Interview Drill Reference
General. Bullets only. Don't recite — thread these as you speak.
When they follow up after a STAR answer
- "What specifically did YOU do?" → your role got buried; "we" swallowed the "I"
- "What was the outcome?" → you didn't close the STAR loop
- "What would you do differently?" → they want growth mindset, not a defense
- "How did you measure success?" → missing data or concrete signal
- "Who else was involved?" → they want collaboration context, not just what got built
Story mapping
Leadership / Influence without authority
Primary ArifZefen — native Android push
- Client had a working iOS app; wanted a direct Android port
- East African users never touched iOS — foreign patterns; even their iOS app broke HIG
- Made the case; argument didn't land → mocked up native screens in Material Design
- #1 Play Store in East Africa, 1M+ users
Fallback SafeTraces — design system before any screens existed
- 3 platforms to ship; instinct was to start designing surfaces immediately
- Argued for shared component library first — no screens yet, no proof it would pay off
- Bet: upfront investment would compound across platforms 2 and 3
- Paid off: components moved freely; later platforms shipped faster
Fallback Souza Group — recommended against the custom build before signing
- Client: per-file permissions for privileged deposition transcripts
- Pre-contract: walked them through Google Drive — model largely solved off-the-shelf
- Declined: data had to stay on-premise (legal requirement, not technophobia)
- Signed with clear requirements; every alternative already discussed; still running a decade later
If all used Pandora — shipped in production instead of presenting designs
- Outside consultant, no authority; established internal workflows
- Skipped wireframes; drove features directly into production
- Made it self-evident by doing, not presenting
Conflict / Difficult colleague
Primary ServiceChannel — PM inserting into workflow
- PM writing tickets directly to devs; demanding Trello → Jira migration
- Reporting phantom no-progress to execs — Trello not surfacing in Jira, board looked stalled
- Process problem, not a people problem — visibility gap, false signals upstream
- Built Trello→Jira integration (pre-Atlassian, required digging); devs stayed in Trello
- Absorbed the PM personally: daily standups directly with her
- Delivered to spec
What looked like a people problem was a process problem. Fix was systemic + relational.
Lightweight pushback Rotten City — built generative color system against business partner's advice
- Partner said don't spend time on it — reasonable scope call for a two-person project
- Built it anyway: custom SCSS function generating palette-matched colors at runtime, ~a day of work
- Rationale: audience is artists; visual credibility is the product's first job; this earned it
- Client loved it immediately; became a defining visual signature
Use when the question is about craft judgment or trusting your read on the audience, not a high-stakes interpersonal conflict.
Fallback TerraTracker — IP dispute with founder
- Founder claimed all developed work as his IP mid-project
- Offered to sell; he had no funding to buy it
- Rebuilt on custom rendering engine — new codebase clear of any claim
- Knew enough to diagnose the right fix: Preact's diffing model was the bottleneck at the sensor's data frequency, not the data pipeline
- Rebuild cleared the performance ceiling; better product came out of the conflict
Prioritization / Tradeoffs
Primary SafeTraces — tiered investment across 3 platforms
- 3 platforms: mobile field, web client, internal admin — all had to ship
- Mobile + web client: client-facing, revenue-relevant → full design treatment
- Admin: deliberate minimal pass — internal, no polish needed to move the needle
- Not cutting corners: deliberate call about where polish produces returns by audience
- Shared system: less-polished surfaces still coherent, not visually broken
Frame as progressive architecture, not tiered polish.
Fallback Capture Collective — web-only constraint as deliberate scope decision
- 4 personas had to launch together — couldn't soft-launch one without the others
- Chose web-only: no native apps, no fragmentation, no parallel design system tracks
- Modularity as core principle: components moved across all 4 surfaces without rework
- 18-hour days for a month; web-only made that pace survivable
Fallback TerraTracker — front-loaded core MVP before partnership deteriorated
- Partnership strained early; IP arrangement ambiguous
- Prioritized core map + directional display first — minimum viable interface
- When it went sideways legally, had something to rebuild from
Fallback StartupOS — scoping an ambitious wish list on a two-person team
- Client wanted a wide variety of module types; we had two people and a live platform
- Continuously cut and deferred — core module types first, shipped well
- Rather than just saying no: iframe injection let us support embedded content; repurposed existing components to get close to intent without full build cost
- Shipped a working product; client's broader business went dark — but that's not the same as a product failure
Scoping is a design skill. Frame as: most valuable output was what we decided not to build.
