Price Confidence Layer

Project Overview

Price Confidence Layer is an AI product concept developed to help first-home buyers navigate conflicting property price signals. Rather than introducing another valuation estimate, the solution creates an explainable trust layer over Domain's existing AVM, comparable sales, agent guide history, and market data. It combines a Confidence Score, transparent reasoning, and an Expected Reality Range so buyers can understand not only what a property might cost, but how much confidence to place in the available signals. The concept is Phase 1 of a broader Decision Confidence Platform, with future extensions into inspection guidance, offer strategy, and auction support.

Key Analysis Areas

What I Did

• Defined first-home buyers as the primary MVP persona based on their limited market experience, high financial risk, and strong need for pricing confidence.
• Synthesized evidence from Reddit discussions, app reviews, news, regulation, and competitor analysis to identify unrealistic price guides, conflicting valuation signals, and a broader trust gap.
• Reframed the problem from a lack of pricing information to a lack of interpretability: buyers already have multiple numbers, but often do not know which signals to trust.
• Mapped the buyer journey before and after the proposed solution, identifying where uncertainty creates wasted inspections, time, and financial cost.
• Designed the product around three components: Confidence Score, Why This Price?, and Expected Reality Range.
• Created a redesigned Domain listing-card prototype that introduces confidence information without disrupting the existing agent listing experience.
• Designed the AI logic around comparable sales, AVM confidence, agent guide history, suburb trends, and listing metadata.
• Embedded explainability, human verification, fairness, and accuracy into the system design, then defined a phased roadmap and measurable success metrics.

Reflection

This project changed the way I think about AI product design. Users did not necessarily need more information: they already had agent guides, AVMs, bank valuations, comparable sales, and market commentary. The real problem was that these signals often conflicted. That shifted our product from prediction to interpretation. Instead of asking whether we could build a better price model, we asked whether we could help buyers understand the uncertainty behind the numbers they already see. It also made me more conscious of the relationship between AI, product design, and trust. In a high-stakes decision such as buying a home, accuracy alone is not enough. Users need to understand why a recommendation exists, how confident the system is, and where uncertainty remains. The Domain workshop reinforced that Product, Design, Data, AI, and Engineering may approach a problem differently, but the strongest solutions emerge when they stay aligned around the same customer problem.