TE
科技回声
首页24小时热榜最新最佳问答展示工作
GitHubTwitter
首页

科技回声

基于 Next.js 构建的科技新闻平台,提供全球科技新闻和讨论内容。

GitHubTwitter

首页

首页最新最佳问答展示工作

资源链接

HackerNews API原版 HackerNewsNext.js

© 2025 科技回声. 版权所有。

Ask HN: How can Netflix and Amazon Prime be this bad at recommending content?

49 点作者 roboy超过 2 年前
I’ve literally spent 20 minutes for the xth time in a row looking for anything that remotely captures me on both platforms and in the end did not watch anything. I’ve spent hundreds of hours binging series and movies on both platforms, but neither seems able to use this to suggest anything I like. I’ll probably cancel at least Netflix.<p>It would be so easy: - Store series I‘ve watched, show me new seasons when they release. - neither does it reliably. - Don’t keep suggesting things I have watched, except in a „watch again“ category. Especially using new pictures, getting me to click on the same stuff only to realize that I have seen this already.<p>Given the billions invested in these UIs, how can they fail so miserably? What am I missing? This makes no economic sense … or am I just a weird edge case, and this kind of UI works for the majority?

38 条评论

boffinism超过 2 年前
If you spent 20 mins looking, it&#x27;s not their recommendation system that&#x27;s failing. It&#x27;s that you don&#x27;t like their content (or how their content is portrayed).<p>The recommendation system is the thing that shows you stuff without you needing to browse for 20 mins.
评论 #34755834 未加载
评论 #34754957 未加载
评论 #34755482 未加载
评论 #34754795 未加载
评论 #34754787 未加载
hombre_fatal超过 2 年前
Maybe recommendations are harder than we think because our tastes are too arbitrary and unpredictable. Maybe overlapping with someone else&#x27;s tastes actually predicts nothing about whether there are any more places we will overlap, so it&#x27;s not the simple data processing problem that we wish it to be.<p>I&#x27;ve tried all sorts of &quot;people who liked X also enjoyed Y&quot; book and movie lists and it never seems any better than a generic list of decent books and movies.<p>One of my favorite movies is The Arrival. To me, it&#x27;s one of the best love stories ever told. Yet when I meet other people who really liked the movie, that wasn&#x27;t the main draw for them. Perhaps it was the intrigue of the time travel or watching interesting characters navigate a complex conflict or perhaps they liked the provocative questions that it explores.<p>Perhaps two people liking The Arrival says little about what they have in common at all, thus there&#x27;s actually no data to go off beyond recommending both parties more critically acclaimed movies. I suspect that this might be the cold reality of recommendations.
评论 #34757464 未加载
评论 #34756214 未加载
评论 #34755063 未加载
Cyberdogs7超过 2 年前
As someone that has worked within one of these orgs, it&#x27;s because the recommendation system has been captured by contract negotiations and forced merchandising. You are not seeing things they think you would like. You are seeing the things they want you to watch.
评论 #34755039 未加载
评论 #34758134 未加载
hauxir超过 2 年前
Totally agree, that&#x27;s why I gave up on it myself and created an aggregator.<p><a href="https:&#x2F;&#x2F;tomatotree.tv" rel="nofollow">https:&#x2F;&#x2F;tomatotree.tv</a>
评论 #34755401 未加载
评论 #34758951 未加载
评论 #34754772 未加载
评论 #34754812 未加载
评论 #34754851 未加载
评论 #34754735 未加载
coenhyde超过 2 年前
IMO Netflix and Amazon Prime&#x27;s problem is they don&#x27;t surface their good content enough. They mix it with their shitty content. They should make it stand out. HBO are good at this. When they make something good, that they know is good, they will rub your face in it with large high quality images.<p>I don&#x27;t have time to shift out the good from the bad with Netflix.
评论 #34754960 未加载
13of40超过 2 年前
Something else conspicuously missing from these services is a checkbox to make it never show me something again. Why do I have to scroll through an endless line of &quot;nope&quot;, &quot;nope&quot;, &quot;nope&quot;, &quot;definitely not&quot;, &quot;no&quot; every night, then look at all the same stuff tomorrow? Does someone get paid when I have to sacrifice a brain cell for each of these minor decisions?
评论 #34754879 未加载
评论 #34754798 未加载
评论 #34754875 未加载
atum47超过 2 年前
I do believe they have some kind of &quot;monetization&quot; going on regarding what they recommend you. Even YouTube apparently.<p>They suggest the same thing over and over under &quot;different&quot; categories. I they go try to force &quot;popular&quot; content on you.<p>One of the reasons I made this joke&#x2F;rant project<p><a href="https:&#x2F;&#x2F;victorribeiro.com&#x2F;recommendation&#x2F;" rel="nofollow">https:&#x2F;&#x2F;victorribeiro.com&#x2F;recommendation&#x2F;</a>
teknofobi超过 2 年前
Because priority number one is getting you to watch their originals, so that you will keep subscribing as the market for movies produced outside the streaming services gets more competitive.<p>The big events gets featured on the home page, for anything you are probably better of searching for directors or reading reviews, and then finding where you can see it.
