RealPage revenue management

What this models

This example runs on the Shared-pricing-engine class serving competing landlord operators network: 11 components and 22 pathways between them. Every context in the Lab is a stylized model, never a reconstruction of any actual deployment, and each assumption behind it carries a provenance label.

Evidence base: 3 assumed · 13 published baseline. In the Lab, the shaded evidence band behind each headline readout draws its width from the least-established class below.

• assumed

  Shape, baselines, demand/capacity and privacy posture are drawn from this deployment's own record, not from a catalogue template. Verified against all 123 shipped orgs with the topology and dynamics signatures: zero collisions on either axis. Nothing here was chosen to move the solver; the read-out is reported, not targeted.

• baseline

  Posture, binding. No defendant admitted wrongdoing and no liability has been adjudicated: the vendor and four landlord operators each settled without an admission, the federal decree carries no fine, and two landlord defendants remained in litigation when the settlements were filed. Every complaint allegation drawn on this network - the pooled nonpublic data, the alleged asymmetry of the price band, the compliance pressure - is drawn as an allegation. The decree itself is proposed and pending the court's public-interest determination; the stipulation was entered in March 2026 and final entry was still outstanding in mid-2026.

• baseline

  The model-to-operator adoption pathway is drawn at full strength from a contested band, not a point estimate, and the record itself contains the dispute. The complaint reports a national acceptance rate of 40 to 50 percent for new leases across January 2017 to June 2023, which read alone would place this pathway much lower. The same complaint reports internal analysis finding nearly 60 percent of final floor-plan prices within 2.5 percent of the recommendation and more than 85 percent within 5 percent, which read alone would place it at the ceiling. The upper end is used because a price that lands within 5 percent of the recommendation is not an independently set price, and because the record's own guidance is that raw acceptance understates influence while the spread metrics overstate independence. A reader who prefers the acceptance figure should read this pathway one step lower.

• baseline

  The two record-to-record pathways between the shared dataset and the publication system are the deployment's defining structure and are drawn at full strength in both directions. The record describes the feedback as documented and central: every accepted recommendation became an executed-lease record in the shared dataset, which recalibrated the next day's recommendations for all participating operators. This is the only bidirectional store pair in the catalogue, and it is what makes this a coordination shape rather than a scoring shape.

• baseline

  The reconciliation check between the publication system and the shared dataset is drawn latent because the record documents its absence directly rather than by inference: the automatic-acceptance setting implemented daily recommendations with no human review and updated advertised rents automatically. The proposed decree's response is structural - automatic ranges become user-set and off by default - which is the clearest available evidence that nothing sat in that position before.

• baseline

  The band-and-escalation screen is drawn as a guardrail carrying a model-to-operator check at a low level rather than as a stronger inhibitor. Two documented properties set that value together: the band demonstrably operated, routing out-of-band changes to escalation or approval, and it is alleged to have constrained price decreases more tightly than increases. A screen that runs harder in one direction than the other catches only part of the error class in question, which is what a bounded check at a low level encodes. The proposed decree requires the band to treat increases and decreases alike.

• baseline

  The model-to-model self-loop is drawn at full strength, the catalogue's strongest, on a market-structure figure rather than a technical one: the complaint alleges at least 80 percent of the commercial revenue-management software market for multifamily housing, and the CEA estimated the software priced at least 10 percent of all US rental units. Independent reporting found that in one Seattle neighbourhood the ten biggest property managers ran 70 percent of units and all ten used the pricing software in at least some of their buildings - which is a concentration finding about the managers, not a claim that all of those units were algorithm-priced.

• baseline

  The independent second price read is drawn latent because the record's own account of the remedy explains why nothing occupies that position: regulators did not audit model correctness. The instruments they used instead were structural - severing the meetings, staling the shared data, coarsening its geography, re-symmetrising the band, re-inserting the human. No source in this record describes any independent evaluation of whether the recommended prices were right.

• baseline

  Both data-leaving pathways run live at a substantial level rather than inactive, which is unusual in this catalogue and is derived rather than assumed. The record documents data agreements giving the vendor lease-transaction data from over 16 million units nationwide, including units of operators who were not customers of the pricing products, so transaction records demonstrably left each operator's own governance. It separately documents vendor-hosted user groups and forums where competing operators' staff met, a channel the decrees expressly close. The privacy flag on the data crossings reflects unit-level nonpublic transaction records moving to a third party at scale; the record frames that sensitivity as competitive rather than as a personal-data finding, and no regulator in these sources made a data-protection determination.

