Research Methods Taxonomy

Tier 1 — Typical

Near-universal among companies conducting any structured research. Tooling is commoditised; skills are widely taught.

Tier 2 — Common

Used regularly by companies with a dedicated research function or meaningful product maturity. Require specialist facilitation, deliberate participant recruitment, or a methodological framework to learn.

Tier 3 — Rare

Largely absent from startup and scaleup research programmes without specific investment. Require specialist hardware, proprietary platforms, clinical credentials, or panel infrastructure built over years. Sorted from least to most specialist.

Legend

Categories definitions

Interviews — Facilitated conversations (individual or group) designed to surface motivations, mental models, experiences, and attitudes. The researcher is the primary instrument. Includes qualitative depth interviews and structured expert elicitation.

Observation — Methods where the researcher watches participants behave in context, either in situ (ethnographic, field) or via a structured task (mystery shopping). Behaviour is observed rather than self-reported.

Surveys — Self-report data collected at scale via structured questionnaires. Captures attitudes, preferences, and stated behaviour. Includes general-purpose instruments and specialised pricing, satisfaction, and prioritisation methodologies.

Behavioral — Task-based or system-recorded data about what users actually do (clicks, navigation paths, task completion, error rates) rather than what they say they do. Largely digital; tooling is the primary enabler.

Physiological — Biometric and neurological measurement methods that capture involuntary physical responses (eye movement, skin conductance, brain activity, facial expression) as proxies for attention, emotion, and cognitive load. Require specialist hardware or certified platforms.

Modeling — Advanced statistical and quantitative methods that take survey or behavioural data as input and produce structured outputs: segmentations, preference weights, positioning maps, market size estimates. Require statistical expertise and often specialist software.

Secondary — Research using existing data sources rather than primary collection. Includes competitive analysis, search data, syndicated panel data, trend platforms, and patent databases. Speed and cost advantages; limited to what already exists.

Panel — Methods built around longitudinal or community-based participant groups, recruited, maintained, and studied over time. Generates tracking data and deep familiarity with a stable cohort. High infrastructure overhead.

Scoring notes and edge cases

In-depth interviews (IDI): why High (B2B) and not Low across the board

The craft elements of IDI — guide writing, facilitation, note-taking, thematic analysis — are learnable in-house, and the tools required are minimal. On that basis, Low might seem defensible. However, the binding constraint for most startups and early scaleups conducting IDI is not craft. It is participant recruitment.

In B2B contexts, reaching qualified participants involves small and tightly protected populations. Cold outreach response rates are low. Gatekeeping is common. Subjects command high incentive expectations and have unpredictable availability. Agency networks such as GLG, Atheneum, and Respondent provide access but at significant per-interview cost (often £200–£500 per completed session). Without a standing panel of engaged customers, recruiting a sample of 12 qualified B2B participants is a logistical project in itself.

In B2C contexts, the problem is substantially more tractable: Prolific, UserInterviews.com, and similar platforms provide screened consumer participants at reasonable cost, and social media recruitment supplements this well. Respondent also serves B2C audiences alongside its well-known B2B offering. Medium reflects the residual coordination effort required even there.

The principle this illustrates: outsource need should reflect the full execution burden — including recruitment — not only the craft of the method itself.

JTBD: two different methods sharing one name

The term "JTBD" in practice refers to two methodologically distinct approaches that share a theoretical ancestor but differ significantly in execution. This document lists them as separate entries.

JTBD / switch interviews (qualitative)
JTBD as a theoretical frame was conceptualised by Tony Ulwick around 1990 and introduced to Clayton Christensen in 1999; Christensen later popularised the term in The Innovator's Solution (2003). The switch interview method was developed by Bob Moesta (with Greg Engle and Chris Spiek) through applied practice. The method involves a small number of semi-structured retrospective interviews — typically 8–12 — focused on a specific switch: the moment a customer decided to hire (or fire) a product in favour of an alternative. The goal is to surface the causal mechanism behind a purchase or adoption decision — the push and pull forces, the anxieties, the social and emotional dimensions.

This approach requires no large sample, no statistical framework, and no specialist software. It is interpretive and qualitative. It is also often practised without any interviews at all — through desk-based job mapping or synthesis from prior research. Outsource need is Low · Low.

