Hi Reader -

I've scaled tens of products well past a million a month in ad spend: both in-house and agency-side, and here's what's funny about operating at that level.

Every measurement argument you see online stops mattering.

The guy with 2k followers and a ring light who cracked the code with a new lookalike stack. The "I scaled a brand to six figures, here are my 9 CBO learnings" guy. The targeting hack that changes everything until Meta deprecates it in March.

Nobody at a million a month is having those conversations.

Because they've learned the thing nobody posts about: the more you spend, the less certain you get. At that scale your users are on web and app at once, and these are surfaces that don't talk to each other.

At $50k a month, everyone is sure. At a million a month, the sure ones have been carried out on stretchers.

The ones still standing lie awake about three things. Not because they haven't worked on them. Because they're the kind of problems that are truly hard to solve.

1. Is the optimization event giving us the best users?

Sooo is the first purchase the event that actually gives you high-LTV users, or looky-loos?

Then the web-app split makes it even more complicated. The first purchase happens in the app. The renewals and the upgrades: the part that actually makes the unit economics work, happen later on the web(or vice-versa).

The algorithm only sees the first one. So it spends like the second one doesn't exist.

The ring light guy assumes the event is always right. Once in a while he discovers cohorts and thinks it’s the next best thing after sliced bread.

2. Do I even know which platform the money lands on?

User installs on the app, pays on the web. Or signs up on the web, gets valuable later inside the app, where the algorithm that bought them is legally blind.

How much of the value happens on web vs. app is hard for most businesses to truly understand.

The honest answer at scale isn't "we solved it." It's "we poke it with a stick every week to make sure it still flinches."

3. Is the algo learning from truth, or from noise?

Every channel takes credit for every conversion. Split that across web and app, where the two sides barely share data, and your dashboard adds up to 140% of your actual revenue while everyone nods like that's a normal thing for numbers to do.

Feed a machine that signal and it does what machines do best. Compounds the error. At speed. In the wrong direction. With total confidence.

Of course these corporations do no evil.

You’re welcome.

---

These problems are hard, which means most of your competitors aren't solving them either. The operator who actually wires web and app together, even imperfectly, sees a version of reality the rest of the category can't.

And at a $1mm+/mo in spend, these wins add up wayyy more than your next best hook or even channel.

That's why these problems are not a headache - they are a head start.

These are hard problems. And they are supposed to be hard.

We've seen enough data points to be able to pattern-match and see the failure modes - and also the ways(even imperfect) that products win in a world with messy measurement.

If you're spending $50k+ to $1mm/month and you've started to suspect your dashboard is telling you a gorgeous bedtime story your bank account doesn't agree with, hit reply with "SIGNAL" and we'll talk.

Cheers -

Shamanth

P.S. Our friends at AppsFlyer have launched their new Web Performance Measurement solution at MAU Vegas this week. If you haven't seen it yet, here's a good read on what it actually means.

PPS: this is a partner email w/ our friends at Appsflyer.
​