strategy · 6 min read · April 2026

UX Research Protocol — NUUN's Method | NUUN Digital

Insight

NUUN's UX research protocol — discovery, generative, and evaluative research blended into one operating model with clear cadence and deliverables.

Categorystrategy
UpdatedApril 2026

Last updated:

Quick answer
NUUN's UX research protocol blends discovery, generative, and evaluative research into one operating model. Discovery runs continuously (PMs and designers). Generative research runs in sprints (external researchers + in-house team). Evaluative research runs before every ship gate. Governance ties all three to product and business outcomes through a shared research ops layer with one taxonomy, one repository, and one cadence.

UX RESEARCH PROTOCOL — NUUN'S METHOD

Quick Answer: NUUN's UX research protocol sequences three research modes — discovery (what problem?), generative (what should we build?), and evaluative (does it work?) — inside a continuous operating model. Each mode has a specified sample, method, deliverable, and decision link. The protocol is method-agnostic but governance-strict: every study ships with a published methodology note and links to the product decisions it shaped.

THE THREE MODES

Discovery research. Run before a product bet is made. Open-ended, generative. Methods: exploratory interviews, ethnography, diary studies, journey mapping. Deliverable: problem frame.

Generative research. Run during design. Concept-testing, co-creation, tree testing, card sorting. Deliverable: design decisions.

Evaluative research. Run during and after build. Usability testing, A/B testing, accessibility auditing, post-launch behavioural analysis. Deliverable: validation and iteration queue.

Most teams run evaluative without discovery and ship products that solve the wrong problem efficiently. Discovery is the highest-leverage mode and the most often skipped.

THE PROTOCOL AT A GLANCE

| Mode | Goal | Typical Method | Sample | Cadence | |---|---|---|---|---| | Discovery | Frame the problem | 8–12 IDIs + field visits | Purposive | Pre-project | | Generative | Shape the solution | Concept test, co-creation | 20–60 | During design | | Evaluative | Validate and iterate | Usability, A/B, accessibility | 5–20 per iteration | Continuous |

DISCOVERY — WHAT TO RUN

In-depth interviews (IDIs). 45–60 minutes. Semi-structured. Aim for saturation, typically 8–12 interviews per segment.

Contextual inquiry / field visits. Observing real work in real context. Highest-leverage method for B2B and complex workflows.

Diary studies. Multi-day self-reported logs. Captures over-time behaviour that IDIs miss.

Journey mapping. Synthesis artefact across touchpoints, emotions, and systems. Living document; updated as evidence grows.

Desk research. Existing data, analytics, competitive scan, literature. Always before primary fieldwork, never in place of it.

GENERATIVE — WHAT TO RUN

Concept testing. Mid-fidelity stimuli (storyboards, clickable prototypes) tested with target users. Qualitative first, then quantitative validation.

Card sorting / tree testing. Information architecture validation. Low cost, high leverage pre-design.

Co-creation workshops. Structured generative sessions with users. Most useful for complex B2B workflows and regulated-industry design.

Quantitative preference testing. When two viable directions exist and the organization disagrees.

EVALUATIVE — WHAT TO RUN

Moderated usability testing. 5–8 participants per iteration. The single most under-used research method in enterprise.

Unmoderated usability. Platform-delivered (Maze, UserTesting, Lyssna) for wider reach and faster turn. Complement to moderated, not replacement.

A/B testing. For live products with enough traffic. Statistical discipline required — sample size, test duration, guardrail metrics.

Accessibility auditing. WCAG 2.2 AA is baseline; WCAG 2.2 AAA for regulated or public-sector contexts. Automated plus human.

Analytics-driven evaluation. Session recordings, funnel analysis, heatmaps. Complementary to direct research, not substitute.

SAMPLING DISCIPLINE

Common misconception: more participants is always better. The five-users heuristic (Nielsen) applies to usability testing at a single design iteration; it does not apply to quantitative preference testing or segmentation-sensitive research.

Our defaults:

  • Discovery IDIs: 8–12 per segment, saturation-based
  • Generative concept tests (qualitative): 15–25
  • Generative concept tests (quantitative): 200+ per cell
  • Moderated usability: 5–8 per iteration
  • Unmoderated usability: 20–40 per iteration
  • Accessibility: 5–10 with assistive-tech users per audit

DELIVERABLES — WHAT WE ACTUALLY SHIP

Discovery phase. Journey map, problem frame, research-backed opportunity areas, video reels of customer moments.

Generative phase. Concept validation report, IA recommendation, annotated prototype with research rationale.

Evaluative phase. Usability findings prioritized by severity and frequency; accessibility audit with remediation roadmap; A/B test readouts with decision.

Every deliverable has a published methodology note with sample, method, and limitations. Video and audio reels accompany readouts — raw customer voice beats any paraphrase.

