust you for those minutes that's like pretty critical and at the 47 maybe it's like 20 minutes worth of content and at 20 minutes they trust you enough to buy it if it fit their problem and then one one rung up it's like 67 minutes and then you keep going up to like to get into the lab they probably would have to have watched hundreds of minutes but like once they cross that threshold then they'll convert so If you think about it, most people are like, well, I should just do YouTube because every YouTube video, if it's 20 and yours are podcasts, like 60 minutes long, the average watch time, let's say is 20 to 30 minutes.
You're getting a lot of those blocks in one shot. And so you can scale someone up that ladder of trust pretty quick, which is why you've had so much success with YouTube, right? Because somebody watches five episodes that might happen in a single night.
boom, you hit 200 minutes and then they'll, they'll be willing to buy anything. My framework for content is I call it the modern content stack, which is basically a triangle. You've got short form, long form, both going into email.
That's like the golden goose. If you, if you could do anything, it'd be one short form platform. I think Instagram is the most valuable or LinkedIn, depending on the audience for you is probably Instagram, YouTube for long form, both into email.
And you pretty much can't lose. Now from a content minutes perspective, you may think if I had to make a trade off of YouTube versus short form, you're going to want YouTube.
And I agree with that, right? The reason most people want to start with short form is because YouTube is so difficult, like so many moving parts, you've already cracked it. However, once you're getting what you're getting from YouTube and you kind of have a formula, that's when you look to expand the universe into short form to kind of get more breadth of the minutes.
To bring it all back to what I was starting with in terms of the narrow ICP, the minutes don't accrue unless the person who watched the video trust you and actually have the problem you're talking about. So if you broadened, or, you know, made more generic the topic you're talking about, yeah, maybe it gets 100 ,000 views instead of 20, but actually none of those 100 ,000 blocks hit because no trust was accrued because the topic you had to talk about was so generic and like surface level that it didn't actually accrue trust.
So for me, it's like what I try to get people to do is very narrowly define your ICP. I like to think of it as an audience of one. So like if there was just one person in a chair in front of you, that's like probably somebody in the lab, you can just picture them right now.
It's like, If I could give value on every single piece of content to that person, there's a lot more people like that and it would scale accordingly. The algorithms in the last like one to two years have gotten so good at targeting the exact type of person you want that you don't actually need more than like 10 to 20 ,000 views per post on Instagram other than for ego reasons.
You can drive a ton of email lead gen from that. So I would say like the first thing is narrow is good. There's things we can talk about in terms of how to expand the...
saturation within that narrow slice, like the way you frame the hooks and the topics, but like you want narrow. If our ICP are other creators who are educational in nature for their audiences, my perspective has been that education has been harder to do well on, especially Instagram and TikTok, but that might be a limiting belief too.
When you see people who are doing education in short form video and doing it well, What is true about those folks that they're being successful in a world where there's a lot of just like great, pure entertainment that is probably more engaging? Yeah.
How do you find doing it? Well, are you talking about views? That's a good question.
I would say doing it well in terms of they are making Instagram worthwhile for them. This is like the ultimate answer. I would say that probably in the back of my mind when I asked that question was like, okay, yes, they're doing good views.
They're doing views on par with people who aren't doing education. You know, the target market for this is actually fairly large. So if views are in the single digit thousands, it feels like there's just a lot being left on the table.
So views as a metric doesn't matter in an absolute sense, but it feels like we are really barely scratching the surface. 100%. So I would say views are not the right metric to study.
Views are a metric to study. And I would say the third most important, when I look at the stack, I also think not all views are created equal, right? So an entertainment video that has a million views is not getting the trust accumulation the same way an education video, it's 40 ,000 views, it's super dense.
So that does make it hard. The hard part is this decoupling from like the thing you can see from other people, which is views. your own, like all of our brains are like see number, want number bigger, compare number, right?
And so it can cause a lot of people to stop or to go down the wrong path when they're anchoring views as the number one metric. And here's what I mean by that. If you know professional creators are the number one target audience and you know, all you have to do is get them to trust you and they will ascend down the funnel because you have all of that dialed.
