3 ms·

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 databas

by judge2020 2y ago

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?)

rahimnathwani 2y ago
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.
sudosysgen 2y ago
> 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.
bobmcnamara 2y ago
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-fixing-algorithm-still-price-fixing https://www.ftc.gov/business-guidance/blog/2024/03/price-fix...
- Gunax 2y ago
  > 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?
  - bobmcnamara 2y ago
    > 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%28withattachments%29_0.pdf https://www.ftc.gov/system/files/ftc_gov/pdf/YardiSOI-filed%...
    - judge2020 2y ago
      > 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.

rahimnathwani 2y ago

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.

sudosysgen 2y ago

> 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.

bobmcnamara 2y ago

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

Gunax 2y ago
> 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?
- bobmcnamara 2y ago
  > 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%28withattachments%29_0.pdf https://www.ftc.gov/system/files/ftc_gov/pdf/YardiSOI-filed%...
  - judge2020 2y ago
    > 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.