Hacker News Reader

Comment by kodah | Hacker News Reader

> Yet as today’s letter points out, subsequent investigations by The Wall Street Journal, Reuters, and The Markup revealed that not only did Amazon employees working on private-label items have access to third-party data, but they routinely used it, even discussing it openly in meetings. “Amazon employees regularly violated the policy—and senior officials knew it.”

That's pretty incredible. I'm curious what kind of data this is; knowing the nature of the data would help us understand how platform operators can abuse their position.

View on HN

Consider: you are on the Amazon Basics product development team. If you do well, you will be promoted, given bonuses, and greatly financially rewarded. If you do poorly, well, Amazon let's go of 6% of employees in every org every year[0].

There are policies in place saying that you definitely can't look at the data around which third party sellers are doing well, selling lots of stuff. But also there's no auditing in place to see if you check it or not. It's all a big data warehouse and you have access to it.

And hey, you've heard everyone else is looking at this data. You're just taking a quick look to get ideas do you don't get fired. What's the harm? You'll help the company do well and won't have to polish your resume.

[0] https://www.businessinsider.com/amazon-performance-review-6-...

They should have used the Google excuse for the middle step.

Construct an algorithm that tells you what you want, based on the data that you're not supposed to be using, and then throw your hands up when anybody asks and say "Nobody really knows how ML models work."

Or before machine learning was an option: management consultants.

"Yes, this is a clean room implementation, guided with a ML model."

    import torch
    design_parameters = torch.chinesewall(existing_product_url)

Consider: not working for an unethical company.

PMs at Amazon, even people who need visa sponsorships, have options (and probably better options). Don't defend their poor choices. Engineers at Amazon have even more options.

> Consider: not working for an unethical company.

You mean one of the companies Amazon is driving out of business? For some reason they aren't hiring.

Capitalism (Society?) rewards unethical behaviors, as long as it's even just a little obfuscated. It's a competition for resources.

Ethical people act ethically and tend to stay far away from the unethical.

It follows that unethical people will this group together to commit their unethical conspiracies. See the Political parties, big oil, pharma, insurance, the fed, hedge funds, etc.

Ethical people have been self selecting out of these lucrative industries forever. What’s left are the clueless and the criminal.

I wouldn't say it's a capitalism problem but more of the erosion of a free market.

Anything that is counter to a free market rewards unethical behaviors. But politicians don't need to be bought. It's just what has been tolerated. This is a great enabler and you can see this with the SEC as just one example.

Regulatory capture occurs [1] and the megacorps infiltrate. People want to get ahead and they hide behind these megacorps, many of which have unethical behavior pushed by the leadership, investors, their boss, and sometimes even condoned by in-house lawyers. If there is no risk to shareholder value and no risk of personal charges, they will do it.

We need some criminal charges levied and some extreme fines. Maybe just set the fines to equal the money they stash overseas? It HAS to grossly outweigh the economic gain from the behavior, so that it's a big hit to shareholder value. [1]

[1] https://en.wikipedia.org/wiki/Regulatory_capture

There are always rewards for unethical behaviors, which is why people engage in them. This is true whether the framework is communism, capitalism, or marriage.

Competition encourages the unethical behaviors and punishes the ethical behaviors. Set the competition knob to 0, everyone slacks, crank it to 11, and everyone cheats.

Despite the fact that managing this balance is fundamental to a capitalist society, it's extremely typical for the latter half of this balance to be ignored, for competition to be framed as an unmitigated positive, and for the Goodhart's Law side of it to be completely swept under the rug.

Culpability should be attributed both to the individual and to the system. Otherwise it's easy to construct systems that bypass responsibility. Like we see here.

I feel like what differs is what type of unethical behaviors different systems encourage.

Capitalism encourages screwing over other people, being cutthroat in business, etc.

Communism encourages different faults, such as passivity, mob mentalities, and not taking responsibility.

Anarchism would discourage considering the effects of 2nd-order+ consequences.