Ambiguity / Problem solving
Primary SafeTraces — blank slate, 3 personas, hardware that didn't fully exist yet
- Hardware existed, zero software — no reference point, no prior version
- 3 personas: field ops (mobile, zero training), corporate clients (web), admins
- Started with field ops — highest stakes, lowest tolerance; empty state was first session
- Shared data model → component library serving all 3 without duplication
- Hardware API delayed; designed against spec, tested later — scope flexed without losing coherence
Fallback TerraTracker — learned nuclear physics to design the data model
- Real-time radiation detection UI, zero domain background
- 128 channels of photon energy data per detector; peaks → isotope ID
- Went deep on physics first — count discrepancy between detectors = source direction
- Designed from physics, not a feature list; no expert validation needed after
Call it "NukeDash" (working title) — don't say TerraTracker or SFPD.
Fallback Fundue — no-account constraint required rethinking session identity
- No account, no phone, no email, no data — ruled out all conventional identity approaches
- Reframed: session, not account — who you're with now, not who you might find
- Features followed: drawing canvas, dual camera (pre-BeReal), in-call game — shared space, not social graph
Vague brief StartupOS — ambitious, undefined scope; built into a live platform
- Client: "decentralized incubator" — ambitious vision, no clear monetization, long wish list
- Two-person team; had to sequence everything around keeping a live platform stable
- Most important design output was scoping decisions — what to build, in what order, what to cut
- Shipped; real users completed modules; client's business eventually went dark
Research to find solution StartupOS — branching logic UI with no obvious pattern
- Conditional branching in a form builder is a directed graph — paths loop, chains repeat, standard tree views don't model it
- First instinct was canvas/flowchart — honest about not having the implementation skills for it
- Researched beyond form builders: Houdini (procedural video software) has the same node-connection problem
- Landed on scoped tree view: only shows branching local to the current question; constrains complexity without hiding it
Pivot mid-project eQuest — freelancer failure as unplanned pivot
- Delegated to unvetted freelancer before HK; corrupted files, late delivery
- Recovery plan: what to salvage, what to rebuild, how to close the gap remotely
Didn't ship CustomerView — acquisition killed the buildout, not the work
- Designed a complete QA analytics platform; designs validated, base components built
- Client got acquired mid-project — international company wanted the data infrastructure, not the UI layer
- Tried to make the case: produced the architecture diagram showing how the two platforms could integrate
- Acquirer's call to make; quality of the work and fate of the project are separate things
Failure / Setback
Primary eQuest — delegated to an unvetted freelancer
- Month in HK committed; eQuest started a few weeks before
- Freelancer for interstitial views — portfolio looked fine; never saw him work
- Slow, inconsistent, corrupted Illustrator files over Dropbox — damage compounded while away
- Worked nights at HK coworking; couldn't close the gap
- Delivered late and subpar; client cancelled dev, refused to pay; settled at steep discount
- Lesson: never delegate to someone you haven't seen work — portfolio ≠ witnessed performance
Fallback DashPerks — product worked, distribution didn't
- White-label hotspot marketplace for college students who lost home internet
- Launched with one college; low conversion — school didn't promote it
- Model mismatch: students who needed it most couldn't afford it; needed subsidy, not direct purchase
- Shelved: neither distribution nor model fixable without a different go-to-market
- Lesson: distribution in the contract; validate payment model before building checkout
Do not use ArifZefen or SafeTraces here — both were delivery successes
Data / Metrics
Primary ArifZefen — OS version adoption data shifted the design baseline
- US default: 1–2 versions back. East Africa: looked up actual OS adoption data first
- Found 3–5 versions behind — required going to look, not an assumption
- Shifted the version floor; graceful degradation became structural, not an edge case
- Current Material Design components couldn't be assumed to exist on their actual devices
Fallback POC Medical — same OS version research on iOS + Android for rural India
- iOS + Android, rural India — device landscape far from Western defaults
- Same step: looked up actual OS adoption for the geography
- Same outcome: version floor shifted; graceful degradation structural on both platforms
- Stakes higher: a medical app breaking in the field ≠ a music app
Kill decision TerraTracker — conference market research over partner's projections
- Market size came only from the founder — never independently verified
- Attended homeland security conference in Miami to talk to actual buyers
- Founder had overstated the market — technology real, market size wasn't
- Combined with IP threat: shelved based on direct evidence, not partner's projections
No instrumented UX metrics from agency work — own it if probed. The data available was market signals + client feedback. Moving in-house is partly to get the feedback loop that consulting structurally can't give you.
