Modernizing Commercial Real Estate Appraisal
A custom appraisal platform designed and built from scratch, with heavy UX work to replace a patchwork of tools the firm had outgrown.
A complicated mix of systems, sources, and handoffs
Partner Valuation Advisors (PVA) is a commercial real estate valuation and advisory firm whose appraisal workflows relied on a complicated mix of internal systems, third-party data sources, Excel, Word, and manual handoffs.
I helped define the UX and product strategy for a new platform intended to bring those workflows together: mapping the existing system, defining the product model, prioritizing the MVP, and designing the experience around a centralized Golden Record.
What the work ran on
For years the firm leaned on Lightbox as its primary research and valuation platform, with CoStar alongside it and a scatter of side tools carrying the rest.
The initiative
- ✳A new custom platform for research, comp selection, valuation, and reporting, not a redesign
- ✳Data migration from the existing system into one validated source
- ✳Delivery: started in discovery, heavy on research findings and feature estimation, then shifted into delivery
The complexity
Every property type needs its own information. An appraisal of a cannabis facility, a self-storage property, or a stretch of right-of-way doesn't ask for the same fields, comparisons, or reasoning as an office building. The platform had to hold 14 property types and still feel like one coherent product to every role using it.
The problem wasn't a bad interface. It was a fragmented system.
PVA's appraisal process depended on information moving between multiple systems and people.
Client and project information lived across internal tools like PLINK, while property and comparable-sales data depended on tools such as Lightbox, third-party sources, Excel workbooks, and manual research.
Different roles interacted with the same information in very different ways, from sales and project coordinators to researchers, appraisers, QC reviewers, leadership, and administrators.
The result was a workflow full of:
- ✳duplicate data entry
- ✳manual handoffs
- ✳disconnected sources of truth
- ✳repetitive research
- ✳spreadsheet-dependent workflows
- ✳limited visibility across project stages
The goal was bigger than a UI redesign: create a centralized platform that could eventually replace fragmented tools and become the foundation for PVA's valuation workflow.
The new system also needed to support PVA's transition away from Lightbox as its primary data tool.
Before designing, I mapped the system.
Rather than jumping directly into screens, I started by understanding how information actually moved through the organization. I mapped:
That exposed an important distinction:
Were organized around the company's systems.
Needed to be organized around the user's work.
That became the foundation for the product strategy.
Current State Mapping
Built a full swimlane map of the existing process across every role, from winning the lead through report delivery, documenting exactly which tool handled which step: Project Tracker and email for the bid, Lightbox and CoStar for research, PLINK for financials, Power BI and Chromeriver for tracking, Word for the final report. This is what made the fragmentation visible and countable, rather than just a general complaint.
Current-state swimlane: Lead/Bid → Project Setup → Research and Comp Search → Comp Ingest and Select → Appraisal Analysis → Report Delivery.
Stakeholder Interviews
Interviewed stakeholders across roles to ground the process map in what people actually experienced, not just what the workflow implied. Two threads came through especially clearly.
"Lightbox is largely bypassed for subject properties due to workflow inefficiencies, lack of data rigidity, and poor field mapping to the essential Excel reporting templates, with AI currently only being utilized for narrative elements."
"Limitations and clunky nature of the Lightbox system requires significant manual, redundant work in Excel and requires juggling multiple sources of information."
Each synthesis mapped pain points, opportunities, and recommendations by theme, e.g. Lightbox inefficiencies, data quality and standardization, and automation opportunities.
The people behind the workflowOne of the biggest challenges was that there wasn't a single “PVA user.” The platform needed to support several roles with very different responsibilities:
Sales / Relationship Managers
Identify opportunities, manage clients, and initiate bids.
Project Coordinators
Manage projects, verify information, and coordinate delivery.
Researchers
Gather and verify property, tax, and market data.
Appraisers
Analyze comparable properties, perform valuation work, and build reports.
QC Reviewers
Review work, identify errors, and ensure compliance.
Leadership
Monitor capacity, performance, pricing, and business outcomes.
System Administrators
Manage access, permissions, auditing, and governance.
The challenge wasn't to create seven separate products. It was to create one shared product model that could support seven different mental models.
Future State & MVP Scoping
Redesigned the full process end to end, then made the call on where to start. Research and Comp Search, and Comp Ingest and Select, were where the manual, redundant work was heaviest, so that's where Phase 1 (P0) drew the line, rather than trying to rebuild all six stages at once.
Future-state swimlane, with the P0 scope bracket over the two stages carrying the most manual overhead.
The product decisionWe didn't try to redesign everything.
The original scope touched nearly every stage of the appraisal lifecycle. Instead of attempting to recreate every existing workflow inside a new application, I helped frame the MVP around the areas where centralization could create the most immediate value. The product needed to:
- 01Establish a centralized database
- 02Reduce manual data entry
- 03Improve the quality and traceability of property data
- 04Make comparable-property research faster
- 05Preserve critical workflows that still depended on Excel and legacy systems
- 06Create a foundation that could support future automation and AI
The PRD ultimately defined success around cycle-time reduction, workflow duration, data quality, search performance, data auditability, and revenue/project visibility.
P0 covered Research and Comp Search and Comp Ingest and Select: a unified comp database with similarity ranking, duplicate detection before save, and a golden record concept for verified property data. Lead/Bid, Project Setup, Appraisal Analysis, and Report Delivery stayed in their existing tools for this phase.
Build the new system while creating bridges to the old one.