评论 #34754892 未加载
xyzsparetimexyz超过 2 年前
A big problem on Netflix is that they don&#x27;t use the original poster or description for movies, instead coming up with their own. I find that these generally don&#x27;t do movies justice.
评论 #34754715 未加载
评论 #34754822 未加载
评论 #34754981 未加载
Raed667超过 2 年前
Maybe the problem is that they don&#x27;t carry more of the content that you like?
curun1r超过 2 年前
Their goal is to keep you subscribed, not to keep you entertained.<p>They know they don’t have enough premium content to keep you as a subscriber if that’s all you watch. That kind of content usually costs a ton to create. So they need to have you watch filler content as well. And just like social media companies have taken advantage of the slot machine effect to increase addictiveness, so too do streamers. If everything you watch is good, you’ll end up watching less and be more likely to cancel.<p>You’re assuming that the goal of a recommendation algorithm is to recommend the thing you’d most want to watch at that specific moment. It’s not. The goal is to maximize their retention. It’s amazing how many of life’s annoyances become instantly explainable when you realize that your interests are not aligned with those of the companies you’re dealing with. Your happiness will always be a secondary concern to their profits.
dgeiser13超过 2 年前
Human beings cannot make good recommendations most of the time. What makes you think an algorithm can?<p>My major problem with almost all recommendation engines and add-ons is that most of the time I want a recommendation that is orthogonal to the last thing I watched.<p>If I just watched a Japanese yakuza crime thriller I <i>don&#x27;t</i> want you to recommend another Japanese yakuza crime thriller. Give me something as good as what I just watched but different. But that&#x27;s me. Many people just want to watch Hallmark Christmas romances until their heads explode.<p>That&#x27;s probably why I still use Criticker. Because it gives you one choice across multiple genres.<p><a href="https:&#x2F;&#x2F;www.criticker.com&#x2F;" rel="nofollow">https:&#x2F;&#x2F;www.criticker.com&#x2F;</a>
ulizzle超过 2 年前
My feeling is that Payola is most likely the reason why the recommendation algorithms are so bad rn.
irvingprime超过 2 年前
Yeah. I especially find it interesting when Netflix says something you couldn&#x27;t pay me to watch is a 92% match with my profile (or however they phrase it).<p>But part of the problem the algorithm has is that content is of very uneven quality. I look at very recent things that, a few years ago, I would have wanted to see, and scroll on by. The first Avengers movie? Kind of rushed but acceptable. Thor Love and Thunder? Garbage. (Not Netflix examples, just examples)<p>We want to see new content but have learned from recent experience that it&#x27;s not worth the trouble. The algorithm simply can&#x27;t understand that.
ZeroGravitas超过 2 年前
I got a tip from someone here and shifted to tracking what I want to watch with a seperate system trakt.tv.<p>This lets it recommend across different services and removes incentives to promote their own content (I&#x27;m assuming someone at Amazon and Netflix is trying to goose their own stats by boosting the content they self create, or that they pay the least for)<p>A link (and a free recommendation for an interesting show):<p><a href="https:&#x2F;&#x2F;trakt.tv&#x2F;shows&#x2F;giri-haji" rel="nofollow">https:&#x2F;&#x2F;trakt.tv&#x2F;shows&#x2F;giri-haji</a><p>It lets you do nerdier searches like by director, writer, actor and similar to openlibrary.org it has some weird old stuff too.
评论 #34754905 未加载
HellDunkel超过 2 年前
All streaming platforms i tried have this problem. Maybe the ranking of shows is commercially funded? Also annoying: netflix shows multiple cover images. Most annoying automatic playback of preview (including sound) in the menu.
评论 #34754785 未加载
评论 #34754861 未加载
er0cksc超过 2 年前
Read an article about how Netflix stages in house CDN &quot;boxes&quot; directly in key ISP locations, directly on the switch. Certain high ranking shows are spread through the CDN for responsiveness. Some of the recommended shows are the ones staged close to the watcher to eliminate lag. <a href="https:&#x2F;&#x2F;about.netflix.com&#x2F;en&#x2F;news&#x2F;how-netflix-works-with-isps-around-the-globe-to-deliver-a-great-viewing-experience" rel="nofollow">https:&#x2F;&#x2F;about.netflix.com&#x2F;en&#x2F;news&#x2F;how-netflix-works-with-isp...</a>
评论 #34755207 未加载
评论 #34755197 未加载
Lio超过 2 年前
I think Netflix needs an <i>&quot;I really don&#x27;t like this or programmes like it&quot;</i> option.<p>...come to think of it so does YouTube.<p>I&#x27;d love to be able to search for a particular genre and exclude everything that comes up.
评论 #34754956 未加载
评论 #34754918 未加载
InitialLastName超过 2 年前
