We Had 6 Brands Produce 500 Ads Last Month. 83% Shouldn't Have Been Made

If you're anything like us, you're starting to hate the ad game. DTC brands spend 90% of their week producing something that's almost guaranteed to fail (most winning creatives emerge from the top 10% of tests, which means 90% of what we build has failed even before we launch it.)

If something does take off, we face even harder problems:

"What happens when they fatigue?"

"Can my team keep this pace up?"

"Should we add more AI?"

"Should we cut all our AI?"

"Why is this so hard?!"

We're just gonna say it out loud: the ad game has gotten much harder, and that sucks. 😔

Building ads that work feels impossible, and now that AI has joined the game, we're producing our tails off...but still seeing "meh" results.

We're tired. We're burnt to a crisp. We've got data coming out of our ears, and everyone keeps telling us "you're just not producing enough. Just make more ads."

At Tether, we're not sure "make more ads" is the solution.

In fact, we don't think you need to produce more ads. We think brands need to produce more accurate ads.

This is the ultimate guide to why most ad concepts fail before they even launch, and what you can do to hop off that damn volume mouse wheel and start building ads that scale. You can use it to save your budget, protect your creative team from burnout, or just see how much money you're leaving on the table by making everything that gets suggested.

Roadmap & Read Time

• Time to Read: ~7 minutes

• What's Inside: The real breakdown of a sample of 1,200 real-time customer insights we analyzed last month, and why we only made 200 of them in ads. Plus, the exact filtering system we use to separate winners from waste.

• The Goal: By the end, you'll know how to stop burning money on ads that never had a chance and finally get off the mouse wheel of endless production.

The Big Picture

6 DTC brands. One month. 500 ads produced and launched.

To get those 500 ads, we started with nearly 1,200 raw insights pulled from real customer conversations. Reviews, Reddit threads, TikTok comments, support tickets...the unfiltered stuff. The things people say when they don't think a brand is listening. Sometimes brutal. Occasionally unhinged. Always honest.

Then we started with something most brands completely skip: we filtered them before we produced them.

The result: only about 200 insights (17%) were actually worth making into ads. That means we killed ~1,000 insights (83%) before they ever saw a production budget.

Those 200 insights became the 500 ads we actually produced (some insights spawned multiple executions with different hooks, creators, or formats.)

And yes, we know most brands would have just made all 1,200 of them and called it "testing."

But here's why this matters: If we hadn't filtered and instead produced ads from all 1,200 insights, we'd be looking at 3,000+ ads and somewhere between $1.5M-$3M in wasted production spend.

We made 500 ads for $250K-$500K instead. Same winners. Way less waste.

Let's break down how:

What Actually Happened to the 500 Ads We Made

Okay, so we filtered down to 200 insights, made 500 ads, and saved a boatload of money on concepts that would've flopped.

Here's what happened when those 500 ads went live:

~300-350 ads (60-70%) still flopped in testing. They didn't hit ROAS targets. Audience didn't engage. Creative fatigue hit in days instead of weeks. We killed them fast and moved on.

~100-125 ads (20-25%) broke even or underperformed. Not terrible. Not great. They covered their production costs but didn't scale. We kept a few running at low budgets for retargeting. The rest got archived.

~50-75 ads (10-15%) actually scaled profitably. These were the ones. Strong hook. Clear message. Kept performing for 2-3 weeks (or sometimes longer) before fatigue. Some we refreshed with new creators and get another run. These paid for everything else and then some.

We're not here to tell you that 99% of our ads now hit every single time, because that's ridiculous. No one has a perfect 100% hit rate. But here's the part that matters...

That 10-15% win rate is 1.5-2x better than the industry standard of 5-10%.

Even with our filtering, we still had a 60-70% failure rate. That's just how creative testing works in DTC right now. Most things don't hit.

But the biggest difference is this:

Without filtering, we would have made 3,000+ ads from all 1,200 insights and still only found 50-75 winners. The win rate doesn't magically get better when you make more bad concepts. You just spend more money finding the same winners buried in a bigger pile of garbage.

We would have spent $1.5M-$3M to get the same 50-75 winning ads we got for $250K-$500K.

That's the entire point of filtering: You can't improve your hit rate on concepts that were never going to work. But you can stop wasting budget making them in the first place.

Why Most Brands Get This Wrong

Most brands think their creative problem is coming up with ideas.

