- Grow 📧 Monetize
- Posts
- 📩 Most ads die in three days
Most of these ads are supposed to die
Of every ad one paid growth team launches, 10 to 20% survive.
Twenty-five percent is what they call crushing it.
This team has bought more than 5 million newsletter subscribers over six years of media buying. They run Meta campaigns for B2C authors and for B2B newsletters as niche as one that goes only to HVAC owners and managers.
And their whole operation is built on one idea: the job is not finding a winning ad. The job is killing losers fast enough that the winners can pay for them.
Here is what that looks like in practice, from the first test budget to the sign-up form.
In partnership with
Join Fin, Anthropic, and Clay at Pioneer on October 7th
Pioneer, the summit where CX leaders redefine what’s possible, is on October 7th.
Join leaders from Fin, Anthropic, and Clay for an insightful conversation on the state of AI transformation.
You’ll discover how some of the most innovative minds in CX have transformed their organizations, learn how they think about CX, and hear how they're planning for what's next.
Join the conversation in San Francisco, or tune in virtually.
*Ad
Your first $1,000 tells you nothing
A $1,000 test almost always looks great.
The CPAs come in low, the data looks like it is crushing, and you decide it is time to scale.
So month two you spend $5K. Month three, $10K. And the numbers look nothing like what you tested.
That is why this buyer will not take on a client spending less than $2,000 to $3,000 in month one.
It is not about the fee. A thousand dollars just does not buy enough data to tell you what happens at real spend.
What they want is $3,000 to $5,000 in month one. At that sample size, they are confident the CPA holds when spend moves to $8K to $10K a month, and from there they scale budgets 20 to 30% at a time.
If you only have $1,000, the advice is to skip the agency and run it yourself. Just do not treat what you learn as a forecast.
A cheap test is not a small version of a real campaign. It is a different campaign.
Volume is the only math that works
If you launch two ads a week and hope both work, the math is against you before you start. When neither hits, you have burned a week or two of spend and learned almost nothing.
So this team produces dozens, sometimes hundreds, of new ads a week across clients.
More than 80% of spend goes to video, including plain text over a moving background, because it runs in the Reels placement.
Statics win in bursts, but they have never seen one hold the top spot for longer than three to six months.
The winners are often ugly. They do not fight it. Drifting toward something prettier just means the client pays for the experiment.
Volume is also why their UGC has moved to AI.
Think about the trade. $250 buys you one creator, one video, and your fingers crossed. The same script across a dozen AI actors buys you a dozen shots at a winner. They ask clients first, and some say no, which is fine.
Founder ads still work, just differently. The writer records each line a few times, the team stitches the best takes together, and nobody has to memorize a script or read it into the lens.
Then every ad gets the same short leash.
Sometimes they can tell within 48 hours. The hard max is three to five days, judged on where Meta is putting the spend, click-through rate, and what that works out to in CPA.
When 80% of what you make fails, making more is the only way the numbers close.
See what's really happening in every deal
Aligned shows you what buyers are doing between meetings inside your most important deals then surfaces the risks, openings, and next steps. No more guessing where a deal stands or finding out a champion went quiet too late. Just real visibility, so you can act before it's too late.
*Ad
The speed limit on a winner
Once an ad wins, the question is how fast you can push it. The answer is slower than you want.
"We're never scaling it by more than 15 to 20% daily. That's just a cardinal rule."
Go faster and you reset Meta's learning phase. The algorithm starts over every day, and the CPA you were scaling disappears.
So they stack four to six of those increases, roughly doubling spend over two weeks, then let it settle and watch whether the CPA holds. If it keeps climbing, the extra budget goes to other winners instead.
The account structure behind it is simple.
The main campaign holds 80 to 90% of the budget, all on proven winners.
The testing campaign gets the rest, at $40 to $50 a day per test. Go lower and you will not have enough conversions to call a winner by day three.
Here is the part most people miss.
When a winner graduates into the main campaign, Meta treats it like a brand new ad, and the CPA you saw in testing may not carry over. So they leave the test version running, watch both for three to five days, and back whichever one holds.
Usually that is the main campaign. Not always.
Meanwhile the team is already building a V2 of the winner, same format and hook, so something is ready when the original fades.
"Most of the testing and creative production that we do is to be prepared for when things do go south."
Scale slower than you want to. Push too hard and the algorithm starts over.
Why one ad lasted two years
Their longest-running winner has been live for more than two years.
That sounds like a creative miracle. It is not.
The ad collected thousands of likes and comments along the way. At this point it looks like a popular post, not an ad, and that social proof keeps feeding its performance.
That only works with a huge audience, though.
On B2B, they have never had an ad last past four to six months. The audience is small enough that the same people see it over and over, and returns drop.
On B2C, ads can run 6 to 12 months or longer. Once one makes it past six to eight months, it will likely run another six.
Think about the difference in scale. A bestselling self-help author has a potential audience in the tens of millions. That HVAC newsletter has maybe 40,000 to 60,000 ideal readers in the entire US.
Same platform, completely different expectations.
Your audience size decides how long an ad lasts before you write a word of copy.
Invest in the Future of Popcorn
Popsmith is reimagining popcorn for a new generation. With 100,000+ customers, $20M+ in lifetime revenue, and sales projected to reach $9M in 2026, the premium popcorn brand is expanding nationwide and raising capital for its next stage of growth.
*Ad
The cheapest ad set was buying the wrong people
Here is a question most publishers cannot answer: of the people Facebook brought you last month, how many are actually your reader?
CPA cannot tell you. It tells you what a subscriber cost, not who you bought.