You just got to get their email and you know that you can monetize them. If you know that's true. and you know that education content is the best way to get trust at scale, let's just hold those things constant, then you studying an entertainment video from someone else is actually distracting you from like the right benchmark, right?
The right benchmark is, am I making a high trust, high value video educationally to get one more content minute on the stack, like in a vacuum, right? Me looking at Logan Paul does nothing for me, right? Me looking at, I'm just thinking of like random entertainment names.
And so, The number one thing I think to go through is what are the best metrics to track? And maybe I could just share like how I'm approaching this on my channels because I basically solve short form in like a pseudo entertainment niche, you could say.
Like I was doing tech, which sounds educational, but really it's like entertainment. And I was doing that for a brand deal strategy. So like that was a game to play.
The game you're playing and the game I'm now playing on my Calway Marketing, completely different game. And so even me, like. I can't really look at tech as a proxy for marketing because it's completely different.
So maybe I can explain like what I'm doing on Cali marketing and how I'm looking at it. The three metrics that I think matter most, number one is conversions per video and conversions measured as emails collected per video. That means every video I have a CTA, every CTA is driving to a lead magnet where I can capture emails and I'm measuring all of those in many chat.
And I can go through like this batch process that I run. That's number one. Number two is followers gained per video.
That is more valuable than views because a follow, when someone goes to follow, it's like five to 10 X more powerful. Cause they're like, that was so good. I can't miss it again.
So before I looked at total views, I would look at followers gained and just say, which one is driving more followers last. I would look at total views now. The reason why total views is helpful when you're looking at competition is because you don't have insight really into how many email subscribers their videos are driving or how many followers gain.
Those are like private metrics behind the wall, but you could look at someone else's and see views. So if you're looking at like what's working for others, you have to settle for views for others because you can't see the other two. Something I admire about Callaway is his willingness to invest in experts to help him skip the learning curve.
That's what he did when he hired the one of 10 team to help with his YouTube channel. You probably know the one of 10 team as the leading tool for researching viral ideas, titles and thumbnails. It's what I use the most on my channel.
The reason the tool works so well is that the team is full of YouTube strategists who really know their stuff. What you may not know is that they also have a consulting arm that's not only worked with Callaway, but Ali Abdaal, Will Tennyson, Mike Shaikh, and a bunch of others too. Collectively, they've helped generate over 2 billion views for their clients.
The 1 of 10 team helped Callaway grow a new channel from zero to over 250 ,000 subscribers in just one year. They helped with this video, this video, and this video too. One video alone brought in over 2 million views and generated more than $100 ,000 in revenue.
If you want to work with the same team that took Callaway from zero subscribers to where he is now, that's the 1 of 10 strategy team. They are very selective with who they work with and only take on about five new clients per month. They focus on the things that actually move the needle.
Ideas, titles, thumbnails, and positioning. You get a dedicated Slack channel and regular one -on -one calls to communicate with their team to ensure each video is set up for success. This is not a course and it's not a one -time solution.
It's direct, ongoing support from a real YouTube expert who has done this for some of YouTube's largest creators. Book a free discovery call with the 1 of 10 team today with the link in the description and let them know that Jay sent you. But when you look at your own channel, the biggest trap I find is only relying on views from your own channel.
Because obviously if the TAM is bigger on a topic, you're going to get more views. And so if you expand the TAM beyond the trust zone, those views are actually empty calories. And so that's like just one anchoring methodology that I think about is, can we have enough lead magnets in the ecosystem that we can drive to a lead magnet on every video and then evaluate emails collected per video?
So the Callaway Marketing Channel I've been running for, I think I'm 35 videos in and the goal is... to get to two X per day. We're doing one per day, seven per week right now.
Every single one of them, I have a CTA, every single one of them I'm measuring. And what's funny is not always are the highest view ones driving the most emails, which is like, you would probably accept that to be true and same as me, but the average person would be like, whoa, I'd expect views equal emails. It's not the case, right?
It's all about what's accruing trust, what's targeting a pain point people really have a need for, and then you have a lead magnet complimenting. So when you're doing these CTAs with these lead magnets, Are you reverse engineering a video topic based on a lead magnet you made?
