For a few years, one of LinkedIn's open secrets was the engagement pod: a group chat where members agreed to like and comment on each other's posts within the first hour, gaming the early-engagement signals that ranking systems use to decide distribution. Pods worked for a while. Then, mostly, they didn't. The story of why is a useful window into how feed ranking evolved.
Why pods worked at first
Early ranking systems leaned heavily on simple velocity: a post collecting engagement quickly was treated as promising and shown to more people. Velocity is trivially gameable — thirty people in a chat can manufacture it in minutes. For a while, manufactured velocity was indistinguishable from the real thing, and pod members watched their reach multiply.
Why they stopped working
Platforms had every incentive to close the loophole, because feeds full of pod-boosted mediocrity drive users away. The countermeasures were straightforward, and platforms have publicly acknowledged cracking down on coordinated engagement:
- Graph analysis. The same cluster of accounts engaging with each other, post after post, at improbable speed, is an easy pattern to flag. Engagement from your habitual reciprocators gets discounted.
- Quality weighting. Systems moved from counting engagement to weighing it — who engaged, how they're connected to you, whether their engagement usually predicts broader interest, whether comments carry substance.
- Downstream signals. If the boosted post reaches a wider test audience and that audience shrugs, distribution stops. Pods could fake the first hour, but never the second audience.
The general law underneath
Pods are one instance of a pattern that recurs across every platform and every era of manipulation: when a proxy for quality becomes a target, the platform's defense is to make the proxy harder to fake — usually by moving closer to measuring the real thing. Follower counts gave way to engagement rates; raw engagement gave way to weighted engagement; every step rewards the same boring strategy, which is being genuinely worth reading.
What to take from it
Anything that games a signal has a shelf life, and the correction usually costs more than the boost was worth. Communities that share each other's work because they honestly value it remain fine — the graph patterns of genuine affinity look different from mechanical reciprocity. The line to stay on the right side of is simple: would this engagement exist without the agreement to produce it?