Comment by endymi0n | Hacker News Reader

Comment by endymi0n | Hacker News Reader

Parent Full thread

The issue here is that no matter how "engaging" discovery is, it's not a viable business model on its own.

I'm working on the topic day to day in an adjacent field (movie & TV show discovery) and I'm currently not aware of any startup that has managed making money with the discovery experience itself.

It takes a lot of effort, data and persistence to get reasonable algorithms churning out something sensible and then, nobody wants to pay for it. If anything, it's usually an upsell to monetizing the content itself, which is why the only serious contenders to me here are basically Youtube and Spotify.

Even Netflix "had it all" with the million dollar recommender prize they set out in 2014, and I know their data science team is top notch. But if I scroll through their experience with my personal account in 2021, I basically still get "what all others are currently watching in {your_country} now" and I'm less than thrilled.

And that's exactly because once you've got the perfect algo together that would be able to recommend you that niche movie Netflix features you've never seen in the interface that you're _really_ interested in (I can recommend "Last Breath" in that regard), the VP Content comes in and tells everyone something like "hey folks, whatever you do, make sure you're only promoting our originals, because they cost us far less in licensing... oh and let's give an extra in-your-face boost to King's Gambler to give it that extra buzz that lands us in Variety"

Discovery is hard, and once you made a name there, it's rather easy to screw it up by monetizing it.

> Even Netflix "had it all" with the million dollar recommender prize they set out in 2014, and I know their data science team is top notch. But if I scroll through their experience with my personal account in 2021, I basically still get "what all others are currently watching in {your_country} now" and I'm less than thrilled.

IMO, Netflix's recommender system suffers from the simple fact that the company itself doesn't want to collect accurate user ratings and make recommendations with them. If they did that, then a not insignificant portion of the content and deals they invested in would get memory-holed due to bad ratings or lack of interest, and that would make the company and its leaders look bad.

Why don’t they want to collect accurate user ratings? Also, they know whether or not you finish something, how many episodes you binge, etc. There are many rating-independent signals.

can you elaborate on this? can't they collect ratings for internal use and not surface them, if that's the issue?

they should be able to tune the recommender so that it doesn't completely forget about lower-rated content the way you're saying

Netflix used to do that. You used to be able to rate programs between 1-5 stars and see the average rating. I believe it also recommended other programs to watch based on what you rated highly.

They did away with that system and replaced it with their custom recommendation engine. I believe, without any evidence, they did this to make bad shows look better. They once hosted a documentary about sex workers (can't recall the name, sorry) that had an average rating of around 2 out of 5 stars, and after the recommendation system change, it turned into a "75% match" on my account.

A movie may have an average rating 2/5 but still be a 75% match for you personally, there's no discrepancy here.

That sounds like functionality tunnel vision, to me. Matching users to movies isn’t an abstract metadata networking exercise— the larger goal is recommending things users enjoy so they keep watching. I’m far less likely to suffer a bad movie if it closely aligns with my interests. Poor representations of my passions and crafts frustrate the hell out of me.

Last.fm are perhaps the most successful discovery-mainly service. They had a ‘radio’ thing with a licensed catalog, but apparently finally dropped it in 2014. Otherwise, they are somehow chugging along for almost twenty years. However, I don't know how well their finances are, lately. The service doesn't seem popular in the past ten years or so, and I myself have just drifted away (in favor of less predictable and comfortable discovery by chance or by research, actually).

> it's not a viable business model on its own

Do you mean not a viable VC-funded business, or not viable at all? From context, I think you mean the former. But there are plenty of "traditional" businesses (e.g. clubs and radios) that make a living through discovery.

Maybe the problem is VC funding then? Maybe one day we will hear of some bootstrapped strartup coming out of IndieHackers?

they don't really make their money through discovery, they make money through things they sell alongside discovery.

for radio, that is ads or subscriptions, but radio revenue has been down like the rest of traditional media. for clubs, that is the multiple rounds of overpriced drinks, and maybe a cover.

Spotify's monthly cost is comparable to about two or three mixed drinks, maybe one in an expensive city like New York or Miami. People blanche at paying any higher; remember how people reacted to Tidal being comparatively expensive?

> they don't really make their money through discovery, they make money through things they sell alongside discovery.

that doesn't contradict the assumption that it's a viable business. Search engines and social media services make money from ads but I don't think it would be fair to say that search engines or social media services are not viable business.

The specific claim was “viable business on its own.”

Really? In my experience clubs hire DJs and radio stations are typically national and have the record labels pitching new content to them. I’m not sure either are interested in the “discovery” we’re talking about.

> clubs hire DJs

yes and DJs can be seen as discovery facilitators as per grandparent's comment, so it is fair to say that this is a viable business, which is based on content discovery - or maybe I misunderstood what you're saying?

> radio stations are typically national and have the record labels pitching new content to them

leaving aside the fact that you just described a very narrow subset of all commercial radios - how does this make radios not related to content discovery?

maybe we just need some local travel agency, people can by something like travel of music

Reply on news.ycombinator.com