User focus / Advocacy
Primary ArifZefen — designed for who actually has the device
- Client: East Africa iOS = Android; just port it
- Reality: never touched iOS — navigation, gestures, all foreign
- Android was their first and only platform — iOS patterns add friction before a song plays
- Pushed native; mocked up key screens when argument alone wasn't landing
- #1 Play Store in East Africa, 1M+ users
Fallback SafeTraces — field operator UX under real conditions
- Industrial sites: high-stress, zero training, first session starts empty
- Corporate clients assumed operators wanted dashboard density — they didn't
- Designed around operator's real task: step-by-step job progress
- Empty state was first-class — "no data" had to tell them what to do next
If both used POC Medical — empathy without access to the user
- Rural India clinicians: outdoors, time pressure, second language, no connectivity
- Zero user access — designed for the hardest case, not the product demo
- Offline-first; high-contrast; risk score as a dial (gestalt before number); localized from day one
If all used Capture Collective — base user at a physical entrance
- Employees and guests verifying docs at a physical entrance, mild time pressure
- Single action per step; QR code = entire home screen — impossible to miss
- Thousands of users across enterprise clients; no reported queue or failures
Mentorship / Growing someone
Primary Robbi — CS dev developed genuine design sensibility over 3 years
- Hired out of college as pure backend — no design background or interest
- Rote delegation → shoddier output; rigid scheduling got in his way
- Schedule flexibility → output improved immediately
- Switched from task lists to explaining the why behind decisions
- Started asking questions, then better ones; pushed back when implementation hurt UX
- 3 years: genuine design sensibility; in design meetings other devs aren't
Tell it from Robbi's perspective — what you noticed, what you changed, what happened for him. Don't frame as a contrast with another management style.
Only one story here. If asked a follow-up mentorship question, pivot to how you structure critique or give feedback.
Knowing when to kill a project
Primary TerraTracker — market validation revealed overstated addressable market
- IP dispute; had already rebuilt on custom engine to secure ownership
- Market size: founder's claim only, never independently validated
- Homeland security conference in Miami to talk to actual buyers
- Founder had overstated — technology real, market wasn't
- Shelved: legal threat + market too small = no viable path
Fallback DashPerks — traction wasn't coming; distribution couldn't be designed around
- Product complete; failure was distribution
- School didn't promote it; couldn't sign more partners
- Both problems beyond design to solve; shelved
Constraints as a creative forcing function
Primary TerraTracker — IP dispute forced a rebuild that produced a better product
- Founder claiming IP → strategic rebuild on custom rendering engine
- New codebase: Nyentek's IP regardless of prior claims
- Preact had hit a performance ceiling; forced rebuild cleared it
- Best version of the UI came out of the conflict
Fallback Fundue — no-account constraint shaped the entire product personality
- No phone, no email, no data — ruled out all conventional identity approaches
- Reframed: presence, not identity — who you're with now
- Features followed: drawing canvas, dual camera (pre-BeReal), in-call game — shared space, not social graph
- Without the constraint, it would have been another generic video app
Design systems
Primary ArifZefen — native mobile design system to manage client iteration
- Multiple iteration rounds with no framework — structural drift accumulating
- Built native Android system from scratch: components, type, interactions standardized
- Kept client aligned; gave eng team a predictable framework
Primary SafeTraces — multi-platform shared system across 3 apps from day one
- 3 platforms to ship; built shared system before touching any surface
- Every component served multiple contexts — bet: upfront investment pays off on platforms 2 and 3
- It did: later platforms faster because the system already existed
Fallback DashPerks — white-label system with self-serve client customization
- Multi-tenant from day one; each college: branded without a new build
- Self-serve customization layer — schools branded themselves without involving us
0-to-1 / Building from scratch
Multiple options — pick by domain proximity to the interviewer:
- ArifZefen — consumer mobile, music, international market
