Every morning, property managers across the country open pricing dashboards and find recommended rents waiting for them. The software has already gathered information, run its calculations, and produced a number for each kind of apartment. A renter sees the result later, perhaps as a renewal offer that went up again.

But the hidden machinery behind that number is starting to get detailed house rules.

This case moved again on July 6, when the Justice Department filed a proposed consent decree with Willow Bridge Property Company, a large residential manager. The proposal would bar Willow Bridge from using pricing software that relies on competitors' sensitive nonpublic data. It would also restrict information sharing with other landlords, require a compliance program, and impose monitoring conditions if the company uses an uncertified third-party pricing system.

But the Federal Register carried a proposal, rather than a finished judgment, on July 16. A judge still has to approve it. Multifamily Dive reported that the agreement contains no financial penalty and that Willow Bridge had not responded to its request for comment before the outlet's July 8 report appeared.

This proposal follows settlements with RealPage and several large property managers in the same federal antitrust case. And together, the documents offer a useful look at what algorithm accountability means once it moves past broad principles and into a working product.

The argument behind the rules

The government's amended complaint alleges that competing landlords fed detailed, nonpublic leasing information into RealPage's revenue-management systems. That information included executed rents, concessions, applications, lease terms, and future occupancy. The software then used the combined pool to recommend prices back to landlords that would otherwise be competing for renters.

That complaint also alleges that parts of the system made accepting a recommendation easier than rejecting it. Managers could accept recommendations in bulk, while departures required individual explanations and could trigger review by a pricing adviser.

But those claims remain allegations. The consent decrees did not produce findings that RealPage or Willow Bridge broke the law, and neither company admitted liability through these agreements. RealPage has denied wrongdoing.

And the two sides give different accounts of the software. Associate Attorney General Stanley Woodward said companies cannot use sensitive data and algorithms “to produce market aligned pricing.” RealPage CEO Dirk Wakeham said his company helps housing operators make “informed, independent decisions in a complex housing market.”

That disagreement will continue. The more revealing material sits inside the remedy.

Follow the data into the screen

That court-approved RealPage decree separates model training from runtime operation. The distinction sounds technical, but it is easy to translate.

Training is how a model learns patterns from past information. Runtime is what happens when the finished system uses current inputs to produce today's recommendation.

Under the decree, RealPage cannot use a competing property's current nonpublic information when its software generates a rent recommendation. Model training may use limited backward-looking nonpublic data only after it is at least 12 months old and no longer tied to an active lease. Affected models must be retrained, and the decree restricts how narrowly they can use geographic information.

And the rules reach the interface that a property manager touches.

An auto-accept feature must use parameters set by the landlord. Limits on upward and downward price movement must be symmetrical. The product cannot obstruct or discourage a manager from rejecting a recommendation. RealPage also cannot reward customers or its own staff for accepting recommended prices.

That matters because a human override can exist on paper while the screen quietly pushes people away from using it. A button, a required explanation, or an extra approval step can shape behavior before anyone announces a formal policy. Here, the remedy treats interface friction as part of the accountability problem.

And the decree adds a monitor, compliance officer, audits, access to records, and employee interviews. The initial monitorship lasts three years and may be extended by up to 18 months. A rule about data has someone assigned to inspect whether the data rule is being followed.

What this cannot tell a renter

These agreements do not establish how much any renter may have overpaid. They do not guarantee that rents will fall. Housing prices move with local supply, demand, financing costs, construction, household income, and decisions made by individual owners.

The agreements also leave other disputes open. Private class actions, state cases, and local laws follow their own paths. Willow Bridge's decree remains a proposal during the public-comment and court-review process. RealPage's decree permits some older nonpublic data in model training under specific limits, so the settlement is more detailed than a simple ban on private data.

And renters still cannot inspect a landlord's proprietary dashboard or model. The people receiving the price have less visibility than the people generating it.

That gap is exactly why the mechanics matter. The decrees give ordinary readers a better set of questions: Who supplied the data? How old is it? What does the default do? Can a manager reject the recommendation without fighting the software? Who checks the answer?

When your next renewal arrives, compare it with current public listings and concessions at the same property and nearby buildings. Save dated screenshots. Then ask in writing whether pricing software influenced the offer and what human review or override was available.

You may receive a partial answer. Keep the record anyway. A rent recommendation can arrive as a clean number, but the system behind it is made of choices. These settlements show where some of those choices can be inspected, limited, and changed.

Public sources