Construction Draw Management · Case study
Draw ManagementPNC Bank
Construction Lending · Draw Management · End-to-End UX

Reducing Construction Payment Cycles From 35 to 6 Days

$12.5M trapped in a 35-day payment cycle. NPS at −12. Three parties with zero shared visibility. This is what a complete redesign of the draw management system actually looked like.

ProjectConstruction Loan Draw Management
ClientPNC Bank
RoleSr. Experience Designer
Duration6 Months · 14 Sprints
PlatformReact Web App
Team1 PM · 6 Engineers · 1 CS
Executive summary

Construction draws took 35 days to pay out, 47% failed on first submission, and lenders spent 15+ hours reviewing each one across five systems. I designed PNC’s contractor and lender portals around one shared record: a checklist showing only the 6–8 documents a draw needs, a 4-stage status tracker, and a side-by-side review with automated pre-validation. Payment cycles fell to 6 days, first-time success rose from 53% to 92%, and NPS moved from −12 to +16.

Challenge

Contractors waited 35 days to get paid while their subcontractors demanded payment. Lenders reviewed each draw across five disconnected systems. Property owners had zero visibility. With NPS at −12, $85M in revenue was at risk.

Solution

Two products sharing the same data. Contractors get a dynamic checklist, real-time upload validation and a 4-stage status timeline. Lenders get a side-by-side review, automated pre-validation and high-risk draws surfaced first.

My role

Sole product designer, end to end. 40+ interviews, loan officers shadowed, contractors observed on job sites. 25+ screens, 8 core workflows and a 40+ component design system. 1 PM · 6 Engineers · 1 Content Strategist.

Impact

6 days payment cycle, from 35
~5 hrs lender review, from 15+
$12.5M capital deployed faster per quarter
45% contractor retention improvement

01.1 · The Domain

What Construction Draw Management Is

When a developer or contractor takes a construction loan, they don't receive the full amount upfront. They draw it down in stages — submitting a draw request to prove work has been completed, which triggers a lender review, inspection, and payment approval before funds are released. This process repeats every few weeks throughout a build. It involves three parties: the contractor waiting to get paid, the lender responsible for verifying work and releasing capital, and the property owner who funded the whole thing. When it works, cash flows on time and projects stay on schedule. When it breaks — and it was broken — contractors wait 35 days per cycle, subcontractors don't get paid, projects stall, and lenders hemorrhage review hours on a manual process that was never designed to scale.

01.2 · The Problem

A Broken Three-Party System

Draw management broke down because three groups needed different things, thought about the process differently, and none of them could see what the others were doing.

Contractors wait 35 days to get paid while their subcontractors demand payment now. Lenders spend 15+ hours manually reviewing each draw across five disconnected systems. Property owners have zero visibility into how their capital is deployed. With NPS at −12 and $85M in revenue at risk, leadership knew the product had to change — or lose to competitors offering faster alternatives.

Contractors
35 Days

Average payment wait

A cash flow crisis. Contractors called support 4–5 DM Sans per draw just to get a status update. 47% of draws failed on first submission.

Lenders
15+ Hrs

Manual review per draw

62% of review time was spent locating documents across 5 systems — not actually reviewing them. High-risk draws were buried in chronological queues.

Property Owners
Zero

Visibility into progress

"I'm funding a $500K project and I have no idea what's happening." Owners relied entirely on contractors forwarding updates — the weakest link.

Fig. 01.2aBefore: how a draw moved between the three parties
Draw request 47% fail first submission Status Radio silence for weeks · 4–5 support calls per draw Updates forwardedby the contractor No direct line to the lender Contractor 35daysAverage payment wait Subcontractors demand payment now Lender 15+hrsManual review per draw 5 systems Property owner Zero visibility into progress
Each party held part of the picture. The contractor could not see review, the lender searched five systems, and the owner heard only what the contractor forwarded.

My Role

  • Sole product designer, end to end: research, both portals, design system, delivery
  • Designed both contractor and lender portals across 25+ screens and 8 core workflows
  • Ran all discovery including interviews, shadowing, and survey analysis
  • Built the design system later adopted across 4 products in the Tech Innovation Group
  • Embedded in sprint planning with PM and engineering — wrote acceptance criteria and did design QA every release
Team: 1 PM · 6 Engineers · 1 Content Strategist

The Business Opportunity

  • Platform: a custom React web app on top of a proprietary loan management system, so every design decision had to work within the existing APIs
  • Differentiate on contractor experience — competitors still used manual processes
  • Unlock faster loan originations by reducing the draw cycle bottleneck
  • Reduce $85M at-risk revenue by recovering NPS from −12
  • Cut operational costs by eliminating preventable support tickets
  • Increase contractor retention and referral volume

Discovery

62%

of lender review time was spent locating documents across 5 systems — not actually reviewing them.

62% locating documents38% reviewing
Across12345disconnected systems
02.1 · Research & Discovery

Understanding the Root Causes

Before sketching a single screen, I spent 4 weeks inside the problem — interviewing, shadowing, and analyzing 6 months of historical data to find what was actually causing the 35-day cycle.

Phase 01 · Week 1

Stakeholder Alignment

A business process workshop to agree on research scope, success metrics and priorities before we went into the field.