A common trap in enterprise modernization is assuming that if you're building a new platform, you can just eliminate the old tools. That wasn't realistic here. Appraisers had deeply established workflows in Excel, and the organization still depended on existing systems and external data sources. The platform needed to support:
The PRD specifically called for one-way Excel synchronization for report workflows while leaving room for more advanced bi-directional workflows later. That let the new product become useful without requiring the organization to change everything overnight.
Information Architecture
Mapped the full sitemap: Projects, Properties, Markets, Clients, and Account, each with their own sub-flows for research, comps, and reporting fields. The most important structural decision was the golden record: a property can have both a verified "golden" record and additional unverified events attached to it, created by merging, editing, or adding data over time, rather than every new upload silently overwriting what was already confirmed.
The Golden Record
PVA's property information could come from multiple sources, and those sources weren't always consistent. Rather than allowing every workflow to maintain its own version of the data, the new platform needed a centralized representation of the property and its attributes.
The Golden Record established a hierarchy for determining which information should be treated as authoritative. This wasn't just an interface decision. It influenced the:
The product requirements explicitly called for a Golden Record strategy to prioritize higher-quality data sources and improve data integrity.
Designing around work, not systemsOne system. Multiple mental models.
A researcher shouldn't have to think like an appraiser. An appraiser shouldn't have to understand how PLINK stores project information. And a project coordinator shouldn't have to jump between systems simply because the organization's technology evolved that way.
So I structured the experience around the user's job to be done, while maintaining a shared underlying product model. That meant designing reusable structures for:
Then layering role-specific actions and views on top.
Visual Direction
Before locking the visual system, pulled together reference points for color, typography, and the density of information a data-heavy platform like this would need to carry, from mapping interfaces to stat-card treatments to icon language.
Wireframes
Took the IA into low-fidelity flows for the core surfaces: the individual property record with space composition and unit-level detail, and the project creation flow that ties a property to the appraisal it belongs to.
One of the biggest opportunities was reducing the amount of information users had to enter manually.
Start from scratch → enter everything → verify later
Identify existing information → pull it into the project → verify → fill the gaps
Project creation needed to connect with existing PLINK information and automatically identify whether a property already existed in the central database. The system could then flag potential duplicates rather than forcing users to recreate existing records.
Make the user verify information instead of making them recreate it. A small interaction decision with a much larger systems impact.
Making comparable research more powerfulFor appraisers, comparable-property research was another major opportunity. The experience needed to support:
Instead of forcing users to move between separate tools, the new platform could bring those activities into one connected workflow:
The experience was designed to preserve the appraiser's existing mental model while reducing the amount of manual movement between tools.
High-Fidelity Design
Brought the wireframes into the final visual system, branded as Comp Works by Partner Valuation Advisors: a comp search builder with metro-area and radius filters, a market comps results view with map and list toggle, and the properties, project detail, and market research surfaces working together as one connected system.
Bringing the system together
The resulting experience centered around a connected product model rather than isolated screens.
Projects
A centralized workspace for managing appraisal assignments, properties, clients, status, deadlines, and project information.
Properties
A consistent representation of property information that could be reused across projects and workflows.
Comparable Data
A searchable, filterable workspace for finding and reviewing relevant comparable properties.
Golden Records
A trusted layer for maintaining authoritative property information.
Tasks & Roles
Role-specific workflows that made ownership and next steps visible.
Integrations
Connections to existing systems and external data sources so users didn't have to manually move information between platforms.
Three decisions shaped the product
Establish a source of truth
Rather than allowing every workflow to maintain its own data, the Golden Record created a shared foundation for the product.
Design around workflows, not organizational silos
The existing experience reflected the way PVA's technology had evolved. The new experience reflected how people actually worked.
Modernize without breaking existing workflows
The goal wasn't to force users away from Excel, PLINK, or external data sources overnight. The new platform created a gradual path toward consolidation.
Designing with the team
This wasn't a design exercise happening in isolation. I worked across product, engineering, and business stakeholders to translate a complicated set of requirements and technical dependencies into a product model that could actually be built. That meant continuously balancing:
My role moved between strategy and execution:
The foundation for what's next
PVA's first version was designed to be more than a replacement for today's tools. The architecture created opportunities for future capabilities including:
These weren't treated as MVP features to force into the first release. Instead, the product was designed with a foundation that could eventually support them.
P0 ships in October, P1 starts in November
The P0 scope, Research and Comp Search, and Comp Ingest and Select, is on track to deliver in October 2026. Phase 1 (P1) work begins the following month, extending the platform toward the remaining parts of the process: Lead/Bid, Project Setup, Appraisal Analysis, and Report Delivery.
The hardest part wasn't designing the interface.
It was deciding what the interface should represent.
In complex enterprise products, the UI is often the visible layer of a much larger system.
The most valuable design work happened before the pixels: understanding the relationships between people, data, workflows, technology, and business rules well enough to define a product model that could scale.
That is where I do my best work.
Domain complexity, fast
The same "comp" means something structurally different depending on whether it's a sale, a lease, or an expense comp, each with its own required fields and place in the appraiser's reasoning.
Becoming the strategic advocate
Not just executing what research surfaced, but driving the calls on data mapping and information architecture that would let a work-heavy platform stay usable instead of collapsing under its own complexity.
Borrowing proven logic
Some of that reasoning borrowed directly from how CoStar already structures comp data, adapting logic the industry already trusted rather than inventing a parallel system from scratch.
MVP scoping discipline
Scoping down to two of six stages meant resisting the pull to redesign the entire process at once, proving the model on the highest-friction stages first before extending it in P1.