Comment by judge2020 | Hacker News Reader

Comment by judge2020 | Hacker News Reader

The private data mentioned in the lawsuits is about how, daily, realpage-set prices (and probably YieldStar as well) contribute their prices to a shared database. I wouldn't say this is actually "private" data because they only do so after rents are set and after they've appeared publicly on the properties' leasing portals and/or websites. It's basically a way to avoid Realpage from having to scrape their own rent data from their own customers' websites nightly (and from having to do so multiple times for each property that is pulling comparison data to use for the upcoming nightly rent determination).

note that realpage isn't using same-day rent prices to enable undercutting or colluding on prices amongst different properties, so you won't find Property As 2br2ba unit costing the exact same amount as neighboring propertys 2br2ba unit on the basis of preventing price competition.

As I understand it, the DOJ's basis is that it's illegal to use an algorithm that is both deterministic and also is using said "comparison data" from properties that use the same algorithm. Is this really a sound legal strategy? Algorithms are only algorithms because they're deterministic; if they're not then you've made some form of a ML model.

This sounds like apartment communities could only use an algorithm to ingest comp prices and make rent decisions if they commissioned a new bespoke algorithm that can only ever be used by them (what happens if the algorithm supplier copies and pastes code? Or delivers basically the same product with only changes to a seed value?)

AIUI it's not the sharing of info that's illegal, but the coordination of decisions. IIRC from an earlier HN discussion, Realpage's clients had to agree to accept the recommended rent at least 85% (?) of the time.

> Algorithms are only algorithms because they're deterministic; if they're not then you've made some form of a ML model.

Not at all, many classical algorithms and data structures are and must be probabilistic. Classical quicksort/quickselect is the best example.

I think it's just a case of the case being easier to argue this way.

Their argument is that the Sherman act says no-doing-X. Then folks used computers to do X. This is just price fixing with extra steps.

https://www.ftc.gov/business-guidance/blog/2024/03/price-fix...

> It’s also irrelevant that the algorithm maker isn’t a direct competitor if you and your competitors each agree to use their product knowing the others are doing the same in concert.

I didn't understand the implications of the FTCs decision here. Presumably no one is colluding to use the algorithm together. They are each independently clients of the algorithm creator.

So is the first user to buy the software okay? But then each subsequent user is colluding with the former clients?

> Presumably no one is colluding to use the algorithm together. They are each independently clients of the algorithm.

If you replace "the algorithm" with "our mutual friends pricing advice", it's the same old blatant price fixing clearly prohibited.

And no need to presume: the FTC arguments are quite clear they were advertising to take each landlords data and tune the pricing as if they owned all properties themselves. It's page 23 in the PDF, 11/23 in the memorandum.

https://www.ftc.gov/system/files/ftc_gov/pdf/YardiSOI-filed%...

> If you replace "the algorithm" with "our mutual friends pricing advice", it's the same old blatant price fixing clearly prohibited.

The difference between this comparison and what's actually happening is that the friend doesn't actually price decisions for "today" with information they learned from other competitors that also asked them for price decisions "today".

Instead, it's like a friend giving you a preset excel spreadsheet where you plug in your competitors' price history and your price history and it spits out a number. Is it really illegal for everyone to use the same spreadsheet with the same math in it?

Another child comment of my original top level comment suggests the collusion was that, to use the algorithm at all, a rental property had to accept its output price at least 85% of the time.

> Instead, it's like a friend giving you a preset excel spreadsheet where you plug in your competitors' price history and your price history and it spits out a number. Is it really illegal for everyone to use the same spreadsheet with the same math in it?

Yes, if the competitors are colluding to share their price history with competitors for the purposes of market manipulation.

> Another child comment of my original top level comment suggests the collusion was that, to use the algorithm at all, a rental property had to accept its output price at least 85% of the time.

That's one of several smoking guns. Even if the recommendations were not required to used but only for initial pricing, that's also a smoking gun.

No need to refer to comments, read the FTC memorandum, it cites their historical cases.

> Yes, if the competitors are colluding to share their price history with competitors for the purposes of market manipulation.

The issue is that this sharing is only done so that they don’t have to waste bandwidth and cpu cycles scraping prices they already know from their competitors’ websites.

Is it collusion if everyone is shopping their competitors prices?

Or is it illegal to use the same math equations in the same order?

and might I add, imagine this algorithm is possibly “the most perfect algorithm at generating positive rent growth” to which there is no improvements possible, so a government body forcing you to create your own algorithm could be de-facto government sabotage. It would be like the government preventing railroad operators from using the same ticket pricing calculation minimumPrice = miles * costPerMile + flatCosts

> Is it collusion if everyone is shopping their competitors prices?

Not at all! Competitors are free to publish their prices, to whoever they want, for loads of reasons, but colluding to raise prices isn't allowed. Realpage advertised this as a pricing cartel as a service though.

Prices for rent used to be much more negotiable than they are today, in part because of Realpage. From personal experience, landlords before Realpage were more willing to take a lower offer on something that had been vacant a while. But maybe they should have kept it vacant if they knew they were already the best price for similar accomodations.

>and might I add, imagine this algorithm is possibly “the most perfect algorithm at generating positive rent growth” to which there is no improvements possible...

A perfect algorithm with global knowledge is what the act is trying to prevent, because tilts the market entirely in favor of supply against demand. Same effect as a monopoly, which is why it's illegal.

A lot of this stuff was hashed out in cases around a hundred years ago, by people who tried and failed similar hope-theres-a-loophole arguments, and there are plenty of loopholes they could've hid behind, but instead:

"According to the complaints, RealPage’s own marketing materials make clear that it “utilizes the competitive data” of competitors, to allow those competitors to “outperform the market,” As RealPage allegedly put it, it offered clients “the ability to ‘outsource daily pricing and ongoing revenue oversight’ to RealPage,” allowing RealPage to “set prices for its clients’ properties ‘as though we [RealPage] own[ed] them ourselves.’”

Reply on news.ycombinator.com