Do you have an idea for a video you love and then you create a lead magnet for that? And now that's just in the bank and maybe I'll use it again when it makes sense in the future. What's your order of operations?
Yeah, so I've done it both ways. So I have an academy where I like teach my whole system to people. And what I've done is I've taken pieces of the academy, extracted them out and turned it into a lead magnet.
So then I'm just like, I have that lead magnet made. If the topic suits it, I'll pull from it. So I guess like step one is you have a couple that you believe in, right?
Now, if I were going to like, what videos do you make? I wouldn't make them based on the lead magnet. I would make them based on what I can talk about next, which is like how you actually identify which topics are working.
I would say it's probably, maybe we can get into this now in terms of like topics. Because one of the questions you asked is like, what's the most important thing to look for, to focus on? By far and away, it's the idea of the video.
You can use almost any hook, any storytelling structure, any editing, any format. If you have a bad topic and I can talk about what bad and good looks like, the video will fail. It'll be medium at best.
But if you have a great topic, you can plug in almost any of those other things and it will almost always work. So all like the 80, 20, 90, 10 is can I pick topics that are validated, relevant to pain points they want, and that I have a non -obvious contrarian take about. So for me, like if I had a pie chart of where topics come from, I would say like at the beginning, 70 to 80 % comes from data for what other creators topically have talked about that have worked.
And I'm looking at views because that's all I have access to. And so I'm informing my first couple batches of content based on that. 20 % comes, one out of every five comes from like an original idea where I'm just like, let's just test this.
That first couple batches, and I do batches of 10. So let's say like 20 to 30 videos is all about testing. I'm basically just trying to validate which topics work best for me.
Once I then know that we can, we can talk about like downstream, but you know, topics are the most important thing. And what I'll do is based on the topic that I choose, I'll then plug in whatever the closest lead magnet would be to that topic. The way I think of it is I don't really want to spend every hour making content.
Like I'd rather live my life. Right. And so I'm trying to like increase, I'm trying to de -risk my effort slash increase my hit rate.
Every time I sit down, I want it to bang so that I get the results. So I don't have to spend more hours doing it. That's kind of like how I think of it.
So it's the . Like what would be the way to de -risk? The way to de -risk is to use data on the topics and then creativity on the contrarian insights.
That combo is like, that's what all the goat creators are doing. I mean, I would say on YouTube and that's what I do on YouTube too, on YouTube and short form. And it makes it so much easier.
And that's basically why I built Sandcastles, right? So like. For anyone who goes to sandcastle .ai, that's like my software tool, AI tool that does a lot of this for me.
The reason why I built that is because I went data first from the beginning because I really did not enjoy the process of spending 10 hours on a video and it failing. I like hated that a lot. So I was, which I'm sure everyone does obviously, but I was like, I can't do this forever.
Like I can't do the guessing. So I was like, how do I use data, but not make myself a clone that's copy pasting everyone else's stuff? Because I also think.
That's where you lose too. If you're a derivative, you're going to lose. So you have to be original, but you have to fish in the right ponds.
That's the key thing. If you're fishing in every pond, some ponds don't have fish. So it's like, how do we figure out where to fish and then use our own technique, our own contrarian insight, our own, you know, whatever the metaphor, our own boat to go get more fish, right?
So the way I do it with Sandcastles, which I think is the best way right now is - Can you show us this? So this is Sandcastles. In the upper left, you can build like separate workspaces for different niches.
So like - You might want to say okay professional creators is like my main one But you might have other like verticals you want to test if you want to if you want to partition them separately, right? So this is my marketing workspace.
Basically. The first thing you want to do is build this channel list So if you go to the channels tab, I've got this list here of like all the top Creators that I've come across or that I know of in my category and like you like I said You can either build this one time or like over time, but how you would do it is Maybe let's just delete all of these so I can show you kind of the process from scratch.
So typically you would come in here and you'd be like, I don't know who to add at all. So you go to describe, and let's just say you're looking for, like for me, social media marketing and growth content. And I can say, I only want people from Instagram or I only want people from micro and small cat, which is like 10K followers to 250K.
