Prototyping and design testing

1) Why do you need it

• Hypothesis → Prototype → Test → Insight → Iteration → Solution.

2) Fidelity levels and when to choose them

Low-fi (sketches/paper): structure and flow; 30-90 min per iteration.
Mid-fi (clickable frame): navigation, texts, states; 0.5-2 days.
High-fi (visual/animation): perception, micro-interactions; 1-5 days.
Code-proto (interactive + real data): performance, availability, complex patterns (KYC, payments); 2-10 days.

Rule: start at the minimum sufficient level to answer the hypothesis question.

3) Prototype formats

Paper/Wireflow: Quick Scripts, Screen Comics.
Clickable (Figma/Framer/ProtoPie): transitions, conditional branches, pseudo-forms.
Wizard-of-Oz: "magic behind the screen," complex logic imitated by the researcher.
Code prototype (React/Flutter): gestures, performance, A11y, API integration.
Content prototype: texts, empty states, errors/successes.

4) Where to test and how

• Moderated: depth, clarifying questions, think-aloud.
• Unmoderated: speed, scale, simple tasks.

• Deleted: wider geography, screen recording/video call.
• Face-to-face: context, eye-tracking/gestures, physical devices.

• Usability tests (task scenarios).
• Cognitive walkthrough.
• Heuristic evaluation (Nielsen 10).
• Tree-testing (navigation).
• 5-second tests (first impression).
• Preference test (screen/style options).
• Desirability testing (emo cards).
• Diary studies (long cycles).

5) Test plan (template)

1. Purpose: what hypothesis we are testing (for example: "the new KYC stream will reduce time-to-verify by 30%").
2. Task scenarios: 4-6 key, with success criterion.
3. Artifact: references to prototypes/versions.
4. Metrics: Task Success, Time-on-Task, Errors, SEQ (1-7), SUS (after the session), UMUX-Lite.
5. Recruit: profile, markets, N (5-8 respondents per cycle).
6. Procedure: entering → warm-up → tasks → interviewing → de-brief.
7. Safety and ethics: consent to record, we do not collect personal data/payments.
8. Logistics: devices/browsers, slot grid, moderator checklist.

6) Set of metrics and norms

Task Success (%): target rate of ≥80% on critical steps.
Time-on-Task: median by scenario (comparison with control).
Error Rate - critical/minor error rate.
SUS: ≥68 - basic norm, ≥80 - excellent.
SEQ (1-7): assessment of the complexity of the task (goal - ≤3).
NPS/CSAT after script: as needed.
Behavioral signals: sticking, returns, back-navigation., Rage-clicks.

7) Recruit and sample

Who to call: real segments (beginners/experienced, regions/languages, devices).
How much: 5-8 per quality cycle; for A/B - statistics (see below).
Screening: we cut off professional testers, we require experience in relevant tasks (for example, online payments).
Motivation: honest, without distortion of behavior (do not pay product bonuses for responsible game tests).

8) Moderator script (summary)

Greeting, goal: "We are testing the design, not you."

Consent to record, confidentiality.
Warm-up: previous experience, analogues.
Tasks: "Imagine what you want..." (minimum guidance).

Neutral prompts: "Tell us what you expect to see next," "What is embarrassing?"

Completion: short survey, thanks.

9) Analysis and prioritization of finds

Problem taxonomy: Navigation, Understanding, Input, Feedback, Waiting/Time, Security, Localization/A11y.
Severity: Blocker/Major/Minor/Cosmetic.
Frequency: How many respondents clashed.
Business impact: conversion, compliance risks (CCM/payments), appeals.
Priority = Severity × Frequency × Impact.

Report (1 page): purpose, selection, key insights (5-7), examples of quotes, video links to moments, solution and owner of the problem.

10) A/B and experimentation

What to test: clearly one factor (CTA text, step order, field format).
Metrics: step/funnel conversion, CTR, time, error rate, cancellations.
Sample size: calculate for the desired effect (for example, uplift 5-10%, power 80%, α = 5%).
Ethics and risks: for payments/conclusions to minimize harm: canary launch, traffic caps, cancelability.
Multivariable tests - only at high volumes.

11) Tools and artifacts

Prototypes: Figma/Framer/ProtoPie; for code - React/Next.
Recording and analytics: screen/camera, heat maps, session replay, events.
Templates: test plan, moderator script, consent form, de-brief checklist, one-page report template, decision log.
Data: fictitious/impersonal, payment stubs and KYC.

12) Availability (A11y) and localization in tests

We check navigation with tabs, focus rings, 'aria-live' for toasts/errors.
AA contrast in light/dark themes.
Test "long languages" (DE/TR), date/currency formats, RTL.
Voiced texts of errors/successes, alternatives to icons.

13) Specificity of key flow (iGaming)

Registration/CUS: clarity of requirements, photo tips, ETA ("up to 2 minutes"), document errors.
Payment/deposit/withdrawal: limits, fees, deadlines, same-method rule; rejection texts; confirmation step.
Bonuses/missions: transparent conditions, progress, statuses; avoiding dark patterns.
Responsible play: neutral tone, easy setting of limits, access to help.

14) Handoff and design-QA

Handoff package: prototype + specifications (states, animations, tokens), microcopy texts, i18n keys, A11y requirements.
Definition of Done (UX): loading/empty/error/success, focus/keyboard, localization are covered.
Design-QA: comparison with layout, interactive, contrast/focus, texts, layout on devices/browsers.

15) Checklists

Before the test

• Hypothesis and research question formulated.
• Fidelity level corresponds to the task.
• Scenarios and success criteria are ready.
• Recruit and slots confirmed, consents prepared.
• Prototype opens on target devices/browsers.
• Blinds for sensitive activities (payment/ACC).

After the test

• Notes and timecodes collected.
• Problems are marked by severity and frequency.
• Solutions defined, owners assigned.
• Updated decision log and roadmap.
• Short report (1 page) and demo video (≤5 min) prepared.

16) Anti-patterns

Test "beauty" instead of scripts and tasks.
Mix multiple hypotheses in one test.
Give leading questions/tips before the user tries.
Show only "happy path" - no errors/empty states.

Ignore A11u/localization "before production."

Draw conclusions on 1-2 respondents without confirmation.
Post A/B without calculating sample power.

17) Quick templates

• "If we take the choice of payment method to the first step, then the deposit conversion will increase by 7%, because we will reduce unexpected restrictions later."

• "Please top up with 500 UAH using your credit card."

• "User completes replenishment without assistance and in ≤90 seconds."

• "How difficult was it to complete the task?" (1 - very easy... 7 - very difficult)

• Blocker: stuck and unable to complete → Major: seriously interferes with → Minor: slows down → Cosmetic: does not interfere.

18) Final cheat sheet

Start with low-fi, raise Fidelity only if necessary.
Test scripts, not screens.
Мерь Task Success/Time/Errors + SEQ/SUS.
Prioritize by severity × frequency × impact.
Maintain A11y and localization in each cycle.
Document the decisions and pass on a comprehensive handoff.