The competition knob is but one knob among many. It's the one America happens to have at about an 8.5 (and one which I dislike greatly), but I don't think this problem is unique to capitalism.

>The competition knob is but one knob among many. It's the one America happens to have at about an 8.5 (and one which I dislike greatly), but I don't think this problem is unique to capitalism.

Given the levels of consolidation we've seen over the past 30 years or so, I'd say that a lack of competition also contributes to unethical behavior. In fact, the consolidation pursued by many corporations is predicated on swallowing up even potential competitors in an effort to consolidate their dominant position.

I'd say that was unethical too.

Every communist state has been rife with communism. Centralization is what increases the reward to corruption.

If you don't follow the perverse incentive du jour in a decentralized system, you'll get rolled.

We're drifting off topic, though. If Amazon heavily incentivizes bad behavior, should they be allowed to reap the rewards and discharge the blame? Absolutely not. They're culpable.

Unfortunately, liberal people are unable to parse most of your comment because it goes against their religion.

And if you are caught, you are a rogue employee.

Imagine you could get out of a robbery charge by claiming it was perpetrated by a rogue glove.

Even better, you can be like Johnson & Johnson and create a new crappy glove, then abuse bankruptcy laws to transfer liability to that new glove so that it becomes responsible for everything. Now you get away with committing robbery and you still get to keep your favorite glove too.

It’s data on what products are selling well, which they use to create clones and then sell under “Amazon Basics” and other brands.

It’s a problem because the small sellers take the financial risk trying to sell a new product, then when Amazon sees it’s doing well they make the clone and get the profits.

I am wondering if Walmart and Costcos store brands are allowed to use seller data!

This is such a recurring take, I wonder how we could deal with it in a constructive way.

To illsutrate my point, please take a look at the results: https://www.google.com/search?q=site%3Anews.ycombinator.com+...

I think it indicates a deeper problem, either on how we perceive these companies, or how Amazon succeeded in making itself look like a random traditional company. I don't know the solution outside a very prominent decade lasting lawsuit we could all point at, the same way we had the Microsoft antitrust lawsuit decades before.

TBH, Walmart is pretty fucking huge. Bigger than Amazon in retail revenue.

Amazon on one hand wants to pretend to be a third party market place to avoid liability, regulation, inventory holding risks and so on, and on the other hand wants to collect all the profit, data and exert control as a first party seller.

They can't have it both ways.

If you still have an Amazon account, ask for your data download from Amazon.com. You will probably be astounded to see that they retain just about every interaction you have ever had on their platform and products. I was a light user and received 88 zip files containing many files within them containing IP addresses, devices, and the most minute details. Now imagine how much they collect about sellers and products. You can assume it is everything they can collect.

TBF every company managing a web application should be doing the same.

You positively need to know what your users come to click on, where they bail, what features they never use, what browsers constantly drop out at this specific part of your website, what IP uses the bot that scrapes your most costly site wide search every millisecond etc.

Sure, any of these data could be derive from more focused, pinpoint check if you knew in advance what exactly you need to know. But let's be realistic, there's no way to anticipate every single use case and only collect the minimum data needed to fulfill it.

I don't see detailed usage data collection as a critical issue as long as it isn't sold/leaked/stored for decades/misused to harm. We can come up with so many cases where it has been effectively misused, but I'd argue it's like saying kitchen knives are misused again and again to kill people and we should stop their use. It's not realistic.

I am well aware that any commercial site tracks everything they can. I was merely trying to answer this guy's question about what data Amazon has on sellers.

Personally I don't like having to show my ID when I walk into a store and then have a guy follow me around with a clipboard writing down everything I look at, touch.. I run out of such stores which is why I deleted my Amazon account.

Point taken !

> Personally I don't like having to show my ID when I walk into a store and then have a guy follow me around with a clipboard writing down everything I look at, touch.. I run out of such stores which is why I deleted my Amazon account.

This...is kinda like the next frontier, with in-store camera tracking and linking it to the customer at checkout (either through the loyalty card or credit card token if applicable). We're already partially in that future IMO.

