Every few months someone asks me why we didn't move to San Francisco when we started Fanlytiq. The assumption behind the question is usually: that's where the money is, that's where the talent is, that's where AI companies are born. And there's truth in all three — but none of those things were the actual problem we were trying to solve.
We built Fanlytiq for digital media and streaming teams. Our customers are video production companies in the Valley, short-form studios in Koreatown, streaming platform analytics departments in Burbank, and social content teams in Silver Lake. Nearly all of them are within thirty miles of our office on Sunset. The choice to stay in Los Angeles wasn't defiance of convention. It was customer proximity treated as product infrastructure.
What Working Near Studios Actually Taught Us
Before starting Fanlytiq, I spent three years at Cirrus Media — a mid-sized streaming data company that no longer exists under that name — doing audience analytics for their content acquisition team. Marcus Adebayo, our CTO, spent the years before that building recommendation signal pipelines at a video distribution startup nearby. Sofia Reyes, who leads product, came from a digital media agency where she spent most of her time translating data reports into content decisions that producers would actually act on.
Between the three of us, we had about nine years of watching the same failure mode play out: a video team would spend days in post-production, publish, and then find out forty-eight hours later — after the algorithm's recommendation window had already closed — that the first minute and a half was responsible for 60% of the drop-off. The fix was usually obvious in retrospect. Cut the B-roll opening. Lead with the conflict. Change the thumbnail to one that sets more accurate expectations.
None of that was new insight. What was new was that we couldn't find a tool that could tell you before the video went live. The existing analytics tooling — and we'd used most of it — was designed for post-publish measurement. It told you what happened. We wanted something that told you what was about to happen.
That gap was obvious to us because we were physically proximate to the teams experiencing it. We saw it in sprint planning meetings at production companies. We heard it in complaints from content leads who were being asked by their managers to explain declining completion rates. We sat in on thumbnail debates that lasted two hours and ended with someone's gut feeling winning.
The LA Media Ecosystem Is Not a Consolation Prize
There's a version of this story where Los Angeles is framed as the startup founder's second choice — you're here because you couldn't attract SF capital or SF talent. I want to push back on that directly.
The talent we needed wasn't distributed evenly across the country. Video ML engineers who had actually worked on recommendation systems for streaming platforms — not hypothetically, but who had shipped models against real watch-time data at real companies — those people cluster in LA. The content strategy experience that informs how Fanlytiq's output needs to be framed so that a video editor can act on it immediately, not just a data scientist — that expertise also concentrates here.
Marcus joined because he'd spent years trying to solve the cold-start problem for video recommendation, and he understood the engagement physics of video content at a level that required having worked near actual production pipelines. Sofia joined because she'd been the frustrated translator between analytics reports and creative teams for long enough that she'd mentally built a better tool herself at least twice.
Neither of them were looking to relocate to a city where media was an abstraction.
Proximity as a Development Shortcut
The practical implication of being based here is that our feedback cycles are unusually short. In our first year, we were doing informal walkthroughs of early Fanlytiq builds with video teams we already knew — not as formal beta users, but as people who'd answer a Slack message and give us thirty minutes on a Thursday afternoon.
That kind of proximity doesn't replace structured user research. We're not saying casual relationship capital substitutes for rigorous product development. But it does mean we caught some design mistakes very early that might have persisted longer otherwise.
The specific example I remember most clearly: our first segment scoring output was organized by confidence interval. We showed drop-off risk as a range — "this segment has a 38–61% probability of causing significant viewer loss." A content director we know pulled up our prototype on her laptop, looked at it for about fifteen seconds, and said: "I can't give this to my editor. She needs to know what to fix, not how unsure you are."
That feedback reshaped how we present scores. Fanlytiq now shows the most likely outcome with a recommended action — not because we've hidden the uncertainty, but because the person who needs to act on the information is not a data scientist weighing confidence intervals. She's an editor with a publish deadline in four hours.
We would have eventually learned that lesson from user interviews. But we learned it in week three of building because we were a ten-minute drive from the person who told us.
What This Means Going Forward
We're not anti-remote and we're not claiming LA is the only place to build a media analytics company. There are teams doing interesting things in New York, London, and Seoul that we pay attention to. The media and entertainment ecosystem is geographically distributed in ways it wasn't five years ago.
But the decision we made in 2023 to stay rooted in Los Angeles rather than chase a geography that seemed more legible to investors has held up. Our customer relationships are stronger because of it. Our product intuition is calibrated by constant, low-friction exposure to the people who use it. And the team we've built reflects the specific mix of video production experience and ML depth that this problem requires.
If you're building a tool for content teams and you're not embedded in the workflow those teams actually run — close enough to watch how they make decisions in real time — you're going to build something slightly wrong for a long time before anyone tells you why. We got lucky that our location made that feedback hard to avoid.