6 participants · Current-state mapping · Success metric definition
Phase 02 · Weeks 2–3

Deep User Research

40+ stakeholder interviews. 3 loan officers shadowed during live draw reviews. 5 contractors observed on active job sites.

40+ interviews · Contextual inquiry · Shadowing · Workflow documentation
Phase 03 · Week 4

Validation & Benchmarking

3 competitive platforms analyzed. 6 months of historical draw data reviewed. 200+ contractor survey responses analyzed for patterns.

3 competitors · 6mo data · 200+ surveys · 50+ draw requests
40+
Stakeholder Interviews
200+
Survey Responses
50+
Draw Requests Analyzed
6 mo
Historical Data

User Segmentation & Prioritization

Prioritized by business impact, frequency, and design complexity before committing to scope.

Persona
Business Impact
Frequency
Complexity
Priority
Contractor
High
Daily
High
★★★★★
Loan Officer
High
Daily
Medium
★★★★★
Property Owner
Medium
Weekly
Low
★★★☆☆
Design first for Contractors and Loan Officers — then adapt for Property Owners in Phase 2.
Success Metrics — Defined Upfront With PM & Leadership
Draw cycle time
Baseline 35 days
Target <10 days
First-time success
Baseline 53%
Target >80%
Lender review time
Baseline 15+ hrs
Target <8 hrs
NPS recovery
Baseline −12
Target >0

Agreed before design work began. Tracked via Mixpanel event instrumentation + support ticket volume dashboards.

02.2 · User Personas

The People Behind the Process

Two primary personas built from 40+ interviews, contextual inquiry, and on-site shadowing. They aren't guesses about age or job title. They're based on what we actually watched people do.

M
Primary Persona #1
The Time-Pressed Contractor
Marcus Thompson
Marcus Thompson, 42
General Contractor · 15 years experience
Manages 3–5 projects simultaneously
Based in: Austin, TX
Behavioral Characteristics
Time-Constrained
Works 6am–7pm, handles admin before/after job site hours
Mobile-First User
85% of logins on iPhone — designed for thumb navigation
On-the-Go
Often working from truck/job site with limited wifi
Cash Flow Focused
Needs predictable payment timing to pay subcontractors
"I'm Trying To…"
  • ✓Get paid quickly to pay my subcontractors
  • ✓Submit draw requests without errors — first time
  • ✓Know exactly what documents are needed upfront
  • ✓Track status without calling the lender
"I'm Frustrated When…"
  • ✗Requirements are unclear or change unexpectedly
  • ✗I submit a request and hear nothing for weeks
  • ✗I get rejection emails with vague reasons
  • ✗The interface is hard to use on my phone
"

I'm not a paperwork expert, I'm a builder. Just tell me what you need and I'll give it to you — but I can't read your mind, and I can't afford to wait 35 days.

47%
First-time failure rate
4.2×
Support calls per draw
85%
Mobile (iPhone) logins
J
Primary Persona #2
The Efficiency-Focused Loan Officer
Working with the Relationship Manager
Jennifer Martinez
Jennifer Martinez, 38
Senior Loan Officer · 8 years experience
Reviews 10–12 draw requests per month
Based in: Dallas, TX
Behavioral Characteristics
Desktop-Primary
Reviews require large screen — 85% of logins on desktop
Process-Oriented
Values consistency and compliance above all else
Risk-Averse
Fears missing compliance red flags — audit anxiety is real
Time-Pressured
Portfolio of 50+ active loans — can't afford slow reviews
"I'm Trying To…"
  • ✓Review draw requests thoroughly but quickly
  • ✓Ensure all compliance requirements are met
  • ✓Catch errors or fraud before approving funds
  • ✓Communicate clearly with contractors about issues
"I'm Frustrated When…"
  • ✗Contractors submit incomplete requests (wastes my time)
  • ✗I have to toggle between 5 different systems
  • ✗Contractors call me repeatedly for status updates
  • ✗I miss a compliance issue and get flagged in audit
"

I don't need the system to make decisions for me — that's my job. But I'd love if it handled the grunt work so I can focus on actual judgment calls.

85%
Desktop logins
2.5
Draws reviewed/week
92%
Compliance checks done
47%
Requests with issues
4.2×
Support calls/request
02.3 · User Pain Points

In Their Own Words

The same frustrations kept coming up with all three groups. These aren't summaries. They're what we heard in interview after interview.

Contractors

"I never know when I'll get paid"

  • Waited 35+ days from submission to payment
  • Uploaded 12–18 documents with unclear requirements
  • Received rejection emails with vague, unhelpful messages
  • Called lender support 4–5 DM Sans per draw just for status
"I'm a contractor, not a paperwork expert. Just tell me what you need — but I can't read your mind, and I can't afford to wait 35 days."
Lenders

"We're drowning in manual review"

  • Spent 15+ hours reviewing each draw request manually
  • Toggled between 5 different systems to verify one document
  • No automated validation of document completeness at all
  • Compliance risk from inconsistent review processes
"Every draw is a treasure hunt. I'm searching through PDFs, cross-referencing budgets, and hoping I didn't miss anything critical."
Property Owners

"Where is my money going?"