So they hire a strategist and buy 5 AI tools that look at past performers, then spit out 500 concepts in 10 minutes. They look at what competitors are doing. They figure if something worked for another brand three months ago, it'll work for them now. (PSA: it won't.) They brainstorm until they have 100+ ideas in a Notion doc that nobody will ever look at again, and they feel good at the end of the day because they did something.

Then they hand everything to their creative team and say: "make all of these."

And that's where everything falls apart.

When you're making 50-100 ads per week, you burn through money, time, and mental capacity fast. Really fast. When you break down the math, we're talking $500-$1,000 per concept to produce and test. And the painful part is: 70%-80% of these ads never make it past testing. Your team is burning themselves into the ground making ads that were dead before they even hit Meta's algorithm.

It turns out the real problem isn't coming up with ideas. We've got tools for that now.

Right now our problem is knowing which ideas are actually worth making. Which is way less exciting than "we have AI that makes unlimited concepts"...but way more valuable if you like keeping your budget in tact.

The Numbers Don't Lie

Here's what happened with our 1,200 insights:

• ~360 were non-starters (30%) - Bad ideas. Already overdone. Off-brand. Just shouldn't be made.

• ~640 were possibles (53%) - Interesting ideas that need more work. Maybe they'll be good later. Not ready now.

• ~200 were immediate greenlights (17%) - Ready to make right now. (Go produce these before competitors figure it out.)

The math here is brutal but it should tell you something valuable: For every 1 insight worth making right away, there are 6 others that shouldn't be touched yet. Maybe not ever.

That 83% is the AI slop you need to filter out before you waste a single dollar.

And here's what makes it scary (to us, and hopefully to you as well): most of that 83% really does sound like a good idea when you see it. It feels like it could work. It looks actionable. But that's exactly what makes it so risky.

Only 1-in-6 insights that actually matter. Only 1-in-6 ever make a really good ad. Everything else is noise or busy work pretending to be strategy.

But there's something else you should know...

What It Costs When You Don't Filter

Let's be real about what happens if you skip filtering and just make everything into an ad.

Let's say you're starting with 1,200 of the best insights you've got. Reviews, customer feedback, call transcriptions, social media scrapes, what-have-you.

If You Make Everything

• 1,200 insights × $500-$1,000 (est.) to produce and test each concept = $600,000-$1,200,000

• 83% of those concepts shouldn't have been made

• Money wasted on bad concepts: $500,000-$1,000,000

• Your creative team is completely burned out

• You convinced yourself this is required to "learn what works"

If You Filter First

• 200 insights × $500-$1,000 (est.) to produce and test each concept = $100,000-200,000

• Way higher hit rate on the ones you do make

• Money saved by not making garbage: $500,000-$1,000,000

• Your team focuses on work that actually matters

• You still have a job

The difference isn't just what it costs to think it up, develop the idea, send it to production, move it through the pipeline and get it in the ad account. Maintaining stability in an ad account is all about momentum. It's about morale. It's about whether your creative team still believes in what they're doing or if they've secretly started updating their LinkedIn.

When your team makes 1,200 ads per month and 1,000 of them bomb, morale doesn't just drop, it disappears. When they make 200 and most of them work, they start believing in the process. They stop sending annoyed Slack messages about "maybe we should have a strategy", and start being more creative and more aligned with the brand again.

Why That 83% Actually Matters

Here's what everyone knows: coming up with ideas is really easy.

Here's what no one knows: the new ad game comes down to taste...not skill.

Everyone has ChatGPT to brainstorm now. Everyone can look at what competitors are running (which is a great way to make ads that looked fresh two months ago and are now completely played out.) Everyone has tools that scrape reviews. Everyone has AI that can make 100 "different" concepts in 10 minutes (most of them are really just the same three ideas wearing different hats).

What separates behemoth brands from everyone else isn't how many ideas they generate. It's knowing which ones to kill.

Knowing which ideas to kill and actually having the guts to kill them right at the moment when everyone's excited about them requires:

• Deep understanding of your brand and what you won't compromise on

• Really knowing your audience beyond basic demographics

• Recognizing when something's been done to death

• Predicting what will actually make people stop scrolling

• Balancing being original with not being so weird that nobody gets it

This is why the "creative strategist" is the hardest job to fill in DTC right now. You can teach someone Canva. You can show them your brand guidelines. You can train them to brief creators. You can even train them to understand data analytics, videography, storytelling, and DR frameworks.

But you can't easily teach taste.

