Content ideas: the list was never the problem

What an idea list actually gives you

Worth being fair to it. A good list is genuinely useful for one thing: it widens the space of what you could talk about. If you are stuck, seeing forty angles is better than staring at a blank page.

What it gives you is a set of **arguments**. "Behind the scenes", "the mistake I made last year", "three ways to use this" - each one is a claim about what will interest your audience.

What it does not give you is any way to tell which of them is right, and no way to find out cheaply. The list is a catalogue of hypotheses with no experiment attached.

So the useful question is not "which of these 50 is the best idea". It is **"how do I turn these into attempts that tell me something?"**

How to use an idea list well

None of this means throw the list away. It means use it for the half it is good at.

The list is a **pool of candidate angles**. What it cannot do is tell you which one is right - so do not treat picking from it as a decision. Treat it as the input to a test.

The version that works:

That last step is where most people quit early. They find one thing that works and then keep looking for a better idea, when the leverage has already moved to the wrapper.

• Read the list and pull out the three or four angles that could plausibly fit your product. Not the ones you like - the ones a buyer might recognise as their own problem.
• For each one, find a video that is already running with that argument. Now you have an angle that was not merely guessed, plus a structure to rebuild from.
• Run the angles against each other, holding execution steady.
• Once an angle wins, stop changing angles and start varying the execution.

Twelve angle archetypes - as candidates, not answers

If you came here wanting a list, here is a better one than a list of topics. These are the **shapes of argument** that recur in short-form commerce video. They are not ideas to copy - they are the containers you pour your own subject into.

**Twelve candidate arguments, not fifty topics.** And here is the point: even twelve is too many to run blind. Pick three. Find a video already working with each argument, and let them compete with the execution held steady - that is the next section.

Why picking a better idea doesn't fix it

Say you take five ideas from the list and make five videos. Each one changed the topic, the hook, the format, and the call to action. One of them does well.

Now try to say why.

[A table] What you changed | What you learn when one wins Topic, hook, format and CTA all at once | Nothing you can reuse. Something worked; you cannot name it. One hook, everything else held steady | "This hook works." That is a thing you can do again.

This is not a subtle point - it is the whole reason creative testing has a method at all. Testing guidance is blunt about it: if the creator, the angle, the hook, the format and the CTA all change at once, you cannot say why the winner won. **Change fewer things, and the winner tells you what to repeat.**

A list of fifty ideas pushes you toward the opposite. Fifty ideas means fifty attempts, each one changing every variable at once. You end up with a pile of results and no knowledge.

**And there is a harder version of this problem.** A 2026 study tracked 56,419 YouTube videos across their complete view histories, including channels that never took off, and measured what actually goes into each one: 3,728 thumbnails, 385 hours of speech, titles and tags. Thirty-six measures of image, speech and text **explained at most 1.1% of the differences in views within a channel.** The features could tell you what kind of video it was - they could not tell you whether it would work.

Sit with that for a moment. **You are being asked to choose between fifty ideas on the basis of the one thing that does not predict the outcome.**

The odds, honestly

It is worth knowing what you are up against before deciding how to spend your effort.

The working number in media-buying circles is **roughly one profitable creative for every 10 to 20 you test**, and one that keeps working through weeks of scaled spend for every **40 to 60**. Treat that as a ceiling rather than an entitlement - accounts with thin budgets, unproven angles, or a single creator's voice usually see worse, and **a run of thirty losers before one hit does not mean the process failed.**

⚠️ Be careful with that figure. The people who publish it say plainly that it is practitioner reporting passed around for years, not a measured result, and that no dataset they have access to can confirm it precisely. There is no equivalent published number for organic creator content at all. **What transfers is the structure of the problem, not the exact ratio.**

And the structure is this: if you need somewhere between ten and sixty attempts to find something that holds, then **the number of attempts you can afford is the real constraint.** Not the quality of idea number 17.