Outcome-Driven Innovation (ODI) surveys
Developed by Tony Ulwick, founder of Strategyn (est. 1991), first applied at Cordis Corporation in 1992 and named ODI around 1999. ODI treats the job executor's desired outcomes as measurable variables. Participants rate each outcome on two dimensions — importance and current satisfaction — and an opportunity algorithm scores underserved areas. This requires a large sample (n ≥ 180, per Strategyn's stated minimum; typically 180–3,000), carefully constructed outcome statements written in a prescribed format, and statistical analysis to produce segment-level opportunity scores.

This is the variant that warrants Medium outsource need and a statistical tooling requirement. It is considerably less common in practice than qualitative JTBD, and more closely resembles a structured quantitative survey programme.

When practitioners say "we do JTBD," they almost always mean the qualitative, interview-based approach — not ODI.

Methodology

Frequency tiering methodology

Methods are assigned to one of three tiers based on how commonly they appear in research programmes within the calibration range, drawing from practitioner surveys, UX and market research curricula, and industry benchmark reports (see references).

• Tier 1 — Typical: Near-universal among companies that conduct any structured research. Low barrier to first use; tooling is commoditised and skills are widely taught. A company with one part-time researcher will likely use all Tier 1 methods within its first year.
• Tier 2 — Common: Regularly used by companies with a dedicated research function or sufficient product maturity to demand more rigorous insight. Require specialist facilitation, deliberate participant recruitment, or a methodological framework that takes meaningful time to learn.
• Tier 3 — Rare: Largely absent from startup and scaleup research programmes without specific investment. Methods at this tier require physical lab infrastructure, clinical credentials, proprietary software with high licence costs, or panel infrastructure built over years.

Frequency ratings are approximations, not empirical measurements. They describe prevalence within the calibration range. Some Tier 3 methods are common practice in industries such as FMCG, pharma, or financial services — the rating reflects digital product company norms, not the broader research industry.

Outsource need methodology

The outsource need score reflects the difficulty of executing a method entirely in-house within the calibration range. It is evaluated across four binding constraint types:

1. Participant access — Can qualified participants be recruited without an agency or specialist panel? Difficulty increases where populations are hard to reach or where volume requirements exceed what organic outreach can supply.
2. Specialist knowhow — Does execution require a trained specialist — ethnographer, statistician, neuro-researcher — whose profile takes years to develop internally?
3. Hardware and software — Does the method require equipment with significant capital cost or platform licences inaccessible at standard SaaS pricing?
4. Normative and benchmark data — Does accurate interpretation require proprietary databases of comparable results that only established research firms possess?

Low: None of the above constraints meaningfully apply. In-house execution is realistic with standard tools and moderate skill.
Medium: One or two constraints apply, or one applies significantly. Outsourcing buys meaningful gains in quality, scale, or speed, but in-house execution remains viable with effort.
High: Three or more constraints apply, or one is categorically inaccessible. In-house execution without specialist support would produce unreliable results.

B2B vs B2C outsource need

B2B and B2C contexts are scored separately because participant dynamics differ substantially.

B2B raises outsource need when: decision-maker populations are small and protected by gatekeepers; NDA requirements are common; subjects command high incentive expectations and have limited availability; and professional panel infrastructure (GLG, Atheneum, Respondent) carries significant per-interview costs.

B2C lowers outsource need when: large participant pools are accessible affordably (Prolific — primarily academic and behavioural research recruitment; UserInterviews; Hotjar Engage); social media recruitment is viable; and sample size targets are achievable without specialist agency support.

The outsource need column always shows values as B2B · B2C, including where both scores are the same.

Calibration detail

All scoring in this document is calibrated to the operational context of startups and scaleups building digital products. The document is product and industry vertical agnostic — methods and scores apply equally across digital product categories.

The calibration range spans a significant spread of internal capability. A pre-seed startup may have no budget, no CRM, and no customer base to draw participants from. A Series B scaleup may have a small research team, an established customer base, and a five-figure per-study budget. Scores reflect the constraints that apply across this range.