Stores already track your movement as you move around though bluetooth beacons.

https://www.nytimes.com/interactive/2019/06/14/opinion/bluet...

Yes physical stores are becoming as hostile as ecommerce sites. When i enter one, my phone stays in airplane mode, wear a mask to evade facial recognition and pay cash.

Not by any means defending Amazon here, genuinely curious.

How does this differ from say Walmart/Target/CVS etc., launching their own private branded labels?

Even they have access to all the data on the best selling products.

The difference is that Target is a retail store, not a marketplace. Target has already bought your products from you when Target sells them to customers, and aggregates data about its own sales of your products.

Amazon is a marketplace, and pays nothing to buy your products when they're offered for sale on its virtual shelves. It purports to provide an open and equal common area for companies (including itself) to sell products, yet only Amazon has access to the treasure trove of data about the behavior of consumers on its site.

> Target has already bought your products from you when Target sells them to customers

This is not true for all products. Some are sold under a pay-per-scan model, where the product is owned by the supplier up until the barcode is scanned at checkout, whereupon it is immediately purchased by the store and sold to the consumer.

For other products, suppliers may also be required to accept unsold goods for a full refund of the store’s purchase price.

Stores are marketplaces too, and shelf space – particularly the valuable shelf space like end caps – is not free.

There are products like that, but to my knowledge, there's not a lot of them and it tends to be certain specialized categories. The only example I can think of where the supplier will refund the product if it doesn't sell is live plants.

Another example of a product that commonly has a full refund clause is books.

But pay-per-scan is much more common. Examples include everything from trading cards to makeup to sandwich bread. The supplier will often restock these products themselves, so the store doesn't even need to track inventory.

I'm having trouble finding any solid numbers on what percent of products are sold pay per scan. I just know that in the retail stores I've worked in, the only examples I know of are live plants and possibly aquarium fish.

Target is an online marketplace for 3rd-party products. They don't go out of their way to advertise it, but they call the program Target+. You can find online-only 3rd-party items by searching their website. They ship the items to you, and you can return them in-store if there's a problem. Most people never notice the middleman supplier.

Walmart does this too. Newegg made the switch several years ago. Large retail outlets which do not have a "marketplace" online option are the exception rather than the rule.

Different companies do different levels of vetting, but the retail ecosystem is vastly more accessible to small-timers than it was in the 2000s.

The most salient difference, in my opinion, is that Amazon stood in front of Congress and lied about it.

If they instead told Congress that this is a long established retail practice, I would be more sympathetic to that argument.

Beside the fact that they lied to Congress about.. I guess there is an expectation.

If you put your products on Amazon marketplace, you are expected to join a marketplace. You don't necessarily expect for Amazon to also join the marketplace by using your data to compete against you.

Also you can use Amazon services and warehouses even for products that you do not wish to sell on Amazon website. Essentially using Amazon as a 3PL service. Again, you might not expect Amazon to use your data for their own products.

If you sell your products to Target, this isn't exactly a marketplace. This is a retail outlet. You know they can measure your sales, measure your competitor sales, etc. And you know they won't lie to you about doing it.

It's been a long time since some of the congressional hearings about anti-competitive practices in the software industry, and I'm afraid I'm too tired to look up the details at the moment, but one of the items I remember from the discussion was that Amazon sales employees were allowed by company policy to look at aggregated figures (per-market, per-industry, etc) for merchants on their platform.

That's fine as-is -- nothing wrong with that in principle -- but what I seem to recall employees began doing (and I can totally imagine how this would become an underground "sneaky trick" that employees would begin to learn and share with each other) was to create so-called aggregate groups that only contained a very small number of merchant businesses.

That way they could claim (and perhaps keep a straight face when saying) that they were only looking at aggregate group-level figures, but the statistical reality of the situation would have been that they were looking at information about a few -- or perhaps even only one -- business.

Reply on news.ycombinator.com