• baseline

  Three operator classes are drawn because the record documents three groups with different employers, different authority and different documented behaviour: vendor advisers assigned to client portfolios who measured acceptance and pushed staff toward compliance, corporate teams that configured the band and the automatic-acceptance range, and on-site staff who could accept or override each recommendation. The adviser-to-front-line peer edge is the case's signature and has no analogue in a single-agency deployment: the operator group closest to the model worked for the model's vendor, and its documented function was to raise acceptance rather than to correct output.

• baseline

  Two input feeds are drawn because the remedy discriminates between them, not for shape. Competitor-derived data must now be at least 12 months old and not drawn from active leases, and the existing demand and supply models may not be trained with a geographic variable narrower than a state; published market and survey data is the class left unrestricted while fresh. Drawing one undifferentiated feed would make the central remedy unrepresentable on the board.

• baseline

  A heavy workload matched by ample capacity is the MiDAS-family counterfactual rather than a strained-frontline reading. Demand is the daily unit-level recommendation stream across a footprint the CEA put at at least 10 percent of all US rental units. Capacity is drawn high because the process the engine displaced was independent price-setting by each operator, and the proposed decree's own remedy is to restore it - automatic acceptance off by default, ranges user-set, band symmetric. That is decree text and antitrust theory about what the counterfactual was, not a finding that the deployment caused harm.

• baseline

  The court-appointed monitor's inbound and outbound pathways are drawn faint, deliberately neither absent nor established. The instruments are documented and specific: three years of monitorship after approval, a seven-year decree term, inspection rights, a written compliance program, a cooperation obligation, and four separate landlord decrees. What does not exist is any published compliance finding, and the decree awaits the court's public-interest determination. Oversight is therefore drawn as present and thin, and no claim is made that it has changed measured behaviour.

• baseline

  The remedy's adequacy is contested on the record and stays contested here. Eight public comments were filed under the Tunney Act; the American Antitrust Institute argued in its own commentary that the decree may not stop the vendor functioning as an algorithmic cartel manager - the institute's phrase, not the enforcement agency's; five state attorneys general objected to the adequacy of the private settlements. The agency defended the decree in its response to comments. Nothing on this network resolves that dispute.

• assumed

  Renters are not modelled here and no renter outcome is computed from anything on this diagram. This deployment makes no individual-level automated decision about any person: renters receive no notice, hold no appeal and never touch the engine, so the affected population is fully external to the loop. The documented harm surface is market-priced and diffuse, and the two figures that describe it - roughly 70 dollars per month per unit in algorithm-priced buildings and more than 3.8 billion dollars across US renters in 2023 - are model-based counterfactual estimates from an economic council, explicitly framed by their authors as a lower bound. They are never presented as measured overcharge and they live in the case file, not on the board. Fairness parameters from person-scored deployments in this domain do not transfer to this shape.

• assumed

  Unit counts are kept three-tier and never conflated: over 16 million units describes the vendor's data reach including non-customers; roughly 4.5 million units across roughly 600 customers is a vendor-reported priced footprint whose attribution the verification pass could not confirm in the source the dossier named, so it is carried in prose with a vendor-claim label and is never used as a load-bearing number here; and the CEA's at-least-10-percent-of-all-rental-units figure is a separate penetration estimate with its own denominator. Vendor performance claims - a marketed 3 to 7 percent revenue lift, one operator's report that its buildings outperformed their markets by 4.8 percent - are marketing and self-report, labelled as such wherever used.

