AI Unit Mix Optimization

The Problem

Unit mix decisions are guesswork—leaving NOI/IRR on the table in every deal

Organizations face these key challenges:

1

Weeks of manual comp pulls and spreadsheet modeling for each site, then assumptions go stale before approvals

2

Overbuilding the wrong unit types leads to slow absorption, discounts, and broker-driven repricing cycles

3

Unit mix and pricing recommendations vary by analyst/broker, making outcomes hard to reproduce or defend to IC/lenders

4

Market shifts (rates, migration, new supply) aren’t incorporated fast enough to adjust mix, phasing, or pricing

Impact When Solved

Higher NOI/IRR from better demand-fit unit mixFaster lease-up/sell-through with optimized pricing and absorption forecastsScale market analysis without scaling headcount

The Shift

Before AI~85% Manual

Human Does

  • Gather comps, listings, and broker intel; manually reconcile conflicting data
  • Build/maintain spreadsheet models and run limited scenario sensitivities
  • Make unit mix decisions based on experience and anecdotal demand signals
  • Prepare IC/lender narratives and defend assumptions

Automation

  • Basic reporting tools pull static comps and market summaries
  • BI dashboards visualize historical data with minimal forecasting
With AI~75% Automated

Human Does

  • Set objectives and constraints (target IRR/NOI, risk tolerance, affordability requirements, design constraints)
  • Review AI recommendations, challenge assumptions, and approve final mix/phasing/pricing strategy
  • Handle exceptions (unique assets, regulatory edge cases) and manage stakeholder communication

AI Handles

  • Continuously ingest and clean multi-source market + geospatial data and detect regime shifts
  • Predict property values, achievable rents/prices, and absorption by unit type and submarket
  • Run constrained optimization across thousands of unit-mix/pricing/phasing configurations
  • Explain drivers (feature importance, scenario deltas) and generate IC-ready outputs with auditable assumptions

Real-World Use Cases

Free access to this report