  • Zero visibility into draw request status at any stage
  • Couldn't track how loan funds were being deployed
  • Relied entirely on contractors to forward any updates
  • No mobile access to project financial status at all
"I'm funding a $500K project and I have no idea what's happening. I feel completely in the dark about my own investment."
02.4 · Critical Insights

Three Findings That Ignited the Solution

After synthesizing 40+ interviews and analyzing 6 months of historical data, three insights emerged that shaped every decision that followed.

"I don't know what you need until you tell me I did it wrong."

— 8 out of 10 contractors expressed this exact frustration

The system kept requirements secret
  • ✕Hunt through email attachments for requirements
  • ✕Interpret legal language without guidance
  • ✕Guess at budget category mappings
  • ✕Submit and pray they got it right
My vision: visible, actionable, real-time
  • ✓Display all requirements upfront before starting
  • ✓Use plain language with examples
  • ✓Validate inputs before submission
  • ✓Guide users toward success
1
Key Finding

The Mystery Requirements Problem

47% first-time failure rate

Draw requests failed on first submission due to missing or incorrect documentation — entirely preventable with better upfront guidance. Contractors were guessing what to submit based on previous rejections, creating a costly trial-and-error loop that delayed payment by weeks.

"I never know what documents they need until they reject my request. I'm just guessing at this point." — Contractor, 12 years experience
Pre-submission validation Dynamic contextual checklist Plain-language explanations
2
Key Finding

The Black Hole Status Problem

4.2× average support calls per draw

Contractors had zero visibility after submission. Radio silence for weeks. The only way to know a request was received — let alone in review — was to call the lender directly. This generated enormous support overhead, eroded trust, and consumed loan officer time that should have been spent reviewing draws.

"I submit a request and then... nothing. Radio silence. I don't know if they received it, if it's being reviewed, or if it was lost." — Contractor
Real-time 4-stage status tracker Automated milestone notifications Direct messaging to loan officer
3
Key Finding

Lenders Want Assistance, Not Automation

Lenders need automated validation, not automated decisions. They want the system to handle grunt work — checking completeness, validating budgets, flagging anomalies — while preserving full human judgment on final approval. Automating decisions would create compliance and liability exposure that no lender would accept.

"I'd love if the system flagged potential issues BEFORE I start my review. That would save me hours." — Loan Officer, Senior
Automated pre-validation panel Exception-based review queue Smart flagging with context

Six Months of Data Confirmed the Findings

Historical draw data quantified each pain point and revealed the precise business cost of not fixing them.

Rejection Reasons

Missing documents38%
Incorrect budget coding24%
Incomplete forms18%
Compliance issues12%
Technical / Other8%

Top Support Ticket Reasons

Where is my request?42%
What documents are needed?31%
How do I submit a draw?15%
Technical issues12%

Loan Officer Time Breakdown

Document review35%
Budget validation25%
Communication back-and-forth20%
Data entry across systems15%

Key Insight: 73% of all support tickets were about status visibility (42%) and requirements clarity (31%) — both completely preventable with better UX. That one finding shaped the whole design strategy.

02.5 · Competitive Analysis

What the Market Was Getting Wrong

I evaluated 3 modern construction lending platforms to identify best practices and map the whitespace for differentiation.

Competitor A

  • Strong contractor mobile app
  • Clean submission interface
  • Weak lender review tools
  • No automated pre-validation
  • Chronological queue only
Key Gap

Fixed the contractor side and gave lenders nothing. Half the problem solved, then called done.

Competitor B

  • Excellent backend automation
  • Real-time status updates
  • Poor contractor UX — overwhelming
  • Automated decisions, not validation
  • Compliance risk for lenders
Key Gap

Automated everything including final decisions — creating compliance liability no regulated lender could accept.

Competitor C

  • Beautiful, modern UI
  • Clean dashboard layout
  • Mystery requirements unsolved
  • No proactive guidance at all
  • Feature-shallow despite polish
Key Gap

Their status tracker exposed 11 internal workflow stages that meant nothing to contractors. We collapsed these into 4 user-meaningful milestones — the core requirements problem still untouched.

The Gap We Owned

No competitor had solved the "mystery requirements" problem. Contractors across all three platforms were still guessing what to submit. This was our biggest and most defensible differentiation opportunity.

The Design Opportunity

Most platforms only cared about lender efficiency and ignored contractors. The real opportunity was designing for both at once, without shortchanging either.

The Strategic Insight

Best-in-class competitors offered real-time status but not proactive guidance. Adding contextual education and "why needed" explanations was entirely uncharted territory in this space.

Define & strategy

45 → 6–8

possible requirements, filtered to the relevant items for each draw context.

6–8 relevant items shownThe rest of the 45, hidden for this draw context
03.1 · Define Phase

Translating Research Into a Design Challenge

Once research was done and the three insights held up, we needed to agree on one clear problem before exploring solutions. This is where we got on the same page.

Point of View Statement

Construction contractors and lenders need a way to exchange draw requests with confidence and visibility because unclear requirements and zero status transparency are causing 47% of draws to fail on first submission and costing the business $85M in at-risk revenue.

How Might We…

The three design challenges the team aligned on before any ideation began.

HMW #1

How might we give contractors the exact information they need to submit correctly — before they start, not after they fail?

