4 ms·
And yet somehow I see this argumentation constantly done everywhere, including this very site. "Only 10% of users...", "Only 5% of Tesla owners...", "Only 3% o
by izacus 4y ago
And yet somehow I see this argumentation constantly done everywhere, including this very site.
"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.
- jerf 4y agoI would submit the root cause of this is not what you might expect. I would say it's a failure to understand that "everything is correlated": https://gwern.net/everything https://gwern.net/everything 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.
- hnfong 4y agoTotally agree but it's even worse than that. 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.
- bagacrap 4y agoNo one thinks 10% of users at Google scale is a small number, or 3% of people at America scale. 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.
- InCityDreams 4y ago...fun at parties.
- illiac786 4y agoIt’s because legal risk matters more to them than end user satisfaction (of a specific end user group maybe): 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.
- ilyt 4y agoI'd assume that in public communication it is not a mistake but deliberate move to diminish the perceived severity of the problem.
- marginalia_nu 4y agoI think it's also easy to miss the fact most likely only a small percentage of Google Drive users actually use the service in a way where they would notice a degradation. The majority of users on most free platforms are largely inactive. If the entire service went down, only a handful of users would be immediately affected.
- boredumb 4y agoI think it stems from people being really poorly equipped to think about scale like that. Even someone who has lived in NYC and/or been to jam packed sports stadiums, you're still only around ~50k humans max and that comes out to around 0.005% of drive users. So if a developer comes up and says a change will only effect 0.1% of users it may seem like a insignificant percentage but the actual impact is something to the effect of 20 full stadiums of people, massive.
- bluedino 4y ago> m packed sports stadiums, you're still only around ~50k humans max Nitpick, but there are 8 stadiums (college football) in the United States with a capacity of over 100,000
- bratbag 4y agoIt's a big leap from seeing some people on here making that mistake, and extrapolating that to everybody.
- izacus 4y agoI'll bet a lot that it's MOST people that do this kind of reasoning mistake pretty commonly - when programming, reading analytics or even debating why someone doesn't deserve to have a mini iPhone.
- hnfong 4y agoRight. IMHO that's one of the few valid reasons to require some basic understanding of mathematical rigor in programming. The mindset that you need to deal with ALL, 100%, of the cases instead of just settling for "90% is prolly enough" is very important to counterbalance these common tendencies you speak of.
- DharmaPolice 4y agoA while back where I work we had a serious mess up with our phone system and to cut a long story short, a number of people had requested a callback through an automated system and those call backs never happened. It was a lot easier to say "A small number of customers - less than 5% were affected" vs "We seriously let down more than a thousand people many of whom are vulnerable." 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.
- talkingtab 4y agoA very good point. Thank you.
- mejutoco 4y agoExactly. Whenever a corporation is fined the news make sure to shows x millions of dollars. I have never seen it in percentage of revenue in that context.
- vlovich123 4y agoEveryone seems to be overlooking that a total count + percentage gives you the number of active users. That’s generally not something broken down at a product level. So you can give percentage or absolute. If you only ever give percentage then you don’t have a problem and it gives you a sense of scale. If you give absolute numbers it doesn’t + you might accidentally give percentage out in a different venue allowing competitors to have a sense of how well your product is doing. 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”
- flangola7 4y agoCorporations can safely be assumed to be evil until proven otherwise.
- unyttigfjelltol 4y agoin fairness, ignoring the 3% edge cases generated by power users... that is a central feature of Google's offerings, i.e., Google Workspace. When Google was 'cool' everyone simply figured those edge case features were 'uncool', now they demand them anyway.