It's actually not that useless. Everything organic, or interacting with something natural follows the power law. Or 80/20 if you prefer.

When working on security and integrity issues, we found 10-20% of all traffic would be inorganic. The more course the metric, the more likely it was to be exactly 20%

To me, knowing nothing about this specfic domain, and just abuse/integrity in general, 45% means it's well over double what I'd expect from an unmanaged source. Well over double, because true double wouldn't be 40% (20/100) + 20 = 40/120 = 33%

This heuristic tells me it's specific, targeted, and well above the background noise youms might ignore for higher priorities. In other words, it's a problem that's actionable.

Here, I assume stoking anger and outrage is the goal. That's why it not being 20% is significant.

▲

crazygringo 7 hours ago | parent [-]

> knowing nothing about this specfic domain

Then your comment isn't actually contributing anything.

And the 80/20 rule doesn't have anything to do with this.

The idea that the baseline value for any statistic at all in the world is 20% is not how anything works.

The 80/20 rule is an informal observation that you get 80% of profits from 20% of customers, or can draw 80% of conclusions from 20% of data. It doesn't say anything about the baseline rate of any arbitrary statistic.

	▲	grayhatter 3 hours ago \| parent [-]
		> Then your comment isn't actually contributing anything. Well, you thought that the 45% was meaningless on it's own too, perhaps the problem is more your context window is too small? But, I'll try again. The behavior of bots translates well across domains, the domain of rage fueling bots has additional nuance that I haven't spent significant time studying. i.e. The generalized heuristic is useful, but this comment is lacking in the domain expertise of rage fueling bots. Perhaps the 45% has more meaning with more experience in rage bait? > The idea that the baseline value for any statistic at all in the world is 20% is not how anything works. That's not what I said. > The 80/20 rule is an informal observation that you get 80% of profits from 20% of customers, or can draw 80% of conclusions from 20% of data. It doesn't say anything about the baseline rate of any arbitrary statistic. The 80/20 rule is a simplification of the power curve. I mentioned it not to imply that 80/20 is somehow magic here, but to reference it as a useful heuristic many people will already be familiar. The power curve does show up seemingly everywhere. The more common example where it shows up in understanding social media, is in the behavior of real users. Roughly 1% will generate content, 20% will interact with content, and 79% will scroll/lurk. This also applies to users on the individual level too. The near exclusive majority will spend 79% of their time reading, 20% clicking like/ignore, and 1% of their time submitting or commenting. I mention the power law, because when working with inauthentic behavior, it becomes anomalous when it diverges from that 20% ish expectation. (This is basically the example you mentioned that you're already familiar, it's the same 80/20 from your examples) Say you have a spam botnet, only about 20% of it will interact with any given post. If it's more or less, that's weird. The reverse works too, you'll never see more than 20% of real organic users interacting with spam/bot content. If you see more, or less, that's weird (that cohort is probably not real users) or in other words: If the thing having something to do with how humans interact with social media doesn't follow the power law, it's weird. How does that apply here? If the number was 20% instead of 45%, it would actually be meaningless (you could just say there are bots). What does that extra 15 points mean? That I can't tell you. But it's not in the range of "boring inorganic behavior", which I say as someone who's spent a lot of time trying to find, classify, and remove inauthentic accounts/behavior.