Note: This document uses funding stage as a practical proxy for operational maturity. The OECD/Eurostat definition (Oslo Manual, 2005) and Nesta's application of it define scaleups by growth rate — ≥20% annualised growth in employees or turnover over three years, with ≥10 employees at the start — not by funding round or headcount. The funding-stage framing used here is common in practitioner and ecosystem contexts but does not map precisely onto the OECD definition.

Note: Some US-centric sources (Crunchbase, Y Combinator) use "startup" loosely through Series C or beyond as a cultural designation. This document uses the European definition, where stage reflects operational maturity rather than growth ambition.

Comparison with mature enterprise:

Note on internal research history: this refers to proprietary longitudinal data accumulated through repeated research cycles — past NPS waves, prior usability benchmarks, historical concept test results — that give interpretive context for new findings. It is distinct from industry or normative benchmarks, which are available externally regardless of company maturity.

Implication for scores: The baseline for all outsource need scores is a seed-stage startup. Scores were not averaged across the full startup–scaleup range — they reflect what a small, under-resourced team with no existing participant pool faces. Constraints ease progressively as companies move through Series A and beyond.

Data sources and references

Primary method references

• Nielsen Norman Group — method-specific articles cited inline throughout. Full library: nngroup.com
• Interaction Design Foundation — method encyclopaedia: interaction-design.org
• User Interviews — UX Research Field Guide: userinterviews.com/ux-research-field-guide
• iMotions — biometric and physiological method guides: imotions.com/blog
• Sawtooth Software — conjoint and max-diff knowledge base: sawtoothsoftware.com/resources/knowledge-base

Stage definition sources

• ScaleUp Institute. (Annual). Annual ScaleUp Review. scaleupinstitute.org.uk — the primary ongoing annual report on UK scaleup activity, published since 2016. (Note: The Scale-Up Report on UK Economic Growth was a one-time independent report by Sherry Coutu CBE, 2014, commissioned by the UK government.)
• OECD/Eurostat. (2005). Oslo Manual: Guidelines for Collecting and Interpreting Innovation Data (3rd ed.). OECD Publishing.
• Startup Europe / European Commission. Startup and scaleup definitions used in EU innovation policy frameworks: ec.europa.eu/growth/sectors/digital-economy/start-ups

Frequency tiering sources

• ESOMAR. (Annual). Global Market Research Report. esomar.org
• Nielsen Norman Group. (2022). When to Use Which User-Experience Research Methods. nngroup.com
• Dovetail. (2023). State of User Research Report. dovetail.com
• User Interviews. (2023). UX Research in the Wild. userinterviews.com

JTBD methodological sources

• Christensen, C.M., Hall, T., Dillon, K., & Duncan, D.S. (2016). Competing Against Luck: The Story of Innovation and Customer Choice. HarperBusiness.
• Ulwick, A.W. (2005). What Customers Want: Using Outcome-Driven Innovation to Create Breakthrough Products and Services. McGraw-Hill.
• Moesta, B., & Engle, G. (2020). Demand-Side Sales 101: Stop Selling and Help Your Customers Make Progress. Lioncrest Publishing.
• Christensen Institute — JTBD overview: christenseninstitute.org/video/what-is-jobs-to-be-done-theory
• Strategyn — ODI methodology: strategyn.com

General research practice

• Goodman, E., Kuniavsky, M., & Moed, A. (2012). Observing the User Experience: A Practitioner's Guide to User Research. Morgan Kaufmann.
• Portigal, S. (2013). Interviewing Users: How to Uncover Compelling Insights. Rosenfeld Media.
• Krug, S. (2014). Don't Make Me Think, Revisited. New Riders.
• Hall, E. (2013). Just Enough Research. A Book Apart.
• Rohrer, C. (2022). When to Use Which User-Experience Research Methods. Nielsen Norman Group. nngroup.com

Biometric and physiological research

• Duchowski, A.T. (2017). Eye Tracking Methodology: Theory and Practice (3rd ed.). Springer.
• iMotions. (2020). Biometric Research Methods: A Guide. imotions.com

Industry bodies

• ESOMAR — global research industry standards: esomar.org
• Market Research Society (MRS) — UK standards and qualifications: mrs.org.uk
• UXPA (User Experience Professionals Association): uxpa.org

47 methods across 8 categories. Calibrated for startups (pre-seed, seed) and scaleups (Series A, Series B) building digital products. Last updated March 2026.