- SafeTraces — B2B enterprise, hardware-software, multi-persona
- Capture Collective — B2B SaaS, compliance, multi-tenant, AI/OCR
- Fundue — consumer iOS, privacy-first, delight
- DashPerks — internal product, white-label, marketplace
- Rotten City — civic-tech, community platform, B2C, designing-in-code
- StartupOS — creator tools, form builder, edtech, two-sided
Portfolio / Craft walkthrough
Primary Tracking California — data exploration platform
- CA's shape creates corner negative space — used for UI controls
- Split-screen opt-in, not default — reference impls overwhelmed casual users
- Empty state deliberate — other impls preload; this starts blank for exploratory intent
- Feature-level degradation: features removed at small screens, not squished
- Designed-in-code in Preact — WCAG/ARIA in component architecture, not retrofitted
Fallback ArifZefen + SafeTraces — for design systems angle
Design decision you're proud of Rotten City — generative SCSS color system
- Community platform for an arts district — audience is visual professionals
- Built a generative SCSS function: HSL values at runtime within ranges tuned to match the client's logo palette
- Each profile gets a distinct generated color; fills empty image slots for artists with fewer than 4 uploads
- Business partner said skip it; built it anyway in ~a day; client loved it
Hardest design problem StartupOS — directed-graph branching UI
- Conditional branching in a form builder: multiple paths, loops, chains — a directed graph, not a tree
- Canvas/flowchart approach was out of scope; researched beyond form builders into Houdini
- Landed on scoped tree view: only shows branching local to the current question
- Constrains complexity without hiding it; shipped and was actively used
TerraTracker: interviews OK — do not name TerraTracker or SFPD in public-facing materials. IP situation with former partner is unresolved.
Cross-functional partnership at Staff scope
Primary Tracking California — convinced engineering to skip static prototyping
- Proposed: wireframes → production components; skip static prototypes entirely
- Required eng to trust judgment before seeing finished designs
- Component iteration catches real problems static comps miss; committed to being in the code
- Earned trust: delivered component quality that matched what was promised
Fallback SafeTraces — peer to the technical lead, not just the designer
- Embedded with eng; in library selection and architecture decisions, not just the spec
- Design system informed library choices — design and eng decisions made together
Shaped product strategy (not just design)
Primary Capture Collective — defined the 4-user-level architecture under deadline pressure
- Client had a hard compliance deadline — no time for normal scoping; deferred high-level product decisions to me
- No spec, no defined roles, no prior art — had to figure out what the product needed to be, fast
- Identified 4 roles by working through the actual compliance workflow end to end: guest, monitor, site manager, super admin
- Each level: distinct permissions, data access, primary task — 4 separate problems on one shared data model
- Thousands of users across enterprise clients before Clear consolidated the category
Fallback SafeTraces — 3-platform architecture was a product call, not just a design one
- Client had hardware; no software product, no brief for what it should be
- Decided 3 separate platforms (mobile field, web client, admin) rather than one unified tool
- That split determined the product — each persona needed a fundamentally different interface, not a scaled version
If both used CustomerView — contributed to integration architecture, not just the UI
- Analytics platform for a call center QA startup; got acquired mid-project by a larger infrastructure company
- Had enough technical context to shape the CR-X integration architecture — split of responsibilities between the two platforms was a real decision, not just a diagram
- Architecture diagram was used in active acquisition negotiations
- Buildout was underway (base components in code) when acquiring company decided they only wanted the data — not a design failure, a business one
Story overlap warnings
Never use the same story twice in one interview loop.
ArifZefen
Covers Leadership/Influence, Data, AND User Focus. Use it once — fall to fallbacks for the other two if it comes up early.
TerraTracker
Covers Ambiguity, Constraints as forcing function, Knowing when to kill, and Data (kill-decision frame). High repeat risk — flag if it's already been used in this loop.