**Two more findings from the same study, and they point in opposite directions.**

The discouraging one: **practice does not help by itself.** With all the content measures held, the study found that forty more videos without a hit add nothing. Nothing accumulates except a hit. So "just post more" is not the answer either - volume without a validated starting point buys you nothing.

The encouraging one: **a hit is worth far more than the video it appears in.** A channel's first hit raised the views of everything it published afterwards by **164% on average**, a result the authors put through seventeen robustness checks. Which means the goal is not to make each video better. It is to get to a validated structure, once - and then everything downstream of it inherits the lift.

One caveat on the study itself: it is a single-author preprint on arXiv, posted in September 2026, and it has not been peer reviewed. The sample is large and the findings survived the authors' own robustness checks, but treat it as strong evidence rather than a settled result - and note that the win-rate figure above comes from practitioner reporting rather than measurement.

An angle and a variation are not the same thing

This distinction decides where you spend your effort.

**A variation changes the wrapper.** A different thumbnail, a swapped colour, a new caption on the same underlying argument.

**An angle changes the argument itself** - the pain point named, the mechanism claimed, the promise made.

And the priority is not obvious: **ten wrappers around one weak argument will not outperform three wrappers around three strong arguments.** Polishing the execution of a bad premise is the most common way to waste a month.

Now look at that idea list again. Every item on it is an **angle**, and a completely unvalidated one - a guess about what your audience cares about. You are being asked to invent the hardest part from scratch, fifty times.

Start from an argument that already ran

If angles are the scarce thing, and inventing them is a guess, there is a shortcut: **take one that has already been tested on an audience.**

You do not have to guess whether a video's argument works. If it is circulating in your niche - views, saves, comments, people remaking it - then an audience has already voted. That is not proof it will work for you, but it is a far better starting point than item 17 on a list written for every industry at once.

So the move is: find one video that is clearly working in your space, and **take the argument, not the content.** Keep the order of the argument, the shape of the hook, where the proof lands. Replace the product, the words, the footage and the voice with yours.

**And there is hard evidence for which part to take.** The same study looked at what creators do after a hit: nearly all of them copy it, both its look and its topic. **Only copying the topic paid. Copying the look added nothing measurable.**

That is the whole argument for taking the argument. Not the edit style, not the colour grade, not the pacing tricks - the thing the video was actually claiming, and the order it claimed it in. Which is, precisely, what a rebuild is for.

**What we do not do: we will not find that video for you, and we do not hand you a list of what is trending.** Finding it is the easy part - you already know your niche better than any tool does. What is hard is everything after that, and that is the part BrandU does.

How to tell a video is actually working

This is the part we do not do for you, so it is worth being concrete about it.

**Judge it on the argument, not the view count.** A video with ten million views from three years ago tells you the platform pushed it. A video with two hundred thousand views from last month, in your category, tells you an audience like yours responded to that specific argument.

Signals that are worth more than raw views:

**What you cannot see:** whether it sold anything. Public signals are engagement, not revenue, and no external tool can show you someone else's conversion numbers. Treat every read as a hypothesis to test on your own account - which is what the rest of this page is for.

• **It is recent.** Formats and hooks age. A four-year-old hit is a history lesson, not a template.
• **The comments restate the point.** When people argue with the argument, or repeat it back, the message landed. When they only comment on the creator, it did not.
• **Other accounts are remaking it.** Imitation is the clearest public signal that the structure travels independently of the person who made it - which is exactly what you need, because you are taking the structure, not the person.
• **The audience resembles your buyer.** A format that works on a teenage audience may say nothing about a 40-year-old deciding on a purchase.

Does the platform change the answer?

This page has to serve people asking the same question about YouTube Shorts, TikTok and Instagram Reels. The honest answer: **the situation is identical, and only the signal differs.**

The question is always the same - what do I post, and how do I know it is any good. The thing that changes is how you can tell a video is already working, because each platform publishes different evidence:

**What does not change:** the failure mode. Filling your calendar with a different premise every day is the same mistake on all three platforms, and it produces the same result - a pile of results and no knowledge.