→ Addresses the 47% first-time failure rate
HMW #2

How might we make draw status visible to all parties without exposing internal lender operations or creating complexity?

→ Addresses the 4.2× support call problem
HMW #3

How might we help lenders review draws faster without automating decisions that carry compliance and liability risk?

→ Addresses 15+ hour manual review cycle
03.2 · Design Vision & Strategy

Strategy, Principles & The Central Design Challenge

With the "How Might We" questions defined, this section established the guiding principles that filtered every design decision that followed.

From Reactive to Proactive

The fundamental shift that defined every design decision.

✕
Reactive (Before)
Step 1
Contractor submits draw (guessing requirements)
Step 2
System accepts submission unchecked
Step 3
Lender manually reviews for 2–3 weeks
Step 4
Lender rejects — missing or incorrect documents
Steps 5–6
Contractor resubmits (frustrated, delayed). Cycle repeats.
Result: 47% first-time failure rate, 35-day average cycle
✓
Proactive (After)
Step 1
System guides contractor through requirements upfront
Step 2
Real-time validation catches errors before submission
Step 3
Automated pre-checks validate against budget and contract
Step 4
Lender receives pre-validated request with all docs complete
Steps 5–6
Streamlined approval. Payment initiated automatically.
Result: 92% first-time success rate, 6-day average cycle
The Central Design Challenge

Build one platform that works for two very different ways of thinking, without making either group work the other's way. Contractors think linearly. Lenders think in queues and exceptions.

Contractors → Task-Oriented, Linear

  • My Projects
  • Submit Draw
  • Track Status

Lenders → Queue-Oriented, Exception-Based

  • Pending Reviews
  • Flagged Items
  • Approve · Request Info · Reject

Five Design Principles — The Non-Negotiables

Every decision was filtered through these principles, derived directly from research and agreed with the PM before design began.

1
Simplicity for the Novice, Power for the Expert
→Default to guided workflow →Provide quick mode for experienced users →Never sacrifice novice usability for expert efficiency
2
Show Progress, Always
→Users should never wonder "where am I?" or "how much left?" →Progress indicators, status labels, timeline views
3
Errors Are Design Failures
→If users make mistakes, we failed to guide them →Prevent errors through design — don't just report them
4
Mobile Is Not An Afterthought
→85% of contractors logged in on iPhone — designed for thumbs and spotty connectivity →Every interaction passes the one-thumb test
5
Compliance Without Complexity
→Lenders need audit trails and documentation →Abstract this complexity entirely from contractors

The "Guide Rail" Approach

Inspired by bowling alley bumpers — rather than letting users fail and correcting them later, guide them toward success from the start.

1 · Show, Don't Hide
  • →Display all requirements upfront before contractor starts
  • →Use plain language, not legal jargon
  • →Show examples of acceptable documents
2 · Validate Early, Validate Often
  • →Check document file types on upload — before submission
  • →Validate budget amounts against contract in real-time
  • →Flag missing required fields with specific guidance
3 · Automate the Obvious
  • →Pre-fill data from loan contract and previous draws
  • →Auto-calculate remaining budget per line item
  • →Suggest document types based on draw stage
4 · Checkpoints, Not Gatekeeping
  • →Break complex form into logical steps
  • →Allow save-and-resume — contractors work in bursts
  • →Progress indicator so users always know where they stand
03.3 · Ideation & Concept Exploration

Three Directions Before One Solution

Before picking a direction, I explored three very different approaches. Each answered the HMW questions in its own way, with its own tradeoffs.

Direction A — Explored & Discarded
Wizard-Based Linear Flow

A fully guided step-by-step wizard that locked users into a strict sequence: Project → Amount → Documents → Review → Submit.

What worked
  • +Low cognitive load per step
  • +Clear progress indication
Why discarded
  • −Contractors need to jump back and forth — documents aren't gathered linearly
  • −Testing showed 68% abandoned when locked out of editing earlier steps
Selected
Direction B — Selected
Dashboard-First + Progressive Disclosure

A context-aware dashboard as home base, with progressive disclosure for requirements. Users could freely navigate while the system guided and validated in real time.

Why selected
  • +Matches contractors' non-linear real-world document gathering
  • +78% used dashboard quick actions — highest task success in testing
  • +Scales to lender portal without architectural changes
Tradeoff accepted

Higher initial development complexity — mitigated by phased component rollout.

Direction C — Explored & Deprioritised
Unified Single-Portal (All Users)

One portal with role-based views — contractors and lenders see the same data filtered by permission. Shared activity timeline visible to all parties.

What worked
  • +Eliminates context switching between portals
Why deprioritised
  • −Compliance risk: lenders cannot have contractor-visible workflows in the same session
  • −Research confirmed: lenders actively didn't want contractors seeing their review process
◎
Convergence Decision — Presented to Stakeholders Sprint 3

Direction B was presented to the PM, engineering lead, and two senior lenders in a working session at end of Sprint 3. We walked through all three directions, the research evidence behind each decision, and the tradeoffs. The team aligned on Direction B with one modification: the lender portal's review queue would be built as a separate authenticated session from the contractor portal — a compliance requirement the lenders surfaced in that session.

03.4 · Information Architecture

Two Portals. One Unified Platform.

Separate navigation for each user type, each built around how that group thinks, with the same data underneath.

