DRAWING 01 · HERO

FIG. 01 · HERO

CASE NO.: 02 / 05
BASED: DEN · REMOTE
ROLE: DESIGN & BUILD
REV.: v2026.04

01 · BRIEF · CASE 02 OF 05

Synthetic users as a research method.

Nine personas across four QA environments surfaced three blind spots self-review couldn't. All three shipped from a prioritized backlog.

Method: Persona-driven QA
Personas: 9 across 4 environments
Artifact under test: This portfolio · v3 pre-ship
Outcome: 10-item prioritized backlog · 3 shipped changes

↓ OUTCOME · FUNNEL

DRAWING 02 · OUTCOMEFIG. 02 · FUNNEL

DRAWING NO.: 02 / 07
CLAIM: 9×4 → 18 → 10 → 3 → 3
BASELINE: SELF-REVIEW GAP
LAST REV.: 06/26

02 · OUTCOME · THREE FINDINGS, ALL SHIPPED

Three findings self-review missed — all three shipped.

OUTCOME.FUNNEL9×4 → 18 → 10 → 3 → 3

3/ 10actionable findings would not have been caught by self-review — all three became shipped changes.

Run B · companion · parallel

1design-systems persona · full-site audit

53line-level fixes · across 8 sections

↳ folds into Run A's prioritized backlog — parallel track, not a fifth stage

Traceability from synthetic persona → specific design decision is what separates research from a feedback loop.

↓ EVIDENCE · WHERE THE METHOD HOLDS

DRAWING 03 · EVIDENCEFIG. 03 · LADDER

DRAWING NO.: 03 / 07
CLAIM: REPLACE / AUGMENT
BASELINE: SELF-REVIEW
LAST REV.: 06/26

03 · EVIDENCE · WHERE THE METHOD HOLDS

What synthetic research replaces, augments, or fails to reach.

EVIDENCE.LADDERWhere synthetic research replaces, augments, or fails real researchN=9 personas · 4 QA envs

Signal × Method

DEV

Structural

Content

UAT

Audience

PROD

Adversarial

REAL USERSRecruited practitioners
N=12 · 6 wk · $8k

IA clarity

High · ground truth

Copy resonance

Ship · observational

Target-fit signal

Ship · highest confidence

Edge stress

Mid · access-limited

SYNTHETICPersona-driven QA
N=9 · 2 hr · $0

Structural gaps

High · replaces real

Positioning drift

High · replaces real

Audience targeting

Mid · augments real

Adversarial review

Ship · pre-flight stress

Synthetic vs real

Replaces

Augmentsreal still required

Replaces

REAL USERS

Recruited practitioners · N=12 · 6 wk · $8k

DEV

IA clarity

High · ground truth

Copy resonance

Ship · observational

UAT

Target-fit signal

Ship · highest confidence

PROD

Edge stress

Mid · access-limited

SYNTHETIC

Persona-driven QA · N=9 · 2 hr · $0

DEV

Structural gaps

High · replaces real

Positioning drift

High · replaces real

UAT

Audience targeting

Mid · augments real

PROD

Adversarial review

Ship · pre-flight stress

decision types where synthetic replaced real research

decision type where real research is still required

$8K → $0

pre-flight cost · self-review baseline

↓ PIPELINE · PORTFOLIO AS A RELEASE

DRAWING 04 · PIPELINEFIG. 04 · ENVIRONMENTS

DRAWING NO.: 04 / 07
CLAIM: 4 ENV · 4 GATES
BASELINE: PRE-SESSION
LAST REV.: 06/26

04 · PIPELINE · PORTFOLIO AS A RELEASE

Two synthetic runs, parallel methodologies.

The pipeline runs different persona classes against different artifacts. Run A stress-tests the work itself across four environments, each environment matched to a persona class with a defined critical frame. Run B audits the design system against a single domain-expert persona. Both runs feed the same prioritized backlog.

DEV#1

StructuralArchetype usersDiagnostic

Can any user find what they need in 60 seconds?

Pass criteria · IA integrity

QA#2

ContentThought-leader personasDiagnostic

Does the copy communicate leadership? Five practitioner worldviews stress-test method and metrics.

Pass criteria · method + metrics hold

UAT#3

AudienceLinkedIn-modeled personasDiagnostic

Would the specific person I'm targeting respond?

Pass criteria · pre-outreach sharpening

PROD#4

AdversarialAdversarial usersDiagnostic

Does it survive a skeptical reviewer with 300 portfolios this month?

Pass criteria · ship only if it does

SHIPsurvives all
four gates

Run B · companionOne design-systems persona running a full-site audit — runs parallel to the pipeline, not as a fifth lane.↳ folds into the prioritized backlog

Systems Purist

Marc S.

Outcome Enforcer

Indi Y.