• **On TikTok**, the clearest public signal is **remakes**. When the same premise shows up from account after account, the argument is travelling on its own - which is exactly the property you are looking for, since you are taking the argument and not the account.
• **On YouTube Shorts**, the useful signal is **whether the same channel keeps returning to a premise.** A Shorts channel that reuses a format across uploads is telling you it has evidence you cannot see.
• **On Instagram Reels**, look at **saves and shares relative to views**, not views. A Reel with modest views but heavy saves has an argument people wanted to keep - the number of views mostly measures how hard the platform pushed it.

One variable at a time, and you already have five

Take the argument as fixed - it came from a video that was already working. Now vary the execution, one dimension at a time.

BrandU gives you five replaceable dimensions, and they map onto a test matrix directly:

**But do not treat the five as equal.** The evidence above says the argument is what pays and the wrapper is not - so the hooks and the product (the claim and the offer) are where the real information is, and the presenter, language and frame are the cheaper checks you make later.

Start at the argument end. Pick one. Hold the other four. Run it.

**Why this beats the list:** every attempt now isolates something. After a month you do not have "some videos did well" - you have a list of things you know about your audience, each one paid for once. That is the difference between forty videos that added nothing and forty videos that told you something.

The volume problem, and why it stops being one

If you need ten to sixty attempts to find something that holds, the binding constraint is production. That is where the traditional answer breaks down: making sixty videos by hand is not a plan, and generating sixty from scratch is a different kind of expensive - each one is a fresh roll of the dice, and you cannot tell what changed.

BrandU generates a video as independent segments, so cost follows the change rather than the video:

And because a variation is mostly a re-render, the hundredth attempt does not cost what the first one did. **A run generates up to 10 variants at a time, and you can run it again for the next 10** - the marginal cost does not climb as you go.

• Rewriting one segment's line re-runs that segment and its voiceover. The other segments are kept.
• Restyling, reordering or swapping an asset only re-renders - the model is not called again.
• Change the hook, a line, the language or a component, and the timeline and related elements re-time automatically.

What you keep is not a video - it is a way of making them

Here is the difference between this and a good week of posting.

If you spend the week making five videos from five ideas off a list, at the end you own five videos. Next Monday you are back at the list, because nothing you produced carries forward. That is what a one-off looks like - and it is why the list keeps being necessary.

A rebuild leaves you with something else: **a working structure, in your own product and voice, that you can keep editing and keep reusing.** The argument is settled, so next week's work is not "what do I talk about" - it is "which part of this proven thing do I vary now".

That is the shift worth making. Not from bad ideas to good ideas, but from **starting over every week to compounding every week.** A hit that took someone else an experiment to find becomes an asset you can run again - and per the study above, a validated structure lifts everything you publish downstream of it.

The line: keep the structure, replace the content

Copying someone's video wholesale is not the move - it does not work, for the obvious reason that it sells their thing, and it usually reads as a copy.

What transfers is the structure. BrandU's rebuild is explicit about which is which:

So the thing you keep is the part that was never really theirs - the order of an argument. The thing you replace is everything that made it theirs.

What BrandU does not do

• **It does not tell you what to post.** No trend feed, no idea generator, no viral library. Finding a video that is working in your niche is your job, and you are better at it than a tool is.
• **It does not guarantee the video performs.** Nothing does. The odds section above is the honest version of this.
• **It does not invent your angle.** It rebuilds one that has already been tested.
• **It does not measure results.** Producing variants is not measuring them.

Three steps

• **Find one video that is clearly working in your space.** Not the most viewed video ever - one that is working now, for an audience that looks like yours.
• **Paste the link.** A page URL is fine; you do not need the file.
• **Vary one dimension at a time.** Start with the hook. Keep a note of what you changed, so the result means something.