In partnership with

How 1,500+ Marketers Are Using AI to Move Faster in 2025

Is your team using AI like the leaders—or still stuck experimenting?

Masters in Marketing’s AI Trends Report breaks down how top marketers are using tools like ChatGPT, Claude, and Breeze to scale content, personalize outreach, and drive real results.

Inside the report, you’ll discover:

• What AI use cases are delivering the strongest ROI today

• How high-performing teams are integrating AI into workflows

• The biggest blockers slowing others down—and how to avoid them

• A 2025 action plan to upgrade your own AI strategy

Download the report. Free when you subscribe to the Masters in Marketing newsletter.

Learn what’s working now, and what’s next.

DOWNLOAD THE REPORT

Why Song Testing Is Broken (And How to Actually Do It Right)

It seems every time I talk to an artist with a label and team, they are increasingly talking about "testing songs." Some artists even tell me their label won't let them release a song until it has tested well.

Even more strange is that when I ask how they're testing them, it seems to be the single stupidest way I have ever seen. But this is the music business, and the reason we call smart people rocket scientists and not Project Managers - they aren't often bringing us their smartest and brightest.

Since I have to tell you, there are ways to test songs, but y'all are doing this really, really wrong. By the time you're done with this, you'll know how to test songs better than many of your favorite artists since there's a lot of stupidity going on today.

Why Are Musicians Testing Songs?

If I'm being real, it beats me since most of the time they do it in such a dumb way that it makes no difference.

“Get the Song to 50k Likes and We’ll Release It!”

The advice from many labels these days is some version of this ridiculous thought that until a song receives 50k likes, it shouldn't be released.

I find this silly since musicians will simply hit “promote” on the one they want to release anyway, even if it's the 9-minute song about how sorry they are for cheating on their partner that everyone agrees is terrible.

Musicians will make 50 videos that get a thousand likes or find a celebrity to do something silly on Cameo and post till they get those likes. This is way too easily gamed and dumb.

The Other Method People Are Using

They put up a few different songs with nearly the same exact video look and sing along or pout their lips and try to look sexy or whatever y'all freaks do, then see which one blows up. That's a lot of what we'll be talking about here today.

How It Usually Goes

What many people do to test songs is choose the hook of, say, 4 songs, but sometimes I see it done with a dozen songs. Their logic is that they can't tell which ones get popular, so why not let the public choose? Doesn't sound so bad, right? Well, lemme cook…

They make 4 TikToks, one for each song. Song 1 gets 800 views, Song 2 gets a whopping 1,009, Song 3 gets 998, and Song 4 gets 1,602 views. So obviously they should put out song 4, right? Not quite.

Statistical Significance

Now is when I beg you not to run, as I use some real college professor talk for a minute. I can see when your dumb asses run - YouTube gives every video this chart, and I see when I use big words y'all's little baby brains run and go watch some video where a guy plays guitar with a rubber chicken instead.

Enjoying this? Forward it to a music friend you’d like to be closer to and start a discussion!

How To Achieve Statistical Significance

There's this term we use called “Statistical Significance”. What this means is we often don't have enough data to be able to tell if something performed better and make a conclusion, since there are other variables at play, and one of them could have corrupted the data.

For example, let's say we were doing this for that motherfucker sombr everyone loves right now. I have been talking to those of you who find men attractive, and something about this dude's neck anatomy really gets some of y'all a little too horned up.

Now let's say in video 4, while he was selling the song, he showed that neck in just the way you like it, you may watch it 4x cause it's getting you all horned up, and now it has 1,600 views instead of 800…which wouldn’t have happened if he didn’t show all that neck in the video.

It is very important that tests get done, factoring in outside forces, and with videos, there are many outside forces.

The Numbers Reality

In many statistics, if you get double the amount of views, it can mean it's a very big deal. But with songs that ain't really the case.

Let's do a really extreme example: take OutKast's classic album "Speakerboxxx." "Hey Ya" has 1.7 billion plays, but "She's Alive" only has 3 million plays. So if Big Boi and Andre were doing this, "Hey Ya" would be 567 times better in the test. That is statistically significant.

Unfortunately, I have seen about 10 too many times that a little extra side boob being the only difference in two cuts of a video makes the one with a quarter second of side boob suddenly perform much better. And that's not cause you all are a bunch of horned up weirdos. That is because small details often change the performance of videos.

But 2 or even 4x on TikTok is often not enough to actually be statistically significant since TikTok's performance swings wildly, and many artists often have videos that do 10 or even 30x, so 2x isn't that big a difference. Especially those of you who are still in algorithmic jail - when you're only getting below 300 views, it's often because the algorithm has no idea who to send you to, so we're probably not even testing it on people who like your style of music.

More Evidence

Let's go look at Ravyn Lenae’s TikTok page since I recently wrote a newsletter about her and Sombr and how they use recognizability as an advantage. We can see in about 6 videos that Ravyn Lenae can go from 1→12x her lowest performing video.

While in many statistics something doing 2 or even 3x better doesn't really mean much, it's just not enough to be statistically significant since too many other factors can be at play.

But I do believe when you get around 3-6x, depending on the number, and you get something that performs better, you can see statistical significance.

You Need To Experiment Multiple Times

But the truth is you can't just run this experiment once - there are too many factors.\

Why You Need Multiple Tests

Anyone who's done a lot of experimenting on the short-form video platforms of TikTok, Reels, and YouTube Shorts will tell you the same thing. The time of day or exact length of a great video will only affect its growth by a percentage point or 10 at max. A great video or idea will defy all the odds.

But like I said before, little details will change performance greatly, so we can't just run this test once. Oftentimes, to get data that is cleaner, we need to run these tests 4 times. Even when we are running these tests with the same text and videos on different songs, the data is not conclusive, in that it will be a mixed bag many of the times, with different songs winning each round.

*Even when you run these experiments 4x, there isn’t always a clear winner.

The fact is, video and what is in the video are big factors, but also so is the mood of the audience, as it's tested each day. So if you want to do one of these tests, you need to look for outliers where the song wins at least 3 rounds of these tests by a statistically significant amount of views and likes.

*But I gotta be real - I personally don't think this is the best way to do this.

The Real Problem

90% of songs don't blow up from TikTok. So many songs need more than a snippet to feel the impact.

Let's remember these days, most songs spread from friends telling each other about them, or the algorithm on YouTube sees people really like them and recommends them more. These tests are nowhere near foolproof. I personally have probably seen more of them than most people alive, and I find it really silly that this idea keeps getting more and more popular.

The Best Way to Test Songs

The way I see the best answer happen for what song to promote is having a community of people who like your music. The best answers come from asking people who actually like your style of music. Not asking your mom or partner which song they like best when they aren't passionate about your style of music.

Finding community and reaching out and making friends with other musicians like you, and starting to become closer to people from doing things like this with friends who make music and like what you do, helps. Being in a community with like-minded musicians or people who like your style, I think, consistently gets better answers than any of the tests I see and comes with tons of other benefits.

And you do this all without having to put out a tease of some songs that, in this day and age, when people are doing weird copyright claims on songs before they come out, can save you a lot of headaches.