Now I'll go up to medium. The reason you want this micro to medium range is because typically, and this is like not intuitive, but If you track videos from massive people like Hormozy, let's say, who has like four or five million followers, typically when he has an outlier, it's not necessarily just because the idea did so well.
It's because he's so famous and has so much built -in brand equity that people will like anything he does. So it's actually a counter signal to study the huge people. You just won't find the best ideas.
Conversely, let's say you're not well -known at all. If you have a banger, that means the idea, the concept, the substance was the reason why. It's not because of your face, because nobody knows you, right?
So you want to study people that are slightly smaller in the, that's where really the golden ideas come from. So let's just do like these three and we'll press search. So what Sandcastles does is it breaks whatever you gave it into three different buckets.
So in this case, social media marketing, social media growth, marketing and growth strategies, like they're all kind of derivatives of this. And this will give you the best ranked channels that we have in our database. We have like 100 ,000 channels or so.
And if you want to see more, you can just load more. And the goal is you just hover over these. And if you want to investigate these, you can just click and open it in a new tab.
Boom, this person's in a new tab. I don't know if it's sharing. It's probably not sharing.
But I open it. Boom, I see her channel. Here it is.
And what you want to do is build out your list. So I could say like social with Philip. Boom, press the plus.
And I build him out here. Now, you can do this quickly by just pressing Add All. And what we do in the onboarding is automatically build a list of 20 for you.
But I find most people want to tailor this pretty tightly. Now, the cool thing about a few hidden features we have here is, let's say, so Jaden is really dope. He's dope in this space.
I've seen his stuff. So I'm going to add him. Now, what's cool is I can hit this little double arrows here.
And what we'll do is... make him the anchor. And now these are all the best fits based on like closest to him.
So if you, let's say you're not good at describing what you want to do, but you know, there's one anchor person that you love. You can just go to him, press this boom. Now this list is like the closest possible thing to him.
So I would just like add all these people in. Right. So that's how, that's one way to do it.
The other way is if you know, someone's good, like this girl, Ava is really good in this space. I'll just search her name, personal brand launch. And if you just wait a second, it will find her.
So she's right here now. Like I said, we only have 100 ,000 channels. Like these are top curated channels and our users, every time a user adds one, it gets added to the database.
But if there's somebody that exists that you know should be there and they're not, it just means nobody has ever added them to Sandcastle. So you can actually do it for free. You just go to add URL and you can type it in.
So we have this, I'll go to parse. If we have you, the picture shows up, boom, I add you, no problem. If we didn't have you, it would be like new channel and you can import it.
And then when you import it, we'll. taking your full backlog. Right.
And so this process of building out your list is super valuable because now you can focus your attention. You can focus your research attention on these channels, which you know, to be good. So what I recommend for people is like almost everybody knows at least one name of someone that's pretty, pretty good in their niche.
Start with that person, search them or add them, and then just go to this kind of refactor. And then once you hit this refactor, you can just be like, okay, boom, these are all the people that are really close to this person. So We'll add a bunch in there.
One thing that's interesting about this approach that's a little bit counter to like the YouTube advice in a similar fashion is a lot of times the advice on YouTube is like, get outside your niche. Find somebody who's doing something in finance or gaming that's interesting and try to pull from their video ideas, which I think makes sense, but I don't know.
It doesn't seem like it would translate exactly to Instagram, but I'm having a hard time articulating why that's true. Yeah, so I think there's like different buckets, right? So when you do the YouTube research, If you're looking for topic inspiration or interest topics on YouTube, you need to have people that are in your space or like very close adjacent.
But if you're looking for title formats or thumbnail inspiration or series ideas, you can go broader, right? For example, if you were doing creator stuff, you typically wouldn't look in the e -com niche for like interest topics, but you could find a title format that's crushing and then translate it over. It's kind of similar here where if you're looking for topics, you need the topics to be.
within niche. But if you want to look for formats or you want to look at editing styles, or you want to look at hook, hook, you know, hook formats, you could do that outside the niche as well. So what we found is like, most people have trouble building a list outside of their core niche.
