Moneyball Creator Marketing Part 1

Back in 2021, I was working as a Client Relations Associate at a Creator Marketing Agency in San Clemente. My job was pretty simple: source influencers from different niches of the internet and convince them to join our marketing campaigns and creator roster.

Sounds great, right? Getting paid to watch YouTube and Twitch all day?

Yeah, that's what I thought too.

After a few weeks of mindlessly scrolling through videos, doing a task I genuinely thought I would love, I was bored out of my mind and I couldn't shake this feeling that there had to be a better way. A more data-driven approach to finding creators and actually maintaining a database of them. I might be good at spotting talent, but I knew a computer would be faster, way more methodical, and able to remember a lot more than my brain ever could.

That idea stuck with me, even if I didn't have the skills to put it together.

Fast Forward a Few Years

Since that first role, I've learned a lot. I ran my own clothing brand, Balnced Clothing; over 2,000 orders heat pressed and screen printed out of my college dorm. I became a Blockchain Consultant at the same talent agency, building out their web3 offerings and building my understanding of what creators want out of audience tools. Then, after completing my Bachelor's degree, I returned as a Talent Manager until mid-2025.

Now I'm pursuing my Masters in Business Analytics at Chapman, and that idea from 2021? It still serves a glaring hole in the market. Data is still the lifeblood of the creator economy. Four years of prospecting creators, negotiating brand deals, and watching the industry change with the introduction of short-form video and streaming, I kept thinking about how much better these marketing campaigns could be if someone had the data science skills to create an accurate footprint of a creator's audience.

The difference now? I have the skills and understanding to build what I've always wanted to.

Welcome to Moneyball Creator Marketing

This post will serve as an introduction to my project and where I'm at with it so far. I wanted to create a methodology for businesses to apply a Moneyball-style approach to influencer marketing. Less gut feelings, more data-driven insights into target audiences and the creators that attract them.

Step 1: Keywords

I started by generating 100 keywords from 5 target games: Pokémon GO, World of Warcraft, League of Legends, Diablo, and Monster Hunter.

Step 2: The First Scrape

I scraped the first 100 YouTube results for each of those keywords. That gave me data on 7,325 videos from 3,121 unique YouTube channels.

Not bad for a starting point.

Step 3: The Huge Mistake

Then I decided to scrape all the videos from all of those channels.

This took… a while. I won't lie, I may have ruined my Thanksgiving doing this project (sorry mum!). And honestly, it's still not fully done. I ran into issues with how YouTube encodes different languages in their URLs and page structures, meaning scraping international creators turned out to be way messier than I expected. I've still got 474 channels I'm working through, but thanks to some suggestions from my professor Kenneth Lee Ph.D., I found a silver lining to my language problems. I can identify geographic audiences!

But here's where I'm at right now:

618,949 long-form videos
315,640 shorts
934,589 videos total

Nearly a million videos. My hard drive is not happy with me. Was this overkill? Probably. But now I've got a massive dataset to actually build something interesting with.

What's Next

Here's the thing—I don't expect anyone to scrape nearly a million videos every time they want to evaluate a creator. That's insane. But that's kind of the point.

This massive dataset is my benchmark. It's the ground truth I can test against.

The real goal is to figure out: how little data do you actually need to make good decisions? What if you only scraped a creator's last 15 videos instead of their entire catalog? Could you still identify the same patterns? Make the same predictions? Have the same level of understanding?

That's what I'm running towards. I'll be creating case studies with much smaller data pulls—practical amounts that someone could realistically gather—and comparing the results against my full benchmark dataset. If I can get 90% of the insight with 5% of the data, that's a method worth sharing.

I'm also developing what I'm calling my "Moneyball Creator Marketing Metrics"—specific measurements I think can help identify undervalued creators before they blow up, or help identify creators with a particularly dedicated fanbase. I'll be breaking those down in future posts as I work through this project.

If you're into data science, creator marketing, or just want to watch me figure this out as I go, follow along. This is the project I've wanted to build since 2021—now it's finally happening.

Got thoughts? Ideas for metrics you'd want to see? Drop them in the comments.