The post got 47 likes. In the context of my account at the time, that was a flop. I'd been averaging 300-400 likes on posts in that format. 47 was a signal that something had gone wrong — wrong timing, wrong topic, wrong framing, wrong something. I noted it and moved on.
Six months later, someone DM'd me about that post. They said it had changed how they thought about their work. They'd shared it with their team. Their team had shared it further. It had been circulating in a Slack channel I'd never heard of, in an industry I didn't know I was reaching. The post that got 47 likes had been read by thousands of people I'd never know about.
I've been thinking about this ever since, because it broke something in my mental model of how content works. The model I had was: likes and shares are a proxy for reach and impact. More engagement equals more reach equals more impact. The 47-like post broke that model. It had low engagement and high impact. Those things are supposed to correlate. They didn't.
The reason, I think, is that the post was specific in a way that my higher-performing posts weren't. The posts that got 300-400 likes were broadly relatable — they were about experiences that a lot of people had, framed in a way that was easy to engage with. The 47-like post was about a specific problem in a specific context. Most people who saw it didn't relate to it. But the people who did relate to it related to it deeply. And those people shared it with other people who would relate to it deeply.
The creative culture conversation about this usually frames it as "niche vs. broad" — the idea that you can either reach a lot of people shallowly or a few people deeply. But I think that's the wrong frame. The 47-like post reached a lot of people. It just reached them through a different mechanism than the algorithm. It reached them through human curation — people sharing it with specific other people because they thought those specific people needed to see it.
The algorithm optimizes for engagement. Human curation optimizes for relevance. These are different things. A post that's highly engaging is one that provokes a reaction — a like, a share, a comment. A post that's highly relevant is one that someone thinks another specific person needs to see. The algorithm can measure engagement. It can't measure relevance. Relevance happens in DMs and Slack channels and email forwards, in the spaces the algorithm can't see.
The maker mindset that I'm trying to build is one that doesn't optimize only for what the algorithm can measure. That makes things that are specific enough to be genuinely relevant to the people who need them, even if that specificity costs engagement from the people who don't. The 47-like post was the most important thing I made that year. I didn't know it at the time. The algorithm definitely didn't know it.
The lesson I've taken from this is not "engagement doesn't matter." Engagement matters. It's a signal. But it's one signal among several, and it's the signal that's easiest to measure, which means it tends to dominate the others. The harder-to-measure signals — the DMs, the shares in private channels, the conversations that happen because of something you made — are often more important.
The product thinking version of this is: don't optimize only for the metrics you can measure. The metrics you can measure are a subset of the things that matter. The things that matter include things you can't measure. Build for those too.
I've been making more specific posts since then. Posts that are clearly not for everyone. Posts that will get 47 likes from the people who don't relate to them and will get forwarded to exactly the right person by the people who do. The algorithm doesn't love this approach. The people who need the posts do.
The content strategy that I find most interesting is the one that treats the algorithm as one distribution channel among several, not as the primary audience. The algorithm distributes to people who might be interested. Human curation distributes to people who definitely are. Both matter. The second one is harder to build for, and more valuable when you get it right.
The post that flopped is still out there. It still gets found occasionally. Someone DM'd me about it last month — a different person, a different industry, the same reaction. The algorithm gave up on it six months ago. The humans haven't.