OPERATING MODEL

A working UX research practice runs three cadences in parallel:

  • Continuous — usability testing every sprint or two, rotating feature area.
  • Monthly — discovery and generative research on near-term roadmap bets.
  • Quarterly — strategic research on longer-horizon questions (new segments, new products, market entry).

Headcount ratio of roughly 1 UX researcher per 8–10 designers-plus-PMs is typical for mature practices. Under-staffed practices lean heavily on external research partners (including NUUN) for surge capacity and specialized studies.

GOVERNANCE

Consent and privacy. Every participant informed, consented, and compensated. Video and audio recorded only with explicit consent. Storage and retention per privacy regime.

Accessibility inclusion. Assistive-tech users included in usability samples for any digital product. Meaningful inclusion, not token.

Bias awareness. Recruitment, moderation, analysis, and reporting reviewed for bias. Diverse researcher teams reduce blind spots.

Methodology transparency. Methodology notes published with every study. Raw data (appropriately anonymized) accessible to product teams.

COMMON MISTAKES

Skipping discovery. Teams jump to evaluation because it feels concrete. They then ship well-tested products that solve the wrong problem.

Over-relying on unmoderated. Great for breadth, poor for depth. Balance with moderated.

Treating A/B tests as UX research. A/B tests show what wins; they do not show why. Pair with qualitative.

Recruiting only your existing users. Produces survivorship bias. Include lapsed, competitive, and prospective users for discovery and strategic work.

Letting research live in a silo. Research must be operational — insights visible to product, design, engineering, and marketing.

FAQ

Q: How much UX research is enough?

A: Measured by decisions informed, not studies run. If product roadmap decisions are being made without research evidence, not enough. If research is landing but nobody acts, the issue is operating model, not volume.

Q: Should UX research live under design or under insights?

A: Either works; consistent reporting matters more than org placement. The most common healthy patterns are UX research under a VP of Design or inside a central Insights function that serves design.

Q: What's the minimum viable UX research program?

A: One trained UX researcher, one recruiting budget, one project of each mode (discovery, generative, evaluative) running in parallel. Below that, teams rely on external partners.

Q: How does NUUN run UX research for clients?

A: Three engagement shapes: discovery sprints (4–6 weeks), in-sprint embedded research (ongoing), and surge project work (3–10 weeks). We staff named senior researchers, not anonymous benches.

Q: Do we need generative AI in UX research?

A: AI accelerates synthesis, recruiting screener design, and transcription. It does not replace fieldwork. We deploy LLM synthesis with governance (PII redaction, human review for high-stakes findings).

Q: What about remote vs in-person research?

A: Remote is default for distributed or enterprise users; in-person for field contexts, assistive-tech research, and strategic workshops. Hybrid is the norm.

Q: How long should a usability test session be?

A: 45–60 minutes for moderated; 15–20 for unmoderated. Longer sessions yield diminishing returns and participant fatigue.

Q: What's the cost profile?

A: A single usability-testing round runs $15K–$35K depending on sample and recruitment. Discovery programs $40K–$150K. Strategic programs $80K–$300K. Embedded research is priced on monthly retainer.

RELATED READING

SOURCES & FURTHER READING

About the author

NUUN Digital UX Research

Written by NUUN's UX research practice

Practice-lead review at publication; method and sources cited inline. Refreshed quarterly.

Frequently asked.

What are the three layers of NUUN's UX research protocol?
Discovery (continuous, run by PMs and designers), generative (sprint-based, run by researchers with product team partnership), and evaluative (pre-ship, run by researchers with structured usability methods). Each has distinct cadence and deliverables.
How does research ops tie the layers together?
A shared research repository, a common research taxonomy, standard recruiting and incentive workflows, and a governance cadence that brings product, design, and research together quarterly to prioritize the research pipeline.
What cadence does the protocol follow?
Discovery: continuous, logged weekly. Generative: 3–6 week sprints per research question. Evaluative: every ship gate, minimum 5 usability sessions plus analytics review. Quarterly research-ops governance sets the pipeline.
How many researchers does this protocol require?
At least one research ops lead and 2–4 research generalists per 50 product/design staff. PMs and designers contribute to discovery continuously, which multiplies researcher leverage.
What deliverables does each research layer produce?
Discovery: weekly insight logs. Generative: research reports with jobs-to-be-done and opportunity maps. Evaluative: usability findings, severity ratings, and ship/no-ship recommendations. All feed a single research repository.
Does this protocol integrate with quantitative research?
Yes. Discovery often uses product analytics; generative integrates with MaxDiff or conjoint where feature prioritization is at stake; evaluative pairs with A/B testing and analytics-driven heuristic review. Qual and quant run in sequence, not parallel.

Research That Ships With The Product

If your team is making roadmap decisions without research evidence, we can stand up a research cadence in 60 days.