Kai Tak, the old Hong Kong airport runway has reopened in a new form. The former airport site became three things: a shopping district, a cluster of office towers, and the preserved runway.
One of them is Airside, a skyscraper and mall built directly above the Kai Tak MTR station. If you remember Kai Tak before, this is exactly where planes once made their hair-raising turn over the city before landing. Now, instead of descending into an airport, people ascend into a mall.
By 2024, Airside had been open less than a year. It was still a new landmark, and like all new landmarks, it needed attention—tourists, businesses, and investors. The question was how to get it.
That summer I ran a digital marketing campaign for Airside’s Olympic events. In one month, it generated 300,000 shares and a reach of over 30 million. Buying that reach on Instagram would have cost about half a million USD. My cost: zero.
Multiple factors contributed, alone, each was small, but together, they compounded.
1. Feedback Loops
The first choice was to make shares the primary metric—above views or likes.
Metrics are a compass: what you measure is what you get.
Shares signal genuine interest, people might like something casually, but they share only if they want to attach their name to it.
Another reason: Meta had recently updated its recommendation algorithm. Content with high share rates would be shown more widely (see point 3).
2. Speed of Iteration
In programming, there’s a saying: speed of iteration beats quality of iteration. Some often obsess over polishing a single campaign until it’s “perfect.” I never built super-detailed plans, more like a tree of rough plans. I published rough ideas, measured results, and doubled down only on what worked. It is my belif that good or great should be the result of correction.
In practice, I ran multiple experiments at once. Instead of one campaign (that's why I called it experimental), I had few micro-strategies competing, the best performing ones survived and were scaled, and the bad ones were killed. In other words, I treated things like a genetic algorithm.
A common objection I faced was: “Bad posts hurt reputation.” In reality, if people don’t see a post, it doesn’t affect you. Attention is scarce.
3. The New Algorithm
Before 2024, platforms like Instagram and Facebook favored follower count. A million followers meant you could go viral; without them, you couldn’t.
In 2024, Meta changed this. Posts are now recommended based on engagement, not follower count. It no longer mattered if you were a celebrity or unknown—if people liked and shared your post enough, the algorithm found an audience for it.
If your content increases user time on the platform, Instagram promotes it for you.
Shifts in information flow, whether in speed or medium, quietly make the impossible possible, and those who notice first gain the edge.
4. Technology as a Multiplier
Technology should:
- Lower iteration cost as much as possible.
- Increase iteration speed as much as possible.
The real goal is to raise your learning rate. Over time, publishing enough content builds intuitive judgment about what works. Like a farmer who has examined thousands of eggs and can tell which will hatch without being able to explain how. Over time you learn to spot winning ideas.
Technology here was not just for mass production, but to accelerate learning and sharpen judgment at increasing speed.