Every comp is a building
that already leased
Comps are the right tool for underwriting an acquisition. They are the wrong tool for deciding what to build, and the reasons are structural rather than sloppy.
Comps report the sticker price
A comp records asking rent. What a landlord actually collects is asking rent minus the concessions it took to fill the unit — two months free, a waived amenity fee, a move-in credit. Landlords hold the headline flat and discount underneath it, because the headline is what the next appraisal reads.
Asking rent
Advertised. Nothing obliges anyone to update it when it stops being achievable.
Effective rent
Net of free months and credits. Routinely well below the headline, and rarely published.
The gap is not a rounding error — in commercial leasing the same deal can land in two comp databases at two different numbers, because each one treats concessions differently. Underwrite a program on the headline and the error compounds through every unit you build at that price.
A neighbourhood that already happened
A comp set drawn today describes buildings leased over the last year or two, to people who were looking then. Your building opens in three to five years. Between those two dates a neighbourhood gains restaurants, loses bodegas, gets a school that starts working, gets a train that runs on weekends — and starts attracting a different person entirely.
Every developer has had the same argument in a conference room: is this a trend or is this temporary? Comps cannot settle it, because the households who would settle it have not moved yet.
The spread is the decision
A market study tells you the average renter pays a premium for outdoor space. In a real generated population, a quarter of households will not pay for it at all and a quarter will pay three times the average.
Those are completely different buildings, and the average is the one number that cannot tell you which to build. Where the spread is wide you are choosing whose amenity to buy. Where it is narrow, it is table stakes and you have no choice at all. A single figure hides exactly the thing you needed.
Nobody records who walked away
A comp is a lease that got signed. There is no dataset of the people who toured and left, or who filtered your building out online and never toured at all — and those are the people a program gets wrong.
A generated household carries its deal-breakers, so switching one amenity off tells you how many households it removes from your pool entirely. You cannot run that query against a comp set. There is nothing to run it on.
Generate the demand you can’t survey
You cannot survey people who have not moved yet — so the only way to have them in the room is to model them, individually, with everything a program decision depends on attached: what they earn, what they will pay, what they need, and what makes them walk.
That is not a better comp. It is a different kind of evidence, and it has one property comps never will: you can ask it a follow-up question and change the building while you watch.
Comps tell you what leased. Prelived tells you who’s coming.
Development is art, not science. We’re not arguing.
Everything above says the evidence you’re given points the wrong way. None of it says your judgment does. A developer who has leased six buildings knows things six hundred modelled households never will, and a model that argued with that would deserve to be ignored.
So Prelived doesn’t. Tell it you never build studios, or that in-unit laundry is non-negotiable in your buildings, and it obeys without negotiation — it is structurally incapable of returning a program that breaks one of your rules. What it does instead is price them: this rule costs you three points and four thousand a month, and that one is free, because the population already agrees with you.
Better colours, a bigger canvas, and a note on what each choice cost. The painting is still yours.
Prelived has not been run against a completed lease-up. Nothing here is validated against an outcome — it is defensible because every number shows the trail that produced it, not because a building has confirmed it yet.