You can't shortcut the experience of watching a thousand ads fail to build the instinct for what will work. You can't automate the judgment between "this is original" and "this is so original nobody will understand what we're selling."

AI can make 1,000 concepts in an hour. But it can't tell you which 830 will waste your money. That takes taste.

Experience. Judgment. And confidence to say "no, we're not making that" when someone really wants to.

That's what the 83% means: it's gap between having ideas and having judgment.

Why Copying Past Winners Is a Trap

Quick thing while we're here: if your whole strategy is looking at what worked for other brands three months (or even three years) ago, you're not behind at all. In fact, you're doing exactly what everyone else is doing, because it works...for a minute.

But eventually, you'll need another ad again. Which means your team will need to produce again, which means you're back to burnout...again.

By the time an ad shows up in someone's ad library (or is scraped by a bot), it's been running long enough to get noticed. Which means your audience has seen it dozens of times already. The freshness that made it work at first is gone. You're copying yesterday's winning lottery numbers and hoping they hit again.

The most efficient brands right now aren't studying what worked last quarter. They're listening to what their customers are saying this week and moving fast enough to be first with concepts competitors haven't noticed yet.

That's why we look at 1,200 insights from real customer conversations before we start producing instead of just looking at ad libraries and hoping.

How to Filter Your Own Insights

If you're drowning in ideas and burning budget on concepts that fail, here's our exact system for getting off that train:

Step 1: Gather Everything You've Got

• Scrape customer conversations: reviews, Reddit, TikTok comments, support tickets, anywhere customers talk when they think you're not listening.

• Pull them into one spot.

• Use AI to find the underlying psychological patterns that pop up.

• Brainstorm with your team without shooting ideas down immediately.

• Goal: Get to several hundred raw insights. You're about to kill most of it anyway.

Step 2: Kill the Non-Starters (30%)

For each insight, ask yourself: Is this...

• Already overdone in the market?

• Off-brand or bad for positioning?

• Too small to matter (under 10% of audience)?

• Risky in a way that could blow up?

If you answer yes to 2 or more of these things, kill it. Don't save it "just in case." Just in case is expensive.

Step 3: Separate Ready from Not Ready (Keep 17%)

For each remaining insight, ask:

• Is this original enough? Or is it just a slight twist on what everyone's doing?

• Is it specific enough? Or just vague inspiration that needs six rounds of direction?

• Is it big enough? Or just a cute idea that won't move numbers?

• Is it proven? Is it a real widespread problem or just three comments and a feeling?

Only greenlight ones that check ALL the boxes. Not most. ALL.

Step 4: Make Only the Greenlights

• Focus your whole team on the 1-in-6 that matter

• Focus on higher quality ideas because you're not rushing through 100 concepts

• Focus on better test results because you're only testing strong ideas

• Focus on getting your team less burnout because they're making work they believe in

Step 5: Check the "Possibles" Monthly

• Some will become greenlights as you get more data

• Most will stay in the "possible" pile forever

• That's fine. The point is protecting your budget from mediocre ideas

The Bottom Line

In 2026, every brand has tools that make unlimited ads. ChatGPT will brainstorm all day for you. AI will scrape and pattern-match until you're drowning in possibilities. Ad libraries will show you what's working (or what was working weeks ago).

The advantage isn't in generation anymore. It's in curation. It's having the judgment to kill 83% of what gets made.

The most efficient brands right now aren't making the most ads. They're making the RIGHT ads: the 17% that are actually worth making, that haven't been done to death, that solve real problems with fresh angles.

Nobody wants to admit this, but we will: when you make everything, nothing is memorable.

When you filter hard and only greenlight insights that truly matter (the ones that check every box, the ones you'd bet on) you:

• Save 80%+ of your budget (and your CFO will be way happier)

• Protect your team from burnout (which matters if you want to keep them)

• Get better test results (because you're only testing strong concepts)

• Build a brand that stands out instead of blending into AI-generated "differentiation".

The market is about to flood with AI-generated sameness. Every brand has the same tools, same ad libraries, same AI prompts. Platforms are getting noisier. Creative fatigue is speeding up. What worked last month barely works this month.

The brands that survive won't be the ones making 1,200 concepts.

They'll be the ones with the guts to kill 83% of them.

Curation. Judgment. Taste.

The ability to look at a thousand ideas and know—really know—which ones are worth making and which ones are expensive distractions dressed up as opportunities. That's a win.