POC Medical
Covers User Focus, Data (OS version research), and Low-resource/constrained. Use it once.
Capture Collective
The only AI/OCR story AND a primary multi-persona story AND the primary product strategy story. If used for one, pull a different project for the others.
CustomerView
Supporting example only — not a headline story. Covers: internal tools / technical users, project that didn't ship (ambiguity), shaped product strategy (architecture input). Use as a third fallback when primaries are exhausted, not as a lead.
Rotten City
One story: craft judgment / scope pushback. Covers: lightweight conflict (partner disagreement), prioritization (craft investment call), portfolio craft. Low-stakes — good for when the question is about design judgment, not a dramatic interpersonal conflict.
StartupOS
Two stories: (1) branching logic UI — ambiguity, portfolio, technical depth; (2) scoping — prioritization, ambiguity, de facto PM. Don't use the same story in one loop.
Opener — Tell me about yourself
Read the room. Pick one. All pivot naturally into the career arc.
LEGO — warm rooms, confident energy
"I'm the official LEGO Preservation Architect — the LEGO Group and the National Trust for Historic Preservation gave me that title after I won a competition at 14, and as far as I know it hasn't been rescinded."
Community — habit and belonging angle
Lead with karaoke + jazz jams as evidence of building ritual and belonging.
"I've thought a lot about what makes people show up to the same thing week after week."
Food science — intellectual character
"I cook from food science textbooks — I need to understand the actual mechanics before I trust the output. I design the same way."
Honest self-aware — straightforward rooms
"I ran my own agency for 11 years. Around year eight I realized I was writing more payroll than pixels, so I decided to fix that."
Elevator pitch — formal or cautious rooms
10 years across B2C consumer mobile, enterprise B2B, and civic tech. Designed-in-code from early on. Left agency work to go deep on one product. Here's why this one.
Standard behavioral questions
What have you been doing for the past year?
- Deliberate break after 11 years running the agency — burnout is the honest word
- Lighter freelance (brand identities, landing pages); chose not to jump back until actually ready
- Active recovery: therapy, yoga, music, getting social again — treated it as a real project
- Ready now in a way I wouldn't have been six months ago — that's why now rather than earlier
- Answer clean, move forward. Don't linger.
What does good product work look like to you?
- Looks like nobody designed it — like this is just the only way the thing could exist
- Design is load-bearing: every element doing real work; remove anything and the whole thing feels wrong
- Failure mode: you can feel the designer's hand — performing "being designed," not being useful
- Art vs. design: design is solving problems within constraints — if you want pure expression, make art
- The knife: a well-designed knife could only be that way — balance, handle, taper. Same quality standard, much more complex problem space.
Where do you want to be in 3–5 years?
- Don't optimize that far out — focus on the quality of the work in front of me
- Something that outlasts the project: team better, system still running, decisions I can point to
- After 11 years of switching: see what happens when I stay long enough for decisions to compound
- Want to be the Staff designer who makes the people around them better, not just the one with the most impressive individual work
How do you stay close to users without formal research?
- Agency background = no formal research function, rarely direct end-user access. Don't pretend otherwise.
- Market signals as proxy — OS adoption data before drawing a single screen
- Designed from the hardest case — POC Medical: rural Indian clinician, zero user access
- Post-launch signals: Play Store rankings, user reviews, app behavior as feedback loops
- One of the things that appeals about embedded work: finally having a research function. Haven't had it — and I want it.
- Name the agency context once, move on.
How do you define what a Staff designer should own vs. delegate?
- Own: irreducible judgment calls — framing, the why. Transmitting the brief isn't enough
- Delegate: execution within set constraints, once you've seen them work — portfolio ≠ witnessed performance
- Learned both ways: spread too thin owning everything; lost work trusting unwitnessed style
- SafeTraces model: lower-stakes first, watch, then more rope — calibration, not a leap
Why are you leaving / why now?
- Hit consulting's ceiling — work I love crowded out by running the business
- After 11 years, ready to go deep on one product rather than broad across many clients
- Want to see the downstream impact of decisions — agency work structurally can't give you that
- If pushed on the gap: name the burnout honestly, then move forward. Don't dwell.