Like most people are not aware of like, who's good, what's good, even like this data backed process. So we try to keep it in niche. And then you can, I'll show you a couple of things we have where you can stumble across new stuff.
That's like outside the niche. But this is what I like to start with. So this is my list.
You can see I just, I had a list on Excel. I just pasted all the URLs in. I added them super quick.
So this is the list that I like. So once you have this, then you go to the videos tab. And so this is, to zoom out, what we're doing here is we're breaking down.
how do you find the best performing topics in your niche so that you know which ponds to fish in? So if you go to the videos tab, what you now see are, these are all the videos from just those channels. And you know that because if you click the channels filter, you can see the channels in here.
So if I wanted to look at like just my channel, I'll click my channel and then I've got, boom, I got all the videos from just my channel in here. Now, the difference between this and like Instagram or TikTok proper is that I can filter and cut this data however I want. So what I like to do is I'll go post it in the last, two months.
I like to set engagement rate at 2 % because that filters out the brand deals. Typically, if you have videos with super high views, but they have 0 % engagement, it's because people paid to boost the views. I don't want to look at the brand deals because that gives me a false sense of what topics worked.
So I just set this bottom engagement at 2%. Anything below that filters out. And then what I do is I go to sort and I'll go either outlier score or views, slightly different ways of calculating it, but let's just do outlier score.
So The outlier score, which is this green badge here, you can see like this one is 8 .2X. This means that the views of this video were 8 .2X higher or above the average views in the rolling three months before this video was posted.
We do rolling three months so that if the channel's dying, the outlier score kind of gets recalculated. You don't get punished for like super fast or super slow growth. But so an outlier score is like a better apples to apples comparison across multiple channels.
So the beauty of this is like, let's say I clear, uh, just this. And I look at the whole cohort.
Now this is in the last two months sorted by outlier score across all these channels, the top performing videos. So like, this is the, the wise investor. I don't even know maybe that maybe he's on the channel, but there's Gannon, uh, Peter visuals, et cetera.
So you go through. So now what you can do, there's a few things that are like super helpful. Let me, let me go by newest so I can show you this.
When you see these videos, the very first thing you're probably wanting to do is either play it in a new tab so you can just watch it. So you can just hover over and play a new tab like this.
Or you want to get deep analysis on this video. Like what is the transcript? How can I understand like what this video was about, right?
And so if you just press this analyze button, we will actually do a deep analysis on this video. And if I'll just go back to one that I've already analyzed. So like this one by Gannon, I can press into it.
You get the transcript. The caption, if you want it, is here. But then all this like deep analysis, right?
So what is the topic? What was the seed of the idea? The topic is like a three to five word kind of description.
This is more like a single sentence description. What is the unique angle of this video? What is the common belief versus contrarian reality?
This is like really where curiosity comes from right here. What kind of like core facts or evidence did he put in this video? Then we go to the hook.
What was the formula for the hook? What was the exact hook? What was the analysis of why this hooked work?
What is the category of the hook? Storytelling, same thing. Category, what the storytelling structure was.
Visuals as well. So the whole point is like, now you're getting all of the deep analysis on this video. We are both taking the transcript and literally watching it with a vision model and giving you insights on this.
Now, at an individual level, you could be like, oh, cool, like I'm just gonna, a lot of people just use sandcastles like this. They come in, they filter. They sort by outlier score and literally they just watch each one.
They're just like watch, watch, watch. And they like watching it themselves. They don't even want the detailed analysis.
This alone prevents you from having to scroll to find ideas. And now you can see the best performers. I'll like expand on what I would take from this verse not to keep it original, but like that's like the lightest way to do it.
The more advanced way is once you analyze a bunch of these and you can go one at a time or you can press bulk analyze like. Let's say you wanted to analyze my channel. You would just search my channel.
You could say bulk analyze 50 videos, or you could look and say, okay, it looks like there's 25 in the last three weeks. I'm just going to bulk analyze 25. Once you have all these analyzed, you can either export the data as a CSV and feed that into Claude, or you can use our MCP, which connects Sandcastle directly to Claude.
You can just install it in like 30 seconds. I can link a video if people want. And then in Claude, I can say.