As others have mentioned, part of the goal is to induce you towards the service&#x27;s self-made programming. They&#x27;re rolling the dice that you will get hooked on something if they show you enough options, and your time spent searching for content (as long as it doesn&#x27;t lead to an unsubscribe) is cheaper for them than time spent consuming content.<p>FWIW, Spotify has similar issues in that they can&#x27;t seem to figure out that new music from artists I&#x27;m &quot;following&quot; should be a slam dunk for the &quot;new music for you&quot; list.
评论 #34755095 未加载
jstummbillig超过 2 年前
Is it thinkable that after &quot;hundreds of hours binging series and movies&quot; you have simply exhausted the options?<p>Me, I haven&#x27;t found a second The Wire yet, no matter how often a recommender claimed I have.
评论 #34754934 未加载
评论 #34754872 未加载
yowzadave超过 2 年前
For me personally, the recommendation engines choose the wrong aspects of shows that I like. They focus too much on genre, and not enough on writing qualities, variety, surprise, etc. Just because I liked one “gritty suspenseful murder mystery” (or whatever the genre is) does not mean I will like another! The execution could be wildly uneven in different programs that fall in the same subject matter bucket. I’d rather see a recommendation that can figure out my tastes across genres.
oneoff786超过 2 年前
You are asking for recommendations and they probably don’t have anything. They’re not going to say that. So you get things you don’t want and things you have already seen.
MoSattler超过 2 年前
If you are looking for movie recommendations, nothing beats movielense <a href="https:&#x2F;&#x2F;movielens.org" rel="nofollow">https:&#x2F;&#x2F;movielens.org</a>
trog超过 2 年前
I have thought for years Netflix (for me, at least) would benefit massively from a &quot;never show me this show again&quot; button.<p>I continually see the same bunch of shows when browsing that I will never, ever watch. They take up space and add to cognitive load and ultimately just annoy me each time I see them in the list.<p>I&#x27;m sure this weighting would also be very useful from an algorithmic recommendations view as well.
prirun超过 2 年前
For Netflix&#x27;s DVD service, pre-streaming, their recommendation system was awesome. You know, the one where they had the contests and paid $1M to any group that could do better by some percentage?<p>IMO, they don&#x27;t use anything like that now for streaming. It&#x27;s as others have said: this is what we want you to watch, not what you necessarily will like.
peterhi超过 2 年前
For Amazon the issue for me has always been that aside from the things I shop for myself I get asked to buy things for other people or presents for family members<p>Amazon doesn&#x27;t know who I bought something for so making sense of why I have bought 4 padlocks (my father in law), a set of sauce pans (wife), a light novel (myself) and rechargeable batteries (myself)
评论 #34754915 未加载
CrypticShift超过 2 年前
Mass media platforms are very shallow when it come to &quot;managing personal information&quot;. Netflix is a disaster in that regard. and yes this kind of UI works for the majority: To be honest, the average person, normally content with switching on a TV, couldn’t care less.
acd超过 2 年前
Recommendation algorithms maximize the time you stay and interact with the platform. As long as you stay on the platform and consume content its working.<p>Its a bit the same as dating algorithms. If the algorithm was to good there would be no paying users of the service.
shahbaby超过 2 年前
When I was growing up every kid watched more or less the same cartoons and played the same games. Life was simpler.<p>Now as an adult there&#x27;s too much out there and too little time.<p>I suspect that this is a big part of why people feel there&#x27;s nothing good to watch.
评论 #34755076 未加载
MoSattler超过 2 年前
I actually went back to pirating movies&#x2F;shows and host them on Plex.<p>I sure have a much smaller collection than any streaming service, but it&#x27;s actually all content I like, and I actually watch stuff instead of just browsing for 20min.
cramjabsyn超过 2 年前
Maybe they push content that is cheaper to license optimizing for some sort of license-cost-per-viewer-hour metric
imachine1980_超过 2 年前
contrasting whit YouTube or tiktok is harder to Know why people like series ,less content, series become good after 40 minutes, people let their content run, do you want a background series or are u seeing in the momentfully ckncentrated, lees feedback look all the way down<p>Edit: are you seeing alone or whit other
JohnFen超过 2 年前
Honestly, I have yet to see a recommendation system that works very well.
评论 #34756296 未加载
pdntspa超过 2 年前
Stop relying on recommendations and figure out some humans that curate
IG_Semmelweiss超过 2 年前
its also possible they wont show you content you really want because its licensing is expensive, and instead they nudge you to content with higher % rev split
Scalene2超过 2 年前
Because it&#x27;s profitable to recommend shit.
ruined超过 2 年前
maybe you don&#x27;t want to watch tv
encryptluks2超过 2 年前
These services get paid to promote content. Did you think they are going to promote legit content?
评论 #34754716 未加载
评论 #34754886 未加载
评论 #34754711 未加载