"Only 10% of users...", "Only 5% of Tesla owners...", "Only 3% of Americans...", "Just a few % of blind people..."
Somehow this is a mistake everyone constantly and repeatably does to make things crappier. And that includes rockstar engineers, normal engineers, managers, product people, everyone.
Once you get to Google's scale it becomes more ridiculous. There was some feature that was being removed from YouTube at one point and the talk was about "only 5% of users" - but that could translate into millions of people.
Percentages are obviously useful to get a feeling for certain things but too often they're used to minimise the impact of something.
At least, that’s my generous interpretation of the incentives at play. I don’t think it’s as simple as “those engineers are dumb” / “corporations are evil”
Ultimately, the legal fiction was created to avoid having to create new rules but it creates all sorts of problems because corporations and groups in general are more like political parties and governments where they’re made up of subgroups with their own agendas and those subgroups have their own individuals with their agendas.
For example, if a corporation mass poisons people, no one typically goes to jail and instead the corporation pays a fine, very similar to what would happen if the government did that (see Flint and what PG&E did in Hinkley). You can put people in jail. You can’t put a corporation in jail and the corporate veil is insanely strong against getting pierced which causes all kinds of moral hazards.
That’s not what was said which is that ALL corporations should be presumed to be doing this unless proven otherwise.
Everywhere else it's pretty much guilty once accused.
When someone keeps violating their parole you start asking them to justify why you shouldn't send them back to jail.
We're past the trial phase. The guilt of corporations has already long been established over the past 100+ years. Furthermore, I am not the government and am not required to give them the presumption of innocence.
On a personal note, my great-great-grandfather didn't die shooting Pinkertons for nothing.
Nitpick, but there are 8 stadiums (college football) in the United States with a capacity of over 100,000
What I mean in this case is that if you say to yourself, "hey, only 4% of my users are using more than 50 gigabytes, and I can afford to lose them", it is very likely you are using a model where those 4% of your users are essentially uniformly randomly chosen from your user base. You can see usage, and you may not be able to see much else, so you assume that usage is uncorrelated to any of the other aspects of these users that you may care about.
In reality, everything is correlated to everything, and what that means in this case is that your 4% of users are not being randomly chosen from your user base. This doesn't mean that every way this sample deviates from random is bad for you; you certainly can just get lucky. For instance, that 4% of users probably includes a huge proportion of your users using it for piracy or similar activities. Cool. You may even be using this as a motivation for the action since you can see this to some extent based on reports and analytics, certainly.
Unfortunately, that 4% probably also includes a disproportionate number of your True Fans, who love Google partially because of this free drive space, and because of that love, drove their work environments to Google Teams and other paid services, thus netting Google profit far beyond the costs of this particular Drive account, but indirectly and where accounting can't see it. True Fans who will perceive this not as a business transaction, but as a betrayal, and who may in their passions take actions to do things like (on the margin) remove their companies from paid Google services, and take time out of their day to create bad press.
I'd say that's the core mistake being made over and over. You think you're just getting rid of your "problem children" but those problem children very frequently have correlations to something you do in fact actually want.
Of course, that doesn't mean you want to be abused either. It just means a more delicate approach may be called for than just hacking off a chunk of your user base when you don't really know what it contains. It ought to be an assumption that any time you're taking an even slightly non-random slice of your user base like this that it may have hidden correlations in it you're not taking account of.
You(r company) have say 30 products, each randomly alienating 4% of their users. The end result is that you don't have a lot of users left.
Option a: admit they screwed up and open themselves up to additional legal risk by doing so Option b: minimize the damage done and annoy the hell of this end users.
Large corporations will always pick option b.
If the entire service went down, only a handful of users would be immediately affected.
Also it's poor taste to compare power users who want 5M+ files to blind users, as if both groups deserve equal effort to protect their interests.