The rigor behind the advisory.
Curator Index is built on original quantitative research into how the world's most exclusive residential markets set value. The work is documented in the open, and it is the reason our counsel can rest on evidence.
What began as faculty-advised work in the Dartmouth College Department of Mathematics has become a working method used every day. The two papers below set out the framework and its application. For the full methods, results, and discussion, read the works themselves.
Before turning to the work itself, though, two short notes on how we think about the problem.
A comparable is conventionally chosen for proximity and square footage, then adjusted by intuition. At the top of the market that method fails by construction, because a home of this caliber is a particular combination of architecture, siting, light, condition, and provenance, and the combination is very likely unique. The parts, however, are not. A cantilevered hillside position, a specific quality of renovation, a certain relationship between a house and its view: these recur across the market even when the whole never does. Evidence transfers at the level of those features rather than at the level of the address, which makes comparability a question of how much prior evidence bears on this particular combination. Understood that way, it becomes something to measure rather than assert, and its scarcity becomes a figure that can be reported rather than a limitation to be worked around quietly.
Most valuation claims in this industry report a single figure, and it is often accuracy against public records, which measures how faithfully data was transcribed rather than how well a home can be priced. The useful questions are narrower. What is the error on properties like this one, in this submarket. How many such sales produced that figure. And does a stated confidence range hold at the rate it promises. We publish our own answers, including the ones that flatter us least, because a claim that cannot be checked is not really a claim.
A Multimodal Machine Learning Framework for Luxury Real Estate Valuation
The foundational research, advised by faculty in the Dartmouth College Department of Mathematics. It shows that structured records, listing photography, and language, learned together, sharpen how value is understood at the top of the market.
This document could not be displayed inline.
Open the PDF ↗Bringing the Framework to Life
The applied continuation, translating the foundational findings into a working method now used in live advisory engagements across Los Angeles.
This document could not be displayed inline.
Open the PDF ↗The research continues, extending into new questions of how value forms across the world's most exclusive markets. Future findings will be published as separate work rather than folded back into these papers.