Prelived generates ten households for every unit you’re building, as individuals rather than as percentages. Everything below follows from that one decision.
The population is generated first as distinct individuals, then clustered, and the segments are whatever comes out. So is how many there are.
5 segments across 600 generated households. Widths are demand share.
33 & 38 · SENIOR ENGINEER & DATA SCIENTIST · SEATTLE
31 · UX DESIGNER · WILLIAMSBURG
36 & 34 · SMALL BUSINESS OWNER & SPEECH THERAPIST · WINDSOR TERRACE
27 & 27 · TATTOO ARTIST & LINE COOK · BED-STUY
26 · PRODUCTION ASSISTANT · ASTORIA
Nothing above was chosen. This run produced 5 segments; most sites land at four to seven, and three or nine are both legitimate results.
The modeled share of synthetic prospects that will not rent without each one. We publish no dollar figure per amenity, and no preference ranking either: nothing we can observe says what a household would pay for a roof deck, or how it weighs one against a bike room. A hard constraint is narrower and it is something the household states — leave it out and those households are gone.
Modeled share of the synthetic prospects that will not rent without each amenity. A gate they state, not a preference we scored: leave it out and those households are gone from the pool entirely. The model does not claim how much anyone else likes it, and puts no price on any of them.
In-unit laundry
42%
Dishwasher
26%
Pet amenities
17%
So here is Nour, taken apart
33 & 38 · senior engineer & data scientist · remote/hybrid · From Seattle
This run was generated without narratives, so this household has no written voice. Every structured field below is generated.
Every PreLive carries a record of where it came from: which sources, which site-specific signals, and which slice of the modeled population produced that individual. It isn’t a summary of the method. It’s the trail itself, and it’s the same one a developer opens when they ask where a number came from.
The structured fields on this PreLive, as they sit in the record.
Every entry names the field it moved. A source that fed nothing isn't listed, because it didn't contribute.
amenity_priority_priors · hard_constraint_rates · between-slice income differences · new-construction rent premium · willingness_to_pay · affordability
Signals that move a Brooklyn project and appear in no dataset.
The Trajectory Principle.
A site delivering in 2030 is generated against the 2030 neighborhood, not the one outside today. Schools, transit, and retail move in between, and they change who considers the area at all.
Drawn from the Domestic relocation, dual-earner transfer group of the modeled population.
Prelived has not been run against a completed lease-up.
These are modeled households, not forecasts of individuals. No real resident is represented, and no one was surveyed.
The methodology is unvalidated. Some parameters are fitted from public tables and some are declared priors; the record says which, per household. That is not the same as an outcome confirming them.
Scope is Brooklyn ground-up rental, 20–150 units. Outside that, the sources and the priors are wrong.
Share of demand is share of your program. That’s the whole arithmetic, and it resolves before the architect is briefed.
21 units · 2BR
15 units · 1BR
10 units · 2BR
7 units · 2BR
7 units · Studio+
This is share arithmetic, not a recommendation. It ignores what each group can pay, so it will happily put units into a type the population cannot afford to fill. The Model prices every type against the same households and moves units accordingly — which is why the mix it returns is rarely this one.
A missed program pays for itself out of Year 1 — concessions to move the units nobody wanted, vacancy drag while they sit, and loss-to-lease on the ones priced under what someone would have paid. No figure is quoted here because we have not measured one; the mechanisms are the argument, and every one of them gets more expensive to reverse after schematic design.
Prelived enters at site acquisition and stays useful through schematic design — the window where program is still cheap to change. The sources, the neighborhood knowledge, and the methodology are tuned to Brooklyn submarkets. That’s a scope, not a soft launch.
Every figure on this page comes from one generation run against 94 West Street, Brooklyn, NY 11222 — a 60-unit ground-up rental delivering 2030-06-01. Methodology version 0.1. No external source was fitted for this run, so the population reflects modeled priors and the curated site signals.
Bring us a site