What this example does not show

• BINDING FRAMING: this is an un-adjudicated coordination story under a decree still pending entry. No defendant admitted wrongdoing, the federal decree carries no fine, and two landlord defendants remained in litigation when the settlements were filed - so every complaint allegation drawn on this network, including the pooled nonpublic data, the alleged one-sided price band and the compliance pressure, stays an allegation rather than a finding. The decree is proposed and pending final entry: the stipulation was entered in March 2026 and the court's public-interest determination was still outstanding in mid-2026. The remedy's own adequacy is disputed on the record, and the phrase algorithmic cartel manager belongs to an antitrust institute's commentary, not to the enforcement agency, which defended the decree in its response to comments. The three-year monitor has published no compliance findings, so nothing here says oversight changed measured behaviour - only that instruments exist.
• Renters are not modelled here. This Lab reads institutional propagation only, and estimates no differential harm to served people. This deployment makes no individual-level automated decision about any person: renters get no notice, hold no appeal, and never touch the engine. The documented harm surface is market-priced and diffuse - roughly seventy dollars a month per unit in algorithm-priced buildings and more than 3.8 billion dollars across United States renters in 2023 - and those are model-based counterfactual estimates from an economic council, explicitly framed by their authors as a lower bound because non-participating landlords also raised rents in response. They are contestable in method, they are never measured overcharge, and they live in the case file rather than on this board.
• Deference here is a contested band, not a number, and the record contains its own dispute. The complaint reports a national acceptance rate of forty to fifty percent for new leases across January 2017 to June 2023, and in the same document reports internal analysis finding nearly sixty percent of final floor-plan prices within two and a half percent of the recommendation and more than eighty-five percent within five percent. The adoption pathway is drawn at the effective end of that band, on the reasoning that a price landing within five percent of a recommendation is not an independently set price. A reader who prefers the acceptance figure should read that pathway one step lower.
• Unit counts are three different denominators and are never mixed. More than sixteen million units describes the vendor's data reach, including operators who were not customers. Roughly four and a half million units across roughly six hundred customers is a vendor-reported priced footprint whose attribution could not be confirmed in the source originally cited for it, so it carries a vendor-claim label and is not used as a load-bearing number anywhere in this cell. At least ten percent of all United States rental units is a separate penetration estimate from an economic analysis with its own method. Vendor performance claims - a marketed three to seven percent revenue lift, one operator's report that its buildings outperformed their markets by 4.8 percent - are marketing and self-report.
• This case sits at the housing-affordability edge of its domain, and that limits what transfers. The people affected are market-rate renters, not benefits claimants or homelessness-services clients, and there is no person-level automated decision anywhere in the shape - so fairness parameters, screening thresholds and override statistics from the person-scored deployments elsewhere in this domain do not carry across to this network, in either direction.

Sources and evidence

What this example rests on, claim by claim. Every entry resolves to the same ledger the Evidence Registry publishes.

• The United States and ten plaintiff states allege in United States v. RealPage, Inc. (M.D.N.C., filed 23 August 2024, amended 7 January 2025) that competing landlords contractually fed a single vendor nonpublic, competitively sensitive information — executed new-lease rents, renewal offers and rates, lease terms, and occupancy signals — to train and run a common pricing algorithm that recommended rents back to all of them: the complaint alleges at least 80 percent of the commercial revenue-management software market for multifamily housing and data agreements reaching over 16 million units nationwide including units of landlords who were not customers, describes an 'Auto Accept' setting that implemented daily recommendations with no human review, a 'Governor' feature alleged to constrain price decreases more than increases, and vendor pricing advisors who monitored client acceptance and pushed property managers toward compliance, and reports a national acceptance rate of 40 to 50 percent for new leases across January 2017 to June 2023 against internal analysis finding nearly 60 percent of final floor-plan prices within 2.5 percent of the recommendation and more than 85 percent within 5 percent — allegations only, with no defendant having admitted wrongdoing and no liability adjudicated.

  empirical
  • Government US Department of Justice Antitrust Division, Complaint: United States and Plaintiff States v. RealPage, Inc. (M.D.N.C., 2024) https://www.justice.gov/atr/media/1365471/dl
  • Government US Department of Justice Office of Public Affairs, Justice Department Sues RealPage for Algorithmic Pricing Scheme that Harms Millions of American Renters (2024) https://www.justice.gov/archives/opa/pr/justice-department-sues-realpage-algorithmic-pricing-scheme-harms-millions-american-renters
  • Trade press Paul, Weiss, Practical Takeaways From the DOJ's Algorithmic Pricing Settlement (2025) https://www.paulweiss.com/insights/client-memos/practical-takeaways-from-the-doj-s-algorithmic-pricing-settlement

• The consent decree the Department of Justice proposed on 24 November 2025 edits the deployment's structure rather than its accuracy: competitor data used in models must be at least 12 months old and not drawn from active leases, the existing demand and supply models may not be trained with a geographic variable narrower than a state, the 'Governor' feature must treat increases and decreases symmetrically, automatic-acceptance ranges must be user-set and off by default, vendor-hosted meetings of competing landlords are barred, and a court-appointed monitor holds sweeping oversight for three years under a seven-year term with inspection rights, a written antitrust compliance program, and a cooperation obligation — with no fine and no admission of liability; the Proposed Final Judgment and Competitive Impact Statement were published on 5 December 2025 (90 FR 56286), the stipulation was entered 26 March 2026, and the Department responded to eight public comments on 8 May 2026, with final public-interest entry by the court still pending as of July 2026 and no compliance findings published by the monitor.