Flow #1

Contractor: First Draw Submission

  1. Login & Detection
    System detected no active draws, surfaced prominent CTA
  2. Guided Onboarding
    Educational modal explains process before upload begins
  3. Draw Amount Entry
    Pre-populated budget items with real-time validation
  4. Document Upload
    Required docs surfaced with examples and instant checking
  5. Review & Submit
    Summary with pre-submission checklist — green means go
  6. Confirmation
    Success state with tracking number and explicit next steps
Flow #2

Lender: Reviews Flagged Draw

  1. Review Queue Entry
    Priority-sorted queue — flagged items always at top
  2. Detail View
    Side-by-side: review checklist left, document viewer right
  3. Flag Analysis
    Budget overrun highlighted with supporting documentation
  4. Decision Point
    Three clear options: Approve, Request Info, or Reject
  5. Action Modal
    Template-based request for additional info saves time
  6. Auto-Progression
    Next draw in queue loads automatically — no interruption

IA Design Principles

Five research-driven guidelines applied to both navigation structures throughout the design.

◎

Context-Aware Dashboard

On every login the system works out where the user is and shows the actions they need

≡

Integrated Education

Help content embedded in the flow — not hidden in a separate help section

✓

Progressive Validation

Real-time checks at each step, not just a wall of errors at final submission

▶

Auto-Progression

Next task loads automatically after completion — no dead ends, no manual navigation

◉

Side-by-Side Layout

An action panel that stays on screen, so lenders don't have to keep switching context

Contractor and Lender Dashboard Architecture

Design

23

context switches per lender review, before the checklist and the documents sat side by side.

04.1 · Design Rationale

Every Decision Had a Why

Six decisions across both portals. Each one has the research behind it, what we built, and what changed.

Six Data-Driven Decisions Across Both Portals

Lender Portal — Design Rationale
1
Exception-Based Queue Instead of Chronological
Why Legacy system showed all draws chronologically. Lenders wasted time on low-risk draws while high-risk items with compliance flags sat unreviewed lower in the queue.
Solution Sorted by risk: Flagged items → Standard reviews → Recently approved. The draws that need attention always come first.
High-risk draws reviewed 2.5× faster. Lenders reported confidence they weren't missing critical compliance items.
2
Side-by-Side Review Layout Instead of Tabbed
Why When we shadowed lenders, they switched context 23 DM Sans per review on average — toggling between their checklist and the documents they were validating against it.
Solution Fixed review checklist on left, scrollable document viewer on right. Both panels always visible simultaneously — context switching eliminated by design.
Review time reduced by 35% — side-by-side layout identified as the primary contributing factor in post-launch analysis.
3
Automated Pre-Validation Front and Center
Why Lenders needed a go/no-go signal before investing time in detailed review. "Is this ready for me?" was the first question — but the legacy system provided no answer upfront.
Solution Pre-validation results sit at the top of every review. All green means a fast approval. A red flag points the lender straight at the one issue to check.
43% of draws now fast-track approved under 2 hours — vs. only 8% in the legacy system.
Contractor Dashboard — Design Rationale
1
Dashboard as Default Landing
Why Research showed contractors log in asking "What needs my attention?" — not "Let me browse projects." The old system made them click around before they could do anything useful.
Solution Dashboard surfaces urgent items immediately with quick actions for the most common tasks. If a draw needs attention, it's the first thing visible — zero navigation required.
78% of contractors use dashboard quick actions instead of navigating — confirmed in prototype testing with actual users.
2
Submit Draw as Top-Level Navigation
Why Legacy system required: Projects → Select Project → Draws → New Draw = 4 clicks + 3 page loads. The most critical action in the product was buried 4 levels deep.
Solution Made it a primary nav item. If user has one active project, skip project selection entirely — go straight to draw amount entry with smart pre-population.
Time to start draw submission: 18s → 4s — measured in prototype testing.
3
Progressive Disclosure for Document Requirements
Why Contractors were overwhelmed seeing 15+ document slots upfront — even though only 6–8 were actually required per draw type. The full list caused abandonment before it started.
Solution Show required docs prominently. Collapse optional documents behind an accordion. 45 possible requirements filtered to 6–8 relevant items per draw context.
Form abandonment: 31% → 12% after progressive disclosure in usability testing.
Real Constraints & Tradeoffs

Not everything shipped as designed. Three honest constraints that shaped what launched:

Scope Cut

Dynamic checklist shipped in two phases

Engineering estimated the fully contextual filtering at 3 sprints. We shipped a 2-tier version (required vs optional) in Sprint 6, then iterated to full loan-type filtering in Sprint 11.

Stakeholder Pushback

Property owner portal deprioritized

Research showed highest owner anxiety, but leadership prioritized contractor and lender ROI for Phase 1. I documented the case for a read-only owner dashboard — it shipped in Phase 2.

Technical Constraint

Real-time status required backend work first

The 4-stage status tracker design was ready in Sprint 4, but required a new event-driven notification system. Design had to hold 3 weeks while engineering built the infrastructure.

04.2 · Design Problem Deep Dives

The Two Hardest Problems

Two challenges required iterative design exploration before arriving at the right approach.

