I’m Sarah Kao. I help companies understand people better—through community, communication, audience insight and better ways of working. Fifteen years of it, mostly in games and entertainment.
The interesting work is never the part somebody has already named.
Different words, one underlying problem: the company has lost track of the people it serves. I build the relationship back—the function, the operating model, the shared language, the feedback routes that keep a business honest about what its audience is really saying.
I like this work because nobody owns it yet. At Fortis Games it meant writing an operating model before writing job descriptions. At King it meant giving teams across several markets one way of reading their players. Neither time was the useful outcome a campaign. It was a capability the company still had afterwards.
Fifteen years, mostly in games
Games studio · pre-launch · function built from zero
There was no team, no tooling and no shared idea of what good looked like. Marketing wanted reach. Product wanted feedback. Leadership wanted reassurance. Each of those is a different function, and hiring against all three would have produced a team that pleased nobody.
So I wrote the operating model before the job descriptions: what community would own, what it would deliberately not own, and how player feedback would reach the people making product decisions while those decisions were still open.
What changed: a capability existed where there had been none, and product and marketing ended up with the same vocabulary for the audience. They could disagree about the right things, quickly, without me in the room. That was the point.
Read the full accountThey fail because they misread their customers, misread each other, or never turn insight into a decision.
Customers understand something different to leadership.
Leadership understands something different to engineering.
Engineering understands something different to marketing.
Most companies try to fix this by communicating more.
I think they should start by understanding better.
My whole career has been spent between teams that don’t naturally speak the same language. Sometimes that’s players and developers. Sometimes executives and technical teams. Sometimes marketing, community and product. Closing those three gaps—what customers mean, what colleagues mean, and what to do about it—is the job.
Which is why my work has never really been social media. Community, done properly, is the richest source of intelligence a business has: it tells you where your assumptions are wrong while there is still time to act. My job is to make that intelligence reach the people making decisions, in language they already use.
In practice I build rather than broadcast. A function that didn’t exist before. An operating model that says who owns what. A feedback route that reaches product while the decision is still open. New tools help—they take friction out of the work and speed up how fast a team can learn. They don’t supply the judgement. That part is still human, and it is the part that decides whether any of it was worth doing.
What actually changes
The tools change. The work doesn’t.
Global entertainment brand
Player sentiment existed everywhere and counted for nothing. I built a standing route from community signal into product planning. What changed: audience understanding became an input to roadmap decisions instead of a retrospective nobody read.
Launch planning and narrative
I wrote the narrative several teams could each say out loud without flinching, then turned it into a launch framework. What changed: the next launch began from a shared story and a named owner for every part of it.
Operating models and ways of working
A documented way of turning customer signal into decisions. What changed: three other teams adopted it without being asked, so it became a company capability rather than my personal habit. Some of it is automated now; the judgement isn’t.
Most companies already have the signal. Very few have anywhere for it to land.
Institutional memory is a design problem, not a documentation problem.
It shows up in tone months before it shows up in numbers. Newer tools help me read it faster. They don’t decide what it means.