How much to spend on testing ads?

I got into some classic Twitter beef the other day with this random marketing guy.

He argued the following:

1. You only need to spend 1x your AOV on any creative test before determining if it’s a winner or loser

2. You also only need to keep the test online for 1-2 days max. Anything more than this is burning money and constricting your margins

A lot of you reading might agree with one or multiple of these takes. If that’s the case, I encourage you to continue reading this email and keep an open mind.

If you already disagree heavily… well… enjoy reading to get that sense of “you showed him!” petty validation.

I never like to disagree with something based on feelings. There must be facts and logic associated with my decision.

I could sit here and tell you about how I’ve managed $100M+ of ad spend over 7 years for all different types of businesses. How I’ve spent $150k in a single day. How I’ve tried every bid strategy and every account structure you can think of to come to my conclusion.

But I’m not going to do that.

Instead, like I mentioned, we’re going to logically break down the arguments.

Let’s go.

Argument #1: You only need to spend 1x your AOV on any creative test before determining if it’s a winner or loser.

THis might sound like a great money-saving tactic on the surface level. But in practice it’s delusional. Let’s say you have a $50 average order value. Do you really believe you only need to spend $50 before you determine if the creative is a winner or not? It’s just not enough data for the algorithm. You’d be lucky to get 1 purchase, and even then, it could be a fluke or from a view-through conversion depending on your selected attribution setting at the ad set level.

My recommendation is quite a stark difference – spend 5x your AOV before deciding. Now, that’s a general rule. You can always turn off a creative test earlier if the underlying metrics are noticeably worse compared to your averages. Maybe that new test has super high CPMs. Maybe the hook rate is under 20%. Maybe you just have a gut feeling based on knowing YOUR ad account and the past trends. But, in general, I want several conversion events to populate before I know if a creative test is profitable or not.

Playing devils advocate to my own recommendation, you can also launch your ads inside a cost control environment. This way, new creatives will only receive spend if the algorithm believes it can acquire customers at your desired efficiency target. There are several other reasons why I wouldn’t suggest this, but we’ll save those for another Vault edition.

Argument #2: You also only need to keep the test online for 1-2 days max. Anything more than this is burning money and constricting your margins

Think back to all the crazy events that have happened in 2025 alone. 

Escalating conflicts between Russia and Ukraine and Iran and Israel. Insane tariffs dropping at a moments’ notice across many key industries within U.S.A. markets.’

Consumer sentiment is nearly impossible to predict day to day.

That’s the macro.

The micro is Meta making algorithmic change without telling you. It’s your new ads getting served to the incorrect pocket of users within the first few thousand impressions before it learns who reacts positively and who ignores it. It’s a technical glitch harming ad delivery for certain ad accounts.

A lot can go wrong.

A lot can also go right.

We want to eliminate the volatility, because our decision making success rate goes down with less data in front of us.

Leave your ads online for a minimum of 4-5 days before making any sweeping decisions… positive or negative.

That doesn’t mean you can’t scale earlier if signs are positive. That also doesn't mean you can’t cut or redistribute the budget. But 1-2 days is nowhere near enough time to be confident.

Paid social advertising is not an exact science. It’s also not a gut feeling. It’s somewhere in the middle. That’s the fun part or the challenging part depending on how you view it.

Alright, nuff said.

Till next week you clowns,