Design Problem #1 — Requirements Clarity
47%
first-time failure rate

Contractors guessing which documents to submit. All 45 possible requirements shown upfront — abandonment before the first field.

How It Works
1
System Analyzes Loan Context
Loan type · Draw # · Project stage
2
Dynamic Checklist Generated
6–8 relevant items shown, not 45
3
Real-Time Upload & Validation
File type · quality · completeness
4
Successful Submission
92% first-time approval confidence
Before
After
1st attempt
53%
89%
Time
22m
11m
Confidence
3.2/5
4.6/5
15 contractors · 60 test submissions
Options Explored
A · Show All Requirements
Complete transparency — but 20+ items caused abandonment before the first field
B · Minimal Requirements
Simple to complete — still led to rejections because required docs were missing
C · Dynamic Contextual Checklist
SELECTED
Only relevant requirements shown. Requires backend logic integration.
→ Post-launch: 92% approval · 71% fewer requirements-related tickets
Requirements for this draw: only the relevant documents, why each is needed, sample and template links, camera upload and a recoverable file error
Design Problem #2 — Status Visibility
4.2×
avg support calls per draw

Contractors described submission as "sending requests into a black hole." Binary statuses (Pending / Approved) created anxiety, not confidence.

The Decision

A 4-stage timeline with an estimated completion window and a short message at each stage. Contractors see where things stand without seeing the lender's internal process.

Options Explored
A · Binary Status
Easiest to implement — Pending / Approved. Doesn't reduce anxiety or show progress.
B · Full Internal Stage Detail
Shows everything, but it exposes the lender's internal process. Too complex.
C · 4-Stage User-Friendly Timeline
SELECTED
Clear progress + estimated completion. Requires internal stage mapping.
→ Status-check calls dropped 64%
Draw status: Submitted, Documents verified, In review, Approved, with estimated review date and lender contact
04.3 · Visual Design System

Built to Scale Across the Tech Innovation Group

With 25+ screens, 3 user types and 6 engineers working at once, we had to have a design system. It was the only way to keep things consistent and keep moving fast.

Five Principles Behind Every Component

These rules shaped every component decision — you'll see them reflected in what follows.

Principle

Clarity Over Style

Make it work first. Make it look good second, never the other way around

Principle

Progressive Disclosure

Show what's needed when it's needed — never front-load complexity

Principle

Accessible by Default

WCAG 2.1 AA built into every component from day one, not retrofitted

Principle

Mobile-First

Every component passes the one-thumb test before desktop refinement

Principle

Consistent with Purpose

Reuse patterns on purpose. Any one-off needs a good reason

Three-Layer Architecture

01
Foundation Layer
Typography · Colors · Spacing · Elevation
02
Component Layer
40+ Reusable Components — adopted across 4 products
03
Pattern Layer
Form, Navigation & Feedback Patterns

Component Decisions — Design Rationale

Component 01 — Inline Validation

Real-Time Feedback That Reduces Errors by 35%

Research insight that drove one of the most impactful pattern decisions in the system.

Inline validation in the draw form: line-item amounts checked against the available budget, with a confirmation that all amounts are within budget
✓
Good Error Messages
"Amount must be less than $50,000"
"Please enter a valid email address"
"Password must be at least 8 characters"
✕
Bad Error Messages
"Invalid input"
"Error"
"Please try again"

Validation fires on blur — not keypress — so users aren't interrupted mid-typing. Every error names the exact problem and tells the user how to fix it. That single decision reduced submission errors by 35%.

Research Insight

Users who received immediate validation feedback completed forms 22% faster and had 35% fewer submission errors.

Component 02 — Notifications & Alerts

Non-Intrusive Feedback + Actionable Critical Alerts

Two tiers: passive in-app states for routine events, and action-required alerts for compliance-critical moments.

Two tiers of feedback: a quiet status message saying no action is needed, and an action-required list of missing documents with upload buttons and a count of what is still missing

Routine confirmations stay quiet and easy to dismiss. Compliance problems become alerts that need action, listing exactly what's missing, with clear buttons — never generic "please try again" messages that send contractors back to square one.

✓ Good
"Please upload: Updated inspection report · Signed lien waiver"
"Amount must be less than $50,000"
✕ Bad
"Error" / "Invalid input"
"Please try again"
Component 03 — Progress Indicators

Two Patterns for Two Different Mental Models

Filling in a multi-step form and checking a status are different jobs, so each got its own pattern.

Two fundamentally different jobs demanded two different patterns. Forms needed a numeric progress bar because users wanted to know how much was left. Status needed named stages because contractors needed language they could use when calling their lender.

Design Decision

The 4-stage tracker (Submitted → Documents Verified → In Review → Approved) maps directly to insight #2 — the "black hole status problem." Every stage transition triggers a push notification, eliminating the 4.2× support call pattern.

Two progress patterns: the numbered form stepper, and the four-stage draw status tracker

WCAG 2.1 Level AA — Built In, Not Bolted On

Every component was accessibility-audited before marking it as complete in Figma.

