← All work
Drawing W-01 · Operations intelligence platform

Portfolio Command

Watches an apartment portfolio every day, flags problems before they hit the P&L, runs the manager incentive program, and turns operating data into reports leadership can act on.

In productionA multifamily operator · several hundred units
Drawing W-01-A · How it runsNot to scale
Sources
Power BI model
Property management system
Market survey entries
System
Daily rule checks
Incentive and report engines
AI trend analysis
Outputs
Dashboard
Manager and owner alerts
Monthly action reports

01 · The challenge

Property managers were paid partly on how their property performed. The numbers lived in Power BI and the property management system, but nobody was watching them continuously. Rising delinquency, missed evictions, expense spikes and rent charges that didn't match the books surfaced weeks late. Managers hand-built monthly reports in spreadsheets, and incentive payouts were calculated by hand.

02 · What we built

  1. 01A rule engine that checks 27 business rules across every property on each daily refresh, in three severity tiers
  2. 02A portfolio dashboard with property drill-downs and individual logins, with access enforced in the database
  3. 03An incentive engine and manager scorecard that calculate payouts automatically, with an audit trail
  4. 04A monthly action report where every flagged issue requires a written response from the manager
  5. 05An AI layer that finds trends before they cross a threshold and drafts recommendations for leadership

03 · Revision history

RevChange
01Rule engine, dashboard and alerts live (spring 2026)
02Incentive engine and manager scorecard
03Monthly action report with month-end freeze and audit trail
04AI trend report, mid-month budget recap and policy assistant (fall 2026)
06 · Inside the system

The parts that do the work.

  • W-01-A
    Rule engine

    27 rules evaluated daily across occupancy, delinquency, renewals, rent growth, NOI vs budget, expenses, work orders, evictions and lease expirations.

    • Info, warning and critical tiers, grouped into the owner's own business categories
    • A daily revenue check that rent charges match the books, with an immediate alert to the owner
    • Rules that can't be evaluated say so, instead of silently passing
  • W-01-B
    Portfolio dashboard

    Portfolio KPIs with month-over-month change, plus property drill-downs that show where the money is going.

    • Expenses by GL line, work orders, delinquency split by current and former residents
    • An 18-month lease expiration outlook that shows managers which months are crowded
    • Separate views for leadership and property managers, enforced server side
  • W-01-C
    Incentive engine

    Calculates monthly and quarterly manager payouts: occupancy and delinquency bonuses, disqualifiers, capital spending penalties, a leasing competition and in-house savings credits.

    • Paired with a manager scorecard, auto-filled where possible
    • Quarter history and a full audit trail
  • W-01-D
    Monthly action report

    Replaced a hand-built spreadsheet. Every flagged issue becomes an action card the manager has to answer in writing.

    • Financials auto-filled from the general ledger, with required explanations for budget variances
    • Goals that carry forward and get measured
    • Frozen at month-end close, so a report opened a year later shows exactly what the manager saw
  • W-01-E
    AI intelligence layer

    Twelve months of history checked for worsening trends, budget drift, peer comparisons and good news. The model groups findings into stories and drafts recommendations.

    • A portfolio view for leadership, plus draft emails per property to review and forward
    • A mid-month budget recap per property with projections and notes
  • W-01-F
    Alerts and digests

    A Monday digest gives each manager their property's issues and the owner the whole portfolio. Operational alerts catch stale data and failed jobs before anyone notices.

07 · How it holds up

Engineering decisions that make it trustworthy.

  1. 01Code writes every number. The model writes judgment only.AI output is checked and rejected if it contains digits, forecasts or invented specifics, with safe fallbacks. That's what lets leadership trust AI-written reports with financial data.
  2. 02What gets approved is what gets sent.Report releases are verified against the exact version that was reviewed.
  3. 03Rules are tested against each other.The dashboard and the email digest share rules written in two languages. 112 test cases have to agree before the Monday digest can send.
  4. 04Security lives in the database, not the screen.Row-level security with explicit tests that access is denied where it should be. An audit caught default permissions that would have allowed deleting records with no trail.
  5. 05Check the data, don't trust it.Validation caught a maintenance budget undercounted by 48%, one property reading another property's competitor data, and an owner alert address that couldn't receive mail.