Look at the last 30 videos from Callaway Marketing's channel. Make sure they're all deep analyzed. Give me a full report of like a breakdown of the topics, the hooks, et cetera, right?
So there's so much you can do with this data individually and as a cohort. So you're coming in here for Callaway Marketing and you're doing your, you're filtering, you're sorting, you have a short list by outlier score. When do you stop and say like, yep.
that's one I'm gonna do. Is it something that like inspiration hits or are you doing a minimum 10 and then saying, okay, which one of these did I feel best? What's your process from here?
So there's two paths. There's looking at what competitors are doing to get new ideas. And then there's looking at your own channel to understand all the other stuff that's working really well.
So like, which videos did I make that crushed? What hooks are working? What formats are working, right?
So I kind of like do both separately per se. So I've got a guy on my team that, literally just does the idea stuff all day long.
So he's looking at all of these competitor channels and he's like trying to, like he looks at this view right here and sometimes he'll go last, you know, 30 days. Sometimes we'll go further out. It depends on like for your niche and my niche, this is pretty evergreen.
So you can go back like four, six months and still get good ideas. For news -based niches where you need more real -time stuff, he'll do 14 days or seven days for tech. So he'll look at this and he'll open this video and he'll be like, okay, is the topic of this interesting to him, first of all?
Because I've kind of trained him on what interesting looks like. Just like, are you curious about this period? Okay, number two, do we have something relevant and non -obvious to say about this general topic?
So let's just look at this one right here. So we'll open this. So let's just watch this.
This video got nearly 25 million views. That's over 200 times more views than the creator's regular content. And the reason comes down to one thing.
Your brain is always trying to guess what happens next. So this is basically like a hook case study where he's explaining the hook or psychology of this, right? So my guy or you or anyone who's doing this, right, would be like, okay, this is a breakdown of an example video that had a viral hook that had traction that he explained.
how to do it. So the close clone, which I don't really like doing close clones would be like, let's take this exact same video and make like this exact same example and make our own spin on it where we break down the hook, but in a different way, we say different stuff. To me, that's still too close.
I want to be more unique. So I would say, okay, interesting. The style that he went through to explain this is he started with the hook itself of the example.
He played it without saying anything. Then he introduced what it was and broke it down. Could we run a similar flow like a similar story structure, but using a different example.
Do we have a different example? Do I have unique insights to say on a different example? That would like start the process.
So if my like head of content pulled this and he thought we would have something good, he would put it in the list for us to review. If he thought it was like too junk foodie, he would pass, right? But the reason why like that's helpful is all these are on target and they're the best performers by the people in my niche.
So it just focuses your yes, no process. super dialed. So he'll do this daily or every other day.
And he builds a Google doc that has tabs on the side and every tab is like a separate idea. He's only putting the ones on there that he thinks we have something good to say. Right.
And it's based on the whole corporate, like I've just trained him over time, like the whole corpus of everything. I believe he's seen all my YouTube videos. He knows kind of like our isms, let's say, then we get on a call.
I like to do it twice a week where he'll pitch me each one and be like, here's what I found this crushed. Here's what I think we could do. What do you think?
If I don't have something unique to say that's like contrarian and interesting, we kill it. If I do have something unique to say and I'm like, ooh, that's cool, we should use this other video as the example or ooh, we could bring in this one framework that I talk about and like, you know, I'll construct it. We green light it and we put it into like approved bucket.
We'll go through that call like we just did one today for an hour or so. And what that gives us is like eight to 10 approved ideas where they're validated by data. I know we have something interesting.
We went back and forth for five minutes on the call for each idea and like kind of brainstormed the best option. And then from there, I know I can go into the scripting phase. So that's kind of like the research flow.
Now he's most for short form, mostly he's using sandcastles. Like he doesn't really scroll that much for YouTube. He's looking at.
or like for tech ideas, he's looking at Twitter and other things like that, like email newsletters around tech. But for like Callaway marketing, the marketing ideas, he's pretty much just looking at this. So that's the, how do we get new ideas just on the topic front?
How are people doing this before there was a tool like this? Were they just like - Manually. Manually with like bookmarks and folders and spending a lot of time on their phone?