C

Color Contrast

4.5:1 for text, 3:1 for UI components — verified across all states including hover and focus

Verified
K

Keyboard Navigation

Complete tab order logic with visible focus indicators throughout every interaction flow

Implemented
A

Screen Reader

ARIA labels, semantic HTML, role attributes — tested with NVDA and VoiceOver during development

Optimized
T

Touch Targets

44×44px minimum on all interactive elements — critical for the 85% of contractors on mobile

Compliant
M

Reduced Motion

All animations respect prefers-reduced-motion media query — zero exceptions

Responsive
S

Text Scaling

Fully functional at 200% text size — no horizontal scrolling, no broken layouts

Flexible
04.4 · Final Design Deliverables

Final Design Deliverables

Six key screens across the two portals, reconstructed with sample data. The actual shipped UI is protected under client NDA.

Fig. 04.4a

One draw request, two portals

The contractor and the lender work in different products, on the same record.

Lender portalReviews the same draw
Lender review queue: needs my review, pre-validated draws, recently approved
Contractor portalSubmits the draw
Contractor creates the draw request: line-item amounts, automatic total and budget check
CHE-002 · Draw #04$120,000Same record in both portals
  1. Submitted
  2. Documents verified
  3. In review
  4. Approved
Draw amount and budget validation
01Draw amount and budget validation Line-item amounts, an automatic total and available-budget feedback.

Testing & delivery

52

participants across card sorting, tree testing and prototype testing, before a line shipped.

15 card sorting25 tree testing12 prototype
Then three prototype rounds: 58% → 72% → 89% direct task success
05.1 · Usability Testing & Validation

Three Methods. 52 Participants.
Zero Guesswork.

We tested every structure and interaction decision before it was built. Research confirmed the structure, and prototype testing showed the gaps we still had to close.

Method 01

Card Sorting

15 participants · Unmoderated

Approach

Gave users 40 cards representing features and content. Asked them to group and label categories without any guidance or prompting from the facilitator.

What We Found

80% agreement on the top-level categories — Dashboard, Projects, Submit Draw. Where people disagreed on sub-categories showed us exactly what to simplify.

Key Outcome

80% Category agreement — confirmed 5-item nav structure
Method 02

Tree Testing

25 participants · Text-only IA

Approach

Text-only site structure. Users asked to find specific information — e.g., "Where would you check the status of a draw you submitted last week?"

What We Found

Users struggled to distinguish "Budget" from "Draws" in project details — a critical navigation issue that led to merging budget info directly into draw detail views.

Task Success Rate

87% vs. 80% target threshold — passed

Avg. Time to Find Information

8.3s After nav restructure — down from 14.2s baseline
Method 03

Prototype Testing

12 participants · Think-aloud protocol

Approach

Interactive prototype with think-aloud. 5 realistic tasks per user type — contractor and loan officer flows tested separately with actual personas.

What We Found

Contractors loved dashboard quick actions but didn't discover them on their own — confirming the need for a first-login onboarding tooltip sequence.

Direct Success Rate

72% No backtracking — led to onboarding addition

What Changed As a Result

Added a first-time user guide surfacing both dashboard quick actions and the Submit Draw button on first login — the two features most critical to task success that went undiscovered without guidance.

Before Testing Findings
  • 7 top-level navigation items — too many
  • Separate Budget and Draws sections — confusing
  • Dashboard quick actions hidden below fold
  • No first-time user guidance or onboarding
  • "Submit Draw" buried 4 levels deep
After Testing Changes
  • 5 streamlined navigation items — confirmed by card sort
  • Budget info merged directly into draw detail views
  • Quick actions prominently positioned above fold
  • First-login tooltip sequence added to onboard users
  • "Submit Draw" promoted to primary navigation

The testing insight that changed the most: Contractors consistently rated the dashboard as "really useful" in post-task surveys — but only 28% found the quick actions on their first attempt without prompting. Once people found them, they were clear. The problem was finding them. This single insight drove the addition of a contextual tooltip overlay on first login.

★
Key Validation Moment
"This is like having a loan officer sitting next to me, telling me exactly what to do. I'd actually use this."

— Contractor with 30 years of experience, Sprint 6 prototype test

That confirmed the vision was resonant. We were solving the right problem.

The Iteration Loop — Test, Change, Retest

Testing wasn't one pass. We went through three rounds of changing the prototype and testing again, and results improved every round.

Round 1 — Sprint 4
Initial prototype, 8 participants
Direct task success: 58%. Key finding: contractors couldn't find Submit Draw, and abandoned the document upload when all 45 requirements were shown simultaneously.
Issues: nav depth + requirement overload
Round 2 — Sprint 6
After nav restructure + progressive disclosure
Direct task success: 72%. Submit Draw now top-level — task time 18s → 4s. New finding: dashboard quick actions not discovered without prompting (only 28% direct).
Progress made, discoverability gap found
Round 3 — Sprint 8
After first-login tooltip overlay added
Direct task success: 89%. Quick action discovery went from 28% → 91%. Time-on-task for first draw submission: 4m 12s average — within the 5-minute target set at project start.
All targets met — cleared for development
Fig. 05.1aThree rounds, two numbers
Direct task successQuick action discovery
0255075100Round 1 · Sprint 4Round 2 · Sprint 6Round 3 · Sprint 858%72%89%28%91%
Round 2 fixed navigation but exposed the discoverability gap; the first-login tooltip in Round 3 closed it.
05.2 · Launch Strategy & Post-Launch Monitoring

How We Shipped and What We Watched

Shipped and launched are two different things. Here's how the rollout was structured, what was instrumented, and what we iterated after the first 30 days.