AI Pragmatist

Ethan M.

Simplicity Thinker

John M.

Discovery Lead

Teresa H.

Persona card · sample · PC-01 / 05 · QA · content

Marc S. — Systems Purist

Worldview anchor: Service systems thinking. Backstage architecture as the test of whether work is service design or product redesign with a service label.
First question: “Where's the backstage? Who delivers this internally, and what changed in their work to make it real?”
Pass criteria: At least one artifact per service-design engagement shows internal capability change, not just a customer-facing model.
Failure mode: Sees a tier structure with no organizational change behind it and stops trusting the service-design framing.
Finding: FAIL · AutodeskThree-tier model legible from the front, backstage architecture invisible.Recommendation logged: service blueprint excerpt or org-behavior narrative required in Autodesk case study.

Card 01 of a 5-archetype system · QA · content lane

Each persona was built from a published worldview, not an invented archetype. For each: what would this person look for first, and what would make them stop reading?

RUN B · Design system audit (parallel run)

Run B is a different methodology against the same portfolio. One persona, a design-system domain expert, auditing visual grammar, voice consistency, cadence, and the cliché ledger across all pages, line by line. Not a stress test. A specification audit.

Output: structured BEFORE/AFTER recommendations across 8 sections of the site. Findings integrated into the Run A prioritized backlog where they overlap; held separately where they don't.

OPEN · 9 PERSONAS / 4 ENVIRONMENTS / 18 FINDINGS

FIG. 04-A · QA REPORT (Run A full audit trail)

OPEN · LINE-BY-LINE / 53 FIXES / 8 SECTIONS

FIG. 04-B · DESIGN CRITIQUE (Run B full audit trail)

↓ FINDINGS · WHAT SELF-REVIEW MISSED

DRAWING 05 · FINDINGSFIG. 05 · SIGNALS

DRAWING NO.: 05 / 07
CLAIM: 3 SIGNALS
BASELINE: SELF-REVIEW MISS
LAST REV.: 06/26

05 · FINDINGS · WHAT SELF-REVIEW MISSED

What self-review missed.

SERVICE RECORD · 03 FINDINGS · ALL SHIPPED

F·01The tell

PRESENTATION-TO-DEPTH RATIO

Experienced reviewers read the ratio of presentation quality to case-study depth as an AI-generation signal. The AI use itself was never the tell.

The fix wasn't less AI. It was more depth.

BECAME SHIPPED CHANGE · Centaur Practice case study expanded from stub to full Service Records cluster.

F·02Metric class

METRIC CATEGORY MIXING

“$50M YOY AOV” and “15 services shipped” belong to different metric categories — outcome vs output. Their coexistence undermines the more defensible numbers.

A recruiter wouldn't notice. A hiring design leader will.

BECAME SHIPPED CHANGE · “15 services shipped” replaced with “Business plan renewal rate exceeded 75% goal in Q1 post-launch” — a behavior-change metric, not an activity count.

F·03Meta

META-IRONY

The Synthetic Users case study — the very methodology being documented here — had no methodology shown. Meta-irony surfaced only by running the session.

The fix: show the persona card, show the hypothesis, show the finding. Which is what this case study now does.

BECAME SHIPPED CHANGE · This page. Restructured from 3-section stub to six-section documented methodology with two artifact links.

RUN B · SELECTED FIXES

Hub title cliché resolved April 2026 — see Hub-only copy footnote below (inventory aligns caseStudiesPage with Selected work / Five builds subhead).
Career arc voice: third person → first person ("Matt's work spans..." → "I started in creative...")
Cliché ledger: 12 verdicts logged. "Outcomes that compound" KEEP. "Exit velocity" CUT.

See FIG. 04-B for full audit.

↓ LESSON · WHAT I'D DO DIFFERENTLY

DRAWING 06 · LESSONFIG. 06 · LIMITS

DRAWING NO.: 06 / 07
CLAIM: PRE-RESEARCH TOOL
BASELINE: UNDER-INVESTED
LAST REV.: 06/26

06 · LESSON · WHAT I'D DO DIFFERENTLY

What I'd do differently.

Synthetic users cannot replace real ones. I under-invested in the ones I shipped.

They surface hypotheses the designer could already form, not unknown unknowns. The quality of a session is bounded by the quality of the persona model. In this session, adversarial personas were the most productive; archetypes surfaced less.

The method is a pre-research sharpening tool. It doesn't replace observation. It makes observation more targeted when you get there.

↓ HANDOFF · THE LARGER STORY

DRAWING 07 · HANDOFFFIG. 07 · HANDOFF

07 · HANDOFF

A method, proven on my own work. The enterprise design it was built to sharpen — next.

Open next case · Autodesk

BACK TO ALL CASE STUDIES