That is why this team treats first-party data as the bare minimum on B2B: a survey right after sign-up, tied to the UTM that says which ad brought each subscriber in.
It matters on B2C too. "60% of my list earns over $150,000" is one data point, and it changes a sponsor conversation.
What has changed is how fast you can use it. They plug AI into beehiiv's MCP and ask things like: of the subscribers who said they are HR C-suite, which came from Facebook, and how does each ad set perform?
Every ad set gets scored on clicks, opens, engagement, and survey answers together. Fourth or fifth place gets its budget cut. A strong one gets 15 to 20% more for two weeks, and then they ask again.
That analysis used to take half a day to a full day per client. Now it is about ten minutes, every client, every week.
I put a case to them I had been thinking through. You are targeting senior employees, and one ad is pulling mostly juniors at a cheap CPA.
They had just run into exactly that. Within two to three days, the survey data showed more than half the spend going to the wrong ad and ad set. They cut it, even though the ads left running cost more per subscriber.
The cheap one was never cheap. It was just buying the wrong people.
CPA tells you what a subscriber cost. First-party data tells you whether you bought the right one.
Lead forms hand you the email from 2001
I asked for a straight yes or no on Facebook lead forms.
No.
The problem is the autofill. The form drops in whatever email the person used to open their Facebook or Instagram account years ago.
"I signed up to this newsletter with an email that I created in 2001 when I made the Facebook account."
There is no friction, so the conversion rate looks incredible: 70 to 80%, against 40 to 60% for a landing page. Your CPA drops. And you are paying for addresses nobody checks.
To be fair, maybe 20 to 25% of established media operators use lead forms, and it may be working for them. The one exception worth making is a call center or SMS team working the leads, where the phone number is probably real.
I watched this play out last month.
A publisher's welcome emails were opening at around 5%. I kept digging into deliverability until I found the lead forms. We sent the same Facebook traffic to a landing page instead, and new sign-ups started opening around 45%.
You know how I feel about open rates, so read that jump for what it is. It is not proof of engagement. It is proof the emails started getting through.
A welcome email opening at 5% is landing somewhere nobody looks. The auto-filled addresses were abandoned inboxes, dragging down sender reputation for everyone who signed up alongside them.
The number that proves those new subscribers are people is the click.
This team has 8 to 10 versions of the same story from the past year. Switch to a landing page, opens go up, CPAs barely move.
The creative was never the problem. The email addresses were.
A 70% conversion rate is a cheap CPA on addresses nobody reads.
The sub-$2 subscriber is a myth
The biggest myth in paid newsletter growth, according to this buyer: that any newsletter can get under $1 or $2 a subscriber.
"Sometimes the lowest is not the most efficient."
Efficient depends on what a subscriber is worth to you, and they work that backwards from the one or two income streams that actually pay. Most newsletters running three or four streams get just 10 to 20% of revenue from the rest.
For a subscriber who engages and converts, efficient might mean $2 to $4, or $4 to $5 and up.
Local newsletters can hit $0.35, for about two to three months, until the market runs out.
Payback runs on the same logic.
Nobody gets 24-hour payback. They have never believed anyone claiming it.
Thirty days is realistic only if you are extremely dialed in, and that level took one large finance publisher years of infrastructure.
An offer between sign-up and the first issue is what about 20 to 25% of their clients run. At 0.5 to 2% conversion, it recoups 30 to 40% of ad costs in the first month, with the rest coming in over three to six months.
Channel choice is the same math again.
LinkedIn runs $10 to $30 a subscriber because it is priced for SaaS companies happy to pay $300 a lead. Unless each subscriber is worth $50 to $100 to you over a year or two, it does not work.
Ads in other newsletters are the opposite. At one large daily business newsletter, that channel brought in the most engaged subscribers of any source. But it is a grind, every placement is a bet, and there are only so many newsletters to buy from.
The better question is not how low your CPA can go. It is how high your income per subscriber lets you pay.
Three pages to fix before the next dollar
So what do you fix before spending anything?
Three pages: a sign-up survey if you are B2B, a thank-you page, and a welcome email.
Having them is not enough, though. Think about the chain. A weak welcome email gets fewer clicks and replies. Fewer clicks means fewer engaged readers. And that shows up in your deliverability long before it shows up in revenue.
Here is your Monday morning.
Pull last month's paid sign-ups by UTM source and answer two questions. How many clicked the welcome email? How many fit your target reader?
If you cannot answer either one, that is the fix, before the next dollar goes out.
What I keep coming back to
1. Every lever depends on knowing who is real. Look at what actually moved the numbers: the survey, the UTM cohorts, killing lead forms, cutting the cheap ad set that was buying juniors. None of it was about creative. All of it was about whether the subscriber on the other end was the person you paid for. If I were building the scoring rubric, clicks would carry the most weight and opens the least.
2. Two operators, two verdicts on LinkedIn, same math. Another operator I follow says LinkedIn costs more per subscriber and less per engaged user. This buyer says $10 to $30 a subscriber does not work unless each one is worth $50 to $100. They can both be right, because the answer depends on cost per engaged subscriber against what an engaged subscriber earns you. If you only track CPL, you cannot tell which one applies to you.
3. Patience is the strategy. Kill in three to five days. Scale 15 to 20% a day. Keep the test version running and build the V2 before the original fades. I have blown up ad campaigns myself, and it was always impatience: scaling too fast, or not having creative ready when a winner died.
Run the Monday check, then reply with your two numbers: welcome-email click rate on paid sign-ups, and the share that fit your target reader. I want to see how far apart those are across this list.
GM.