Phased Rollout Strategy

Instead of switching everyone over at once, we launched in three phases to lower the risk and learn before going wider.

Phase 1 — Week 1–2

Pilot: 50 Contractors

Selected mix of high-volume and first-time users. Legacy system remained active as fallback. Daily monitoring of submission success rates and support ticket volume.

Signal: 88% first-time success in week 1
Phase 2 — Week 3–6

Expand: All Contractors + Lenders

Full lender portal rolled out alongside contractor expansion. Feature flags used to control rollout by loan portfolio. Weekly design QA with engineering to catch anything built differently from the design.

Signal: Support tickets dropped 61% vs. baseline
Phase 3 — Week 7+

Full Cutover + Legacy Sunset

Legacy system decommissioned. Full 90-day measurement window began. Continued bi-weekly design reviews to address edge cases surfaced from real usage at scale.

Signal: All baseline metrics exceeded

What Was Instrumented

  • Draw submission start → completion funnel (Mixpanel)
  • Step-level abandonment rates across the 4-step form
  • Document upload success/failure by document type
  • Lender queue interaction patterns (time per draw, skip rate)
  • Status tracker open rate — proxy for anxiety reduction
  • Support ticket categories and volume (Zendesk integration)
  • Session recordings for first-time contractor logins (FullStory)

What We Iterated Post-Launch

  • Dashboard tooltip: Returning users kept re-triggering the first-login overlay — limited it to the first two sessions in Sprint 12
  • Document naming: Contractors confused by legal document names — added plain-language aliases in Sprint 12
  • Lender bulk actions: Loan officers wanted to approve multiple low-risk draws at once — added in Sprint 13
  • Mobile upload: Camera upload failing on older iOS — patched Sprint 11, resolved in Sprint 12
  • Status ETA: Contractors wanted estimated approval date — added to Phase 2 backlog based on 22% support ticket mentions
◎
Design QA Process

Every sprint release included a design QA pass — comparing live implementation against Figma specs across 5 breakpoints. I maintained a running Jira board of implementation discrepancies. 23 issues were filed across the 14-sprint delivery, of which 19 were resolved before the Phase 3 cutover. The 4 remaining were deprioritised with PM sign-off and carried into the Phase 2 backlog.

Impact & reflection

35→6days

From 35 days to 6, for every draw.

Before35 days
SubmitRejected · 47% of drawsGuess and resubmitCall for status · 4–5×Review across 5 systems · 15+ hrsPaid
After6 days
Paid
Checklist of 6–8 documentsValidated on upload4-stage statusSide-by-side review · ~5 hrs
29 days back in the contractor’s cash flow, every draw
One draw request, before and after. Each bar’s total length is to scale; the stages inside it are in order, not measured. The loop, the status calls and the five-system review are what the redesign removed.
06.1 · Outcomes & Impact

Measured Against What We Set Out to Achieve

Four success metrics agreed with PM and leadership before design work began. Here is how each performed — baseline, target, and actual — across the 90-day window after full launch.

Metric
Draw cycle time
First-time success
Lender review time
NPS score
Baseline
35 days
53%
15+ hrs
−12
Target
< 10 days
> 80%
< 8 hrs
> 0
Actual
6 days ✓
92% ✓
~5 hrs ✓
+16 ✓
Fig. 06.1aBaseline → actual, against the target zone
Target zoneBaselineActual, 90 days after full launch
Draw cycle time
35 days 6 days 0 days40 days
Target < 10 days
First-time success
53% 92% 0%100%
Target > 80%
Lender review time
15+ hrs ~5 hrs 0 hrs16 hrs
Target < 8 hrs
NPS score
−12 +16 −2020
Target > 0
All four metrics landed inside their target zones.
83%
Faster Payment Cycles
35 DAYS → 6 DAYS
92%
First-Time Success Rate
UP FROM 53%
67%
Faster Lender Review
15+ HRS → ~5 HRS PER DRAW
$12.5M
Capital deployed faster per quarter
+28pts
NPS — from −12 to +16
45%
Contractor retention improvement
06.2 · Reflection

What This Project Taught Me

Four things I took into every engagement after this. Earned, not borrowed.

01
Lesson 01

Research Compounds

The same 40 interviews that surfaced the mystery requirements problem also revealed the black hole status problem and the "assistance not automation" insight. Good research keeps paying off. One round of interviews fed several design decisions at once.

02
Lesson 02

Design Systems Are Products

The 40+ component library wasn't a deliverable — it was a force multiplier. When it was adopted across 4 products, that was the design work compounding beyond its original scope.

03
Lesson 03

Asymmetric Users Need Asymmetric Design

Contractors and lenders needed fundamentally different products sharing the same data. The biggest mistake I could have made was designing for a "middle ground" user who doesn't exist.

04
Lesson 04

Ship Value, Not Completeness

The 2-tier checklist shipped first. The loan-type filtering came in Sprint 11. Both moved the metric. Waiting for the complete solution would have delayed real user value by 5 sprints. Progressive delivery isn't a compromise — it's a strategy.

2-tier checklist → Sprint 11 filtering → 53% to 92% first-time success
← Back to Work