Yeah, like they either had a person like a social media strategist or them and they were scrolling for like two to three hours a day and they were having to go through all the noise, all the nonsense is not on target. And then they'd have to see all the ones that are like mid performers. And then finally they'd get to some bangers.
And the ratio from my research was like one out of 10 or one out of 20 would be like even close to potentially relevant. And then you'd have to see it. You'd bookmark it or you'd save the link into a notes file.
You'd have a bunch. And then you'd go through this process of like, is it relevant? Is it not?
So this just speeds up. a lot of the research time. So you don't have to spend all that time doing the finding, right?
Now we also have like this automations feature, which I think is pretty sick. You can set up these rules that will automatically deep analyze whichever videos you want based on the conditions, right? So what I'll do is we have this one on that says for any of the channels in our list, as long as there's a 5X outlier score on the video and it has above 2 % engagement.
We want you to auto process it. So it'll automatically be analyzed. And what that means is when it's analyzed, it's ready to be pulled from in cloud, right?
Cause we have the MCP set up. So we could just set up a daily scrape. That's like, give me the best performing videos of the day today, every day.
And we can set that up as a scheduled task. It can do that. I mean, it could, it could, it could just give you the video to like link out without analyzing it, but by analyzing it, we can also have it say.
what are the common topics that are working and it can just know because it's already done all this analysis for you automatically so this is just like one way to speed it up because you can tell like my whole thing is like how do we speed this up as fast as possible and get as much signal as we can in the workflow this clearly solves my topic ideation and even to a degree hook problem because i can see how this is working the next place where i feel like i would fall down is if the video is too long or in some way like the structure of the whole video goes off the rails is the answer to just use this analyze feature and look at the structure of these and make sure i'm following fairly closely to it and then just be relentless in the edit yeah so there's two ways so when you're when you're first starting and you don't have a ton of your own data like you don't have a ton of your own videos the best way would be find a creator that you like like let's say ganon
Find the creator where you like the flow of it's not like you like the look of his stuff because that's the visuals It's more like you like the way he explains stuff you like the flow of his storytelling You can use this transcript and you can either look just look at it and be like, okay Roughly speaking these the sections.
We also have a feature right here. It's show sections You just press this so it breaks it down into like the five parts or they're not always five But this in this case, there's five parts so you could then say okay cool Like if I'm gonna storytell, but I don't want to I don't want to copy his stuff We need a hook.
We need an intro to the series. Then in this case, he's doing a setup tutorial. There's a live demo and a payoff.
So like at the, at the roughest strokes, you could copy that structure at the most detailed strokes. You could take his exact script and then try to translate it into your own. Now you can do this manually, or of course you can take a hundred scripts from Gannon, put it into Claude and be like, I want to understand what the best performing structures are.
Storytelling wise from the scripts, make me a skill file template. And then when I jump in a topic, write a script for me in that way.
Now, I likely would not just go with that. I would take that and then I'd tweak it. That's going to be 90 % of the way there.
But that gets you, again, it puts guardrails on what you should do when you don't know. What I recommend doing, once you have a minimum amount of videos for yourself, like 10 to 20 videos, let's go to mine, I have all of my videos analyzed. So of course, I can look one by one.
If I sort by outlier score, my eye is trained on content now where I can see let me go 60 days. So we have all of them in there.
I can see like, okay, this was about kind of content strategy. This was about AI tools. This was about storytelling.
This was about formats. This was about strategy, strategy. This one I think is storytelling, storytelling.
So I'm starting to see like common buckets. Like I can just look at it and see, but on the, on the story structure side, the storytelling, like with the script, you could just take these. And once they're analyzed, you can, again, like export the data or use Claude and just say.
hey, I've made 20 videos. Can you look at the top five performers and just help me understand like what is the common structure or flow of the script that's working well for me? Let's turn that into a template or turn that into a skill file.
And then when I feed you a new topic, can you write the script or at least guide me into what those buckets should be? To your point, like that'll speed up a lot of the confusion. I still think you need, like writing is the number one.
creative tool. Like I believe writing is the most important skill. I think you believe that as well.
