Agreed on all of this, including YouTube. For example, I am regularly surprised that YouTube's frontpage will recommend some completely obscure band with 1,000 followers, and it turns out that I like it! YouTube has tons of information on me, but I don't know how it has enough information about that band to know that it would match my taste.
You're assuming that recommendation engines are tuned to *your* success metrics. What if it's exactly perfect for whatever metrics that LinkedIn has tuned it for? In other words: a discussion about recommendation engines is incomplete without addressing the competing needs of the various consumers of the system.
I agree with 99% of this excellent piece, with the one exception that YouTube algorithm isn’t near good enough, either. The fact that they are incentivized to “keep me watching” is what also incentivizes creators to have click-baity content or video watch paths that inevitably lead to controversy and/or political polarization. A better YouTube algorithm would (in-time) result in better content AND be better for society.
I am constantly surprised at how good YouTube’s recommendation algorithm is. I don’t use instagram or TikTok but I imagine their recommendation algorithms are similarly good. In any event job recommendations isn’t LinkedIn’s business. LinkedIn’s business is extracting fees from companies who mindlessly pay their employees’ premium membership, and ads. Incentives drive outcomes.
Spot on, except I am less happy with the YouTube algorithm - it keeps suggesting videos I've already watched, on a ridiculous assumption (stated by YouTube developers) that already watched videos have high probability of being watched again. This might work for the content like music clips, but for someone interested in exploration and discovery it's so annoying, especially given that there is no way to turn this kind of suggestions off.
Amazon appears to earn more $$ margin from selling ads these days than from selling products, so even if you search for **exactly** the product you want they will bury the search result beneath 10-12 "recommended" (i.e. sponsored) products that seem to be deliberately not exactly what you are searching for.
Totally agree! Bad recommendations are almost always an incentives problem, not a technology problem. If the slot is really just ad inventory, it will optimize for revenue/internal agendas, not for fit. For careers, the right metric is outcomes (reply/interview/offer) with coherence across level + skills + comp band, not clicks. This is exactly the bar we’re trying to hit at Refer.
Pleasantly surprised to learn I'd apparently be an excellent candidate for roofing project management
Agreed on all of this, including YouTube. For example, I am regularly surprised that YouTube's frontpage will recommend some completely obscure band with 1,000 followers, and it turns out that I like it! YouTube has tons of information on me, but I don't know how it has enough information about that band to know that it would match my taste.
My top recommendation was Hungarian Voice Acting...
I don't speak Hungarian.
whatt???? insane!
You're assuming that recommendation engines are tuned to *your* success metrics. What if it's exactly perfect for whatever metrics that LinkedIn has tuned it for? In other words: a discussion about recommendation engines is incomplete without addressing the competing needs of the various consumers of the system.
agree. LinkedIn does not do things for its user
I agree with 99% of this excellent piece, with the one exception that YouTube algorithm isn’t near good enough, either. The fact that they are incentivized to “keep me watching” is what also incentivizes creators to have click-baity content or video watch paths that inevitably lead to controversy and/or political polarization. A better YouTube algorithm would (in-time) result in better content AND be better for society.
I am constantly surprised at how good YouTube’s recommendation algorithm is. I don’t use instagram or TikTok but I imagine their recommendation algorithms are similarly good. In any event job recommendations isn’t LinkedIn’s business. LinkedIn’s business is extracting fees from companies who mindlessly pay their employees’ premium membership, and ads. Incentives drive outcomes.
Thanks, Auren - I enjoyed reading the rant :)
Spot on, except I am less happy with the YouTube algorithm - it keeps suggesting videos I've already watched, on a ridiculous assumption (stated by YouTube developers) that already watched videos have high probability of being watched again. This might work for the content like music clips, but for someone interested in exploration and discovery it's so annoying, especially given that there is no way to turn this kind of suggestions off.
Amazon appears to earn more $$ margin from selling ads these days than from selling products, so even if you search for **exactly** the product you want they will bury the search result beneath 10-12 "recommended" (i.e. sponsored) products that seem to be deliberately not exactly what you are searching for.
Never thought of you as an athleisure guy.
I'm trying so hard to avoid suggesting that they learned that from the Dems. Oops failed.
Feel free to delete. Couldn't resist a wide open pitch like that.
They recommended me for position titled “Former Board Member”
Totally agree! Bad recommendations are almost always an incentives problem, not a technology problem. If the slot is really just ad inventory, it will optimize for revenue/internal agendas, not for fit. For careers, the right metric is outcomes (reply/interview/offer) with coherence across level + skills + comp band, not clicks. This is exactly the bar we’re trying to hit at Refer.