Hey Reader -

Client calls us. ROAS is down. Nobody knows why.

Welcome to subscription apps. It's fun here.

Here's the thing. You only see trials. That's it. That's what Meta shows you.

Whether those trials become paying customers? That data lives in some other system. Shows up days later. Sometimes weeks.

So when ROAS tanks, you're staring at a dashboard that's basically shrugging at you. 🤷🏻‍♂️

But wait. It gets even more fun.

Even when that conversion data finally rolls in, you can't just say "this creative caused that conversion." The data doesn't connect like that. Attribution is a fairy tale we tell ourselves to feel better.

So you're left with two options:

a. Guess.

b. Or do actual statistics.

Most people guess.

Now look. These aren't dumb people. They make educated guesses.

They've been doing this a while. They have theories.

• "It's probably the algorithm. Meta changed something."
• "Seasonality. Has to be."
• "Q1 is always weird, let's wait it out."
• "The creatives arent really resonating."

Sounds reasonable. Sounds smart, even.

Still guessing.

And at most companies, this is when someone says "let's loop in analytics."

Translation: let's make this someone else's problem.

The creative agency throws up their hands. "Not our department. Let us know what you find."

Everyone sits around for three weeks waiting for a report.

Meanwhile....

We do things a little differently.

We track a TON of metrics in our internal dashboards that dont come out-of-the-box from Meta and the platforms.

But this time, the normal view wasn't telling us anything useful. UGH. WHY WHY WHY.

So we went deeper.

We pulled 55 days of ad data. Matched it against backend conversion data.

And since you can't draw a straight line from creative to conversion: because that's not how reality works: we used non-deterministic statistical methods to find the signal in the noise.

Pearson correlation analysis. T-statistics. The whole thing.

A creative agency. Running real statistics. On purpose. For fun.

I know. We're a blast at parties.

Here's what we found.

The creative mix had shifted toward statics. And that shift changed who the algorithm was finding. Different optimization. Different users. Worse conversion downstream.

Same trials. Worse customers.

Mystery solved. With stats.

Not intuition. Not experience. Not "I've been doing this for 10 years and I think it's probably..."

Most agencies would've made an educated guess, launched some new creatives, and hoped for the best.

We went deep. Found the root cause. Fixed it.

Anyway... if you are spending $50k/day+ and want to scale your creative with a team that nerds out on this kind of thing, hit reply and say 'NERD' - and we'll talk.

Later,

Shamanth

PS: From the socials:

​I havent opened Meta Ads Manager for 24 hours​

​Chat GPT wants $60 CPM. Here is what that buys you elsewhere.​

​What makes an ad convert? It's not the hook in and of itself.​

​This is how many ads you need for subscription apps(up to $1 million/mo in ad spend)​

​Ladder was an App of the Year Finalist in 2025. BruteForce AI ran semantic analysis on 69 of their ads​