  empirical
  • Government US Department of Justice Office of Public Affairs, Justice Department Requires RealPage to End the Sharing of Competitively Sensitive Information and Alignment of Pricing Among Competitors (2025) https://www.justice.gov/opa/pr/justice-department-requires-realpage-end-sharing-competitively-sensitive-information-and
  • Government Federal Register, United States et al. v. RealPage, Inc. et al.: Response to Public Comments (91 FR 25373, May 8, 2026) https://www.federalregister.gov/documents/2026/05/08/2026-09147/united-states-et-al-v-realpage-inc-et-al-response-to-public-comments
  • Trade press Hogan Lovells, Proposed DOJ settlement provides guidance on use of competitive information in algorithmic pricing tools (2025) https://www.hoganlovells.com/en/publications/proposed-doj-settlement-provides-guidance-on-use-of-competitive-information
  • Trade press Wilson Sonsini, DOJ Settles Its Algorithmic Price-Fixing Case Against RealPage (2025) https://www.wsgr.com/en/insights/doj-settles-its-algorithmic-price-fixing-case-against-realpage.html

• The White House Council of Economic Advisers estimated in December 2024 that rental pricing algorithms cost United States renters more than $3.8 billion in 2023, roughly $70 per month per unit in algorithm-priced buildings, with the software pricing at least 10 percent of all US rental units and nearly one in four multifamily rental units — model-based counterfactual estimates the council explicitly framed as a lower bound because non-participating landlords also raised rents in response, and never measured overcharge; separately, the Middle District of Tennessee preliminarily approved 26 private settlements involving 27 landlord defendants totalling $141.8 million on 21 November 2025 for a class of renters of covered properties between 18 October 2018 and 21 November 2025, over objections from five state attorneys general, with a second batch of roughly $218 million announced in 2026.

  empirical
  • Government White House Council of Economic Advisers, The Cost of Anticompetitive Pricing Algorithms in Rental Housing (2024) https://bidenwhitehouse.archives.gov/cea/written-materials/2024/12/17/the-cost-of-anticompetitive-pricing-algorithms-in-rental-housing/
  • Trade press American Bar Association Antitrust Law Section, RealPage and Certain Landlords MDL litigation update (2025) https://www.americanbar.org/groups/antitrust_law/resources/newsletters/realpage-and-certain-landlords-mdl/
  • Trade press Multifamily Dive, Apartment owners to pay $218M in second batch of RealPage settlements (2026) https://www.multifamilydive.com/news/realpage-settlement-algorithmic-pricing/820745/

Where this connects

Institutional pressures in this domain

• Workload surge — Demand outruns staffing; per-case attention shrinks and review becomes triage.
• Austerity & recovery incentives — Cost-cutting and overpayment-recovery targets tilt the system toward denial and enforcement errors.
• Vendor opacity — The deploying institution cannot inspect the model, data, or update pipeline it is accountable for.
• Data & policy drift — The world, the intake process, and the rules change under a system trained on how things used to be — two mechanisms with different remedies: the statistical properties of what the system processes move (concept drift), or the mixture of inputs arriving in deployment differs from the mixture it was trained on (covariate shift).
• Compliance over substance — Paper controls (sign-offs, checklists) satisfy audits while the behavior they describe erodes.

All of them in context on the Housing & homelessness services domain page.

Levers available here and the patterns behind them

• Store less data — Data minimization
• Vet connections — Connection authorization
• Gate record entries — Human-in-the-loop write gating
• Check copied records — Reconcile copied records
• Peer sharing rules — Peer-edge governance
• Review on schedule — Oversight cadence & retrospectives
• Mark AI-written records — Provenance labeling
• Keep skills sharp — Deskilling-arrest mandate
• Assign a challenger — Structured dissent
• Gate vendor updates — Vendor quality gate
• Pause AI on alarms — Deployment circuit-breaker

Documented case histories

• One engine, many rivals: a shared rent-setting model and the record it writes back
• Allegheny Housing Assessment
• VI-SPDAT
• LA's coordinated-entry triage revision: the fix that needed fixing
• LA County Homelessness Prevention Unit
• Santa Clara County Homelessness Prevention System
• Homebase Risk Assessment Questionnaire
• Xantura OneView (predictive homelessness flagging)
• London's Strategic Insights Tool: one linked memory of rough sleeping read by every borough
• CHAI (chronic-homelessness prediction)
• Calgary Drop-In Centre: interpretable screening a shelter's own staff choose to check
• San Jose's camera car: a low-precision detector aimed at who is sleeping outside
• Imagine LA Benefit Navigator copilot
• SafeRent Tenant Screening Score
• CrimSAFE criminal-record tenant screening