So I never want to like fully outsource every word I say to this, but like as a 75, 85, 90 % of the way there first draft automatic, it's pretty sick. And then you can tweak from there. Yeah, I agree.
The conflicting feelings that I get when I think about processes like this is if you remove all of your own creative work and writing from it, which you totally could. Like I could make this totally data -driven and totally like automated from this.
And it would probably get results. And I just know that would be like not satisfying for me either. Like what I like about these analyzed tools and these headers that you have in this transcript, I can teach myself to fish with this tool.
I don't have to let the tool fish for me. Exactly. The way I like to think of it is, so in the content workflow, you've got topic, you've got research or like substance, you've got hook, script, record, edit, post.
In each of those little modules within the flow, you can decide to turn it all the way to 100 on data or all the way to zero on manual and you can pick where you want. So like for me, where I'm at right now is I really like using 100 % data to guide me on the topic buckets so that I know where to start fishing. Then I look at my own data, like the most valuable thing, like just on YouTube, the most valuable data for you on YouTube is your own channel.
Same thing here. But until you have your own data, you just guiding on the topics that make the most sense from other people is the best way to do it. For substance, the contrarian insight, I'm relying fully on my creativity.
I'm not really like taking other people's insights because I find that's really like the sauce that makes it unique. And that's the creative challenge for me is like. okay, how do I take this 30 days of content topic and make it unique?
So I came up with this new mechanism and like that did pretty well. So on the substance, I'm doing it full creativity on the hooks and the scripting. What exactly what you said is where I'm at, which is like, the goal is to make your brain full of what works so that when you flow naturally creatively, it's already infused in there.
Right? So like you just, when I look at this, I'm like, okay, let me study right here. Let me study what this is.
When I make another one, I don't copy paste. I just, read it just like copy work or like any form of training yourself on a skill.
I'm reading it and I'm like, Oh, okay. I see like how that flowed and that didn't. Okay.
Now I know. So when I go to write from scratch, it's kind of like in my brain already, the way of doing this, that's the like most manual way to use data, but still be fully creative. You can of course dial that a little bit more where it's like, cool.
I'm going to start copy pasting this one that worked. It's my own words, but like, I'm going to start here and then edit around it. The full.
Automated would be like literally just rip this and change a couple words So to your point you can season to taste on this depending on how much creative expression you want I don't think you can fully build an ai factory pumping it out with zero creativity and cut through I just think there's too much supply and it's going to trend towards average Because most people are doing that and they're coming up with the same stuff.
So this is where like I call it data enabled creativity. That's my like big theme or thesis that I believe in is like using the data to guide you and then allowing your, basically freeing up all of your time to be creative and using data for everything else.
What if I took one of your, one of these videos that we're looking at on screen right now, and I removed all visuals, the cuts stayed there, but it was just you and it was just captions. How big of an impact do you think that would make on the performance of that video? A pure yap with zero graphics, which is kind of what you're describing, right?
Like full A roll U and captions. basically the way the framework I use to look at this is like , right? So if you have no visuals on the screen, you're relying on the viewer's imagination to come up with, if they're a visual learner like I am, they have to imagine what you're saying and try to get comprehension from that.
The more technical or the more complex or the more like out there ideas you're talking about, the harder it is for people to imagine. It's a spectrum, right? It's like the more work you, the more front end effort you apply to the production, The easier it is to carve a moat visually because there's just less people putting in that time.
However, to your point, the more resource intensive it is. So you have to just experiment along that spectrum and figure out what you are happy with. And then ultimately what converts into emails.
That's a really critical thing too. Amazing. Well, I will link Sandcastles in the description for this.
That's fantastic. If you're a business owner, hopefully this is helpful to you. Definitely free stuff below.
So feel free to check it out. now that you've seen exactly how calloway structures his content research you'll love seeing behind the scenes with riley brown too watch this interview next
The Hook
The bait, then the rug-pull.
Kallaway has racked up over a billion views and a 50% hit rate on his short-form videos, and he says the secret has almost nothing to do with views. In this conversation with Jay Clouse, he lays out the exact research system, and the software he built to run it, before he ever hits record.
How they asked for the click.
MENTIONED ON CAMERA
