Days-to-Lease — The Neuman Group

How long an active rental will take to lease — built entirely from signed leases (what actually rented, and how many days it sat), never from asking prices. Every for-rent unit gets an estimate. Below is the running accuracy test: on the day each unit is first seen we freeze its prediction, then grade it against the day it actually leases.

How the number is built

est_days = base × ( 1 + 0.30·(seasonal−1) + 0.50·(momentum−1) + 0.20·(price−1) )

BASE

The segment's own history

Median days-on-market over the last 12 months for that exact zip + bedroom count. Falls back to city, then county, then statewide when a segment is thin.

30% SEASONAL

The month it hits the market

Nov/Dec listings lease ~13% slower than the spring lull. Measured across every signed lease since 2019.

50% MOMENTUM

Is this market speeding up or slowing?

The segment's last-3-month median days vs its own 12-month baseline. Statewide it fell from 52 days (Dec) to 33 (Jul) — the heaviest-weighted leg because right now it moves the most.

20% PRICE POSITION

Asking rent vs the going rate

How the ask compares to what that zip + bed size actually signs for. Fair-or-under leases at baseline; only heavy overpricing (>+60%) meaningfully drags it out. The lever the landlord controls.

Forward test — the real experiment (units that lease AFTER we froze the prediction)

Backtest — proof it works today (units that had already leased the moment we snapshotted)

Calibration — predicted band vs what actually happened

Predicted band	Leases graded	We predicted (median)	Actually took (median)
No leases graded yet.

A well-calibrated model has the two right-hand columns close together in every band. This fills in as units lease over the coming days.