Hacker News Reader

Comment by ignoramous | Hacker News Reader

I believe this behaviour is endemic to companies trying to morph into a "tech company" with MBAs leading the charge.

> Customer intimacy. Modern tech companies collect, store, organize, and analyze years of user data. This data is virtual gold, as it enables companies to run targeted ads and personalize the customer experience. The key difference between a customer walking into a Walmart supercenter and Amazon's online store is that Amazon instantaneously reorganizes the whole store (layout, displays, product offerings, etc.) in a way that's tailormade for that customer.

https://archive.is/wK1bC (why wework is not a tech company, hbr.org, 2019)

https://archive.is/G7rSR (why zomato.com is a tech company, hbr.org, 2021)

View on HN

The irony there is I have never ever bought anything because Amazon recommended it.

Amazon is better than Walmart because you can search for and buy what you want when you want and have it show up the next day, and the prices are generally good enough.

Reviews are a nice touch, but those just need to be “good enough” to differentiate cheap chinese crap.

It’s not rocket science.

Bezos at one time mentioned that 35% of all Amazons sales came from recommendations. That’s not an insignificant number.

But, are those recommendations based on what you're looking at? I could definitely see users purchasing one of the recommended items if they come to the site with an item in mind, but not necessarily a brand.

I often come to Amazon and search for something like "washers" and skim through the "recommended" items until I find a pack that's a good value. But from an attribution standpoint, that sale is more the result of "browsing the aisle" than a recommendation. It's not like they recommended hairspray, which i bought in addition to the thing I came there to buy.

Agreed. I think highly contextualized recommendations are the most helpful.

As another example, I frequently find the "often purchased together" section to be very helpful. That might be the only recommendation system that I've ever used more than once.

I would say that's an example of recommendations working. However, in my personal experience, I rarely find "general" recommendations helpful.

I wonder what specifically Bezos meant? Similarly, I often search for something specific and see "sponsored" results that are what I was searching for, with the same item right below it as part of the organic search results. Or, I'll browse for items, then come back to purchase it later seeing it under "recently viewed." Do those count? If so, I purchase recommended things all the time. If not, I don't think I've ever bought something they organically offered on their main page with 20 years of purchase and browsing history.

Looking at the recommendations right now, they're not very good. Maybe I'm a hard sell? Sure, sidewalk chalk. You might run out of that and want more. Stuff to rekey your lock. I needed that when I bought my house a few years ago...I'm not sure why I would need that now. An air conditioner support bracket. I bought one last year much earlier in the summer.

The entire isle was constructed on the fly just for you, including all the items you "compared", and their prices, and the thing you eventually "picked".

Tech companies play this game as well. This has always been a core part of Netflix's pitch: our recommendations will keep users tied to the platform and also reduce the cost to acquire IP as we can always find something the user is interested in.

I used to believe their pitch. However, it doesn't seem like it ended up working. People hear about a show and want to watch that specific show. The recommendations only do so much.

Interesting. I would disagree. They've shifted from supremely niche recommendations ("British murder mysteries during wwII") to everyone watching the entirety tiger king within the same 2 wk period and a coarse grain search for british titles not even providing an exhaustive list.

If anything they have taken recommendations and remade them in their own image/strategy/goals rather than our personalized watching behavior. Great examples continue to be Bridgerton, the Office which then had a meme-based resurgence, that one with Kevin Spacey, etc.

I just think they don't really care.

They care about retention, but they've realized that content >> discoverability for that.

People will find out about content. What they will not do is watch bad content, no matter how specifically and accurately it speaks to them.

You're agreeing with me. When they were serving up niche recommendations it was an intentional business decision. If they could convince people to watch niche things, it reduces their IP costs as those things are cheaper to license and they're also in a stronger bargaining position as they don't need any one specific show.

What happened is consumers didn't agree. People wanted to watch the Office, not another feel-good comedy Netflix recommended. That forced Netflix to abandon their original strategy and spend a lot more on content.

Gold is knowing what the customer wants/needs because then you can actually sell them something. But I doubt any amount of "data" will be able to figure that out.

I believe this behaviour is endemic to companies trying to morph into a "tech company" with MBAs leading the charge.

They're also going to pocket Amazon's fee.

>> with MBAs leading the charge.

So at what point on the MBA path does a person get forced into removing the logic part of the their brain?

MBA's seems to make decisions that are using an alien form that non-MBA's can not seem to understand, and perceive as illogical.....

Because MBAs will always hold two items in tension: profit and user desire.

If these two items are out of whack, an MBA will always favor profit and that too in incredibly creative, often marketing-based ways that rethink the business model. They will convince the user of what they need the user to believe to make the largest possible profit.

ICs often think of profit and user needs/desire but they usually hunt for a place where it overlaps and if it doesn't, they'll hunt again. Profit is demanded for value/convenience to the customer but it is often a 1:1 relationship.

Reply on news.ycombinator.com