Residential pricing decision system

Offer to exit,
as one decision.

A transparent laboratory connecting valuation uncertainty, seller conversion, buyer price response, inventory risk, and the acquisition offer. This evidence layer is location-neutral and built entirely from generated behavior.

CONTROLLED EXPERIMENT · AUDITABLE · REPRODUCIBLE

Evidence, before optimization

Predictive quality and behavioral recovery are measured only on the independent evaluation environment.

$32,777
Home-value anchor median absolute error
Lower is better
90.0%
90% interval coverage
Split conformal
0.184
Seller-acceptance Brier
Versus pooled-rate baseline
0.435
Sale-hazard log loss
Right censoring retained

What the models learned

Fitted responses are shown beside known simulator truth; truth is used for evaluation, never model fitting.

Contemporaneous value anchor on shifted evaluation homes
Contemporaneous value anchor on shifted evaluation homes · controlled evaluation environment
Seller response: fitted versus simulator truth
Seller response: fitted versus simulator truth · controlled evaluation environment
List-price response and cumulative sell-through
List-price response and cumulative sell-through · controlled evaluation environment

Three decisions under uncertainty

Open a case to inspect the selected acquisition offer and the no-sale markdown path.

PRICE

Healthy Demand Margin Protection

How much price should we preserve when demand is healthy?
$377,474automated offer
59.2%acceptance probability
$10,902expected profit / lead
11.9%approx. loss probability
Decision gate: best supported offer has positive risk-adjusted value
No-sale path: W0: hold → W1: hold → W2: hold → W3: hold → W4: hold → W5: hold → W6: hold → W7: hold → W8: hold → W9: hold → W10: hold → W11: hold → W12: hold → W13: hold → W14: hold → W15: hold → W16: hold
ABSTAIN

Stale Inventory High Carry

When does a deeper markdown beat another costly week on market?
$456,224best grid candidate
6.1%acceptance probability
$-925expected profit / lead
2.5%approx. loss probability
Decision gate: every supported offer has non-positive risk-adjusted value
No-sale path: W0: cut 5% → W1: cut 2.5% → W2: cut 1% → W3: cut 1% → W4: hold → W5: hold → W6: hold → W7: hold → W8: hold → W9: hold → W10: hold → W11: cut 2.5% → W12: cut 1% → W13: cut 2.5% → W14: hold → W15: hold → W16: hold
ABSTAIN

Sparse Support Downside Protection

How should uncertainty change both the offer and exit policy?
$622,610best grid candidate
3.0%acceptance probability
$-42expected profit / lead
1.7%approx. loss probability
Decision gate: every supported offer has non-positive risk-adjusted value
No-sale path: W0: hold → W1: hold → W2: cut 1% → W3: cut 1% → W4: cut 1% → W5: cut 1% → W6: cut 1% → W7: hold → W8: hold → W9: hold → W10: hold → W11: cut 2.5% → W12: cut 1% → W13: hold → W14: hold → W15: hold → W16: hold