Decision Intelligence · 60 US Big Cities · $100M Develop & Hold

Pick a city. Pick a horizon. Decide — and record it.

Expected return vs the complexity of running it and the complexity of getting in, at 3, 5 and 10-year holds. Every number traces to the graph.

One relational model

All 60 cities and every 3 / 5 / 10-year horizon are scored by a single model over one graph — not 60 forecasts stitched together.

Many predictions at once

Change the city, the horizon or the objective and the whole board re-scores instantly. Every market is re-predicted in parallel, the moment you ask.

Zero retraining wait

A classic ML model retrains for each new target. Pragma's relational method answers new questions in-context — no re-fitting, no waiting.

01 · Decide

hold horizon

02 · Return vs complexity

03 · The screen

04 · Decision log

method · sources · limits

Segments are fitted (k-means on 8 features), not asserted; each segment carries a posture that seeds the per-city verdict. Expected return = income yield compounding on rent growth + value trend, annualized per horizon; 3-yr values carry the ±12.6% backtested error band — 5 and 10-yr bands are wider and unbacktested. Run complexity = regulatory/tax regime (verified for OH, IL, NY, MA, CA, OR). Entry complexity = market depth for a $100M ticket + supply crowding + basis. Sources: Zillow, IRS, Census, HMDA/CFPB, HUD, FRED, state statutes. Residential indices proxy multifamily; asset-level underwriting needs CoStar/Yardi. Data analysis — not registered investment advice.