How much should I spend to test whether Meta ads can sell a new offer?

When someone wants to try Meta ads for something new, I push them toward a capped test instead of thousands of dollars. Something like $500. See what comes in for $500.

It's an experiment. I can't say for sure it will work. There are a hundred different ways to run ads, so you try different creatives and different locations, learn what works, and then go harder on the one that's working.

How small can a first test be?

Small. For someone learning on their own budget, I'd do $5 or $10 a day for a week and iterate from there. You won't optimize it perfectly. You'll learn a lot doing it anyway.

What numbers should I decide first?

What a buyer can cost and still leave you making money. When I tested a custom song product, the question was simple: could we get buyers for under $20 on a $50 song, or under $10 on a $30 song?

For leads, you won't know quality until you have volume. Track how many you reach and how many move forward, and the real numbers show up.

What does a failed test look like?

That song test. About $80 spent, one sale, and two more people who filled out the order form but never paid. The one buyer had already asked for three revisions.

"We ran the test and unfortunately, I don't think we passed, which kind of sucks because we need to get good at Meta ads and not just for newsletters."

What I took from it:

• Broad was too broad. We ran it with no targeting. I thought the creative was good enough, so the next version went to specific interests instead of Advantage+, aimed at the people most likely to want it: moms of kids graduating high school.
• The price might be wrong. I wondered if we needed to raise it to justify the ad cost.
• A small test fails cheaply. $80 was enough to see it wasn't working as it was.

Why should the ad name a specific group?

Because generic ads disappear in the feed. For the song product, the plan was never an ad that says we'll make you a custom song for $50. It was: are you a real estate agent? Want your own custom song? Listen to this. Then a separate version for insurance agents.

Same idea for a friend selling AI training. People scroll past general AI pitches all day. "Hey, insurance agent, learn how to use AI for insurance" stands out. A demographic niche, like dads, probably beats plain generic but not as much as an industry does. If the training is AI basics, you don't need to be an industry expert, and an AI coding agent can get you enough background in a day or less.

Once something works in the test, give it more budget. How I structure tests on my own newsletters is in running a main and a test campaign.