The Market Research Workflow: Evidence from 105 Current Vacancies

MTF Institute Research Team

doi:10.5281/zenodo.22254955

The Market Research Workflow: Evidence from 105 Current Vacancies

The complete open archive - a visually reviewed PDF, the rights-reviewed 105-row dataset, coding summary, quality record, methods appendix and data dictionary - is preserved at Zenodo DOI 10.5281/zenodo.22254955. The direct public PDF is available here.

MTF Institute Research Report
Research date: 1 September 2026
Author: MTF Institute Research Team
Status: Published MTF Institute Research Report

Abstract

This report examines 105 unique public vacancy results retrieved on 1 September 2026 from 635 saved results and 569 dedupe groups. It admitted direct and bounded adjacent roles only when cleaned evidence showed role context and a substantive output; adjectival analytical and title-restatement analytics did not qualify. Interface or related-card text, promotional execution, unrelated research, adverse status, duplicates and twelve reviewed records were excluded.

Survey or questionnaire work was visible in 51 of 105 accepted snippets (48.6%), and reporting or presentation in 48 (45.7%). Synthesis appeared in 18 (17.1%); data analysis and quantitative analysis in 15 each (14.3%); recommendations in 13 (12.4%); qualitative analysis and stakeholder communication in 12 each (11.4%); fieldwork in eight (7.6%); and interviews or focus groups in seven (6.7%). Quality operations appeared three times, sampling and secondary research twice each, and research brief/design once. These are lower-bound phrase-visibility counts, not full-duty or workforce-prevalence estimates.

The findings support a professional learning sequence centred on decision framing, research design, sampling, recruitment, interviews, surveys, fieldwork quality, qualitative and quantitative analysis, triangulation, insight development and report delivery. The corpus does not support turning that sequence into advertising training. It also does not justify claims about the global size of the pure-research profession. U.S. Bureau of Labor Statistics figures of 952,700 employed people and about 82,000 annual openings refer to the combined occupation “Market research analysts and marketing specialists,” not to dedicated researchers alone. The report therefore treats those figures only as broad occupational context and bases its workflow conclusions on the accepted vacancy corpus.

1. Introduction

Organizations commission market research because a consequential decision contains uncertainty. A team may need to decide whether to enter a market, which customer group to study first, why adoption differs across groups, how people understand an offer, or whether an apparent pattern deserves further investigation. The researcher’s contribution is not merely to collect opinions or create charts. It is to design a credible path from the decision and the uncertainty to evidence that can be examined, challenged and used with proportionate confidence.

That path matters because poor research can look persuasive. A questionnaire can contain leading questions. An interview sample can omit a material context. A percentage can hide a small or biased base. A summary can turn a minority view into a general conclusion. A polished report can conceal contradictory evidence or imply causation that the study did not establish. Professional market research therefore combines method, operations, interpretation and communication. It requires decisions about who should be represented, what may be collected, how evidence should be handled, what analysis fits the design, what remains uncertain and how a decision owner should use the result.

Public occupational data establishes broad relevance but does not by itself define this role. The U.S. Bureau of Labor Statistics groups market research analysts together with marketing specialists under Standard Occupational Classification 13-1161. For the 2025–2035 projection period, the published line reports 952,700 employed people in 2025, projected employment of 1,019,000 in 2035, 7% growth and approximately 82,000 openings per year on average. The category is intentionally broad. It includes work beyond dedicated research practice, so its employment and openings figures cannot be restated as the number of market researchers or live market-research vacancies. In this report, those values provide combined-occupation context only.

The purpose of the present study is narrower: to identify recurring elements of research-heavy work visible in a reproducible public-vacancy sample. The study begins from the distinction between research and marketing execution. A research role may support product, brand, commercial, customer-experience or strategy decisions, but its professional output is evidence and interpretation. An advertising role plans or executes messages, media, campaigns or acquisition activity. Both may use information about customers, yet they own different processes and produce different work. The sampling and exclusion rules were designed to retain the former and reject the latter.

The evidence is also deliberately bounded. The source material consists of saved public-search snippets rather than full employer vacancy bodies. This choice protects source rights and makes the retained record compact, but it constrains interpretation. A snippet may expose one duty while omitting several others. It may omit the employer, exact location, posting age or complete method requirements. Consequently, the analysis asks what was visible, not what every accepted role certainly contained. Frequency counts should be read as minimum visible signals within this corpus, not as estimates of workforce prevalence.

The report has two audiences: the research and education community interested in professional preparation, and curriculum designers deciding what a beginner should do after a practice-led course. For both, it makes the chain from source frame to interpretation explicit. It describes the questions, sampling, exclusions, deduplication, coding, counts, limitations and curriculum implications, then sets out responsible-AI and rights boundaries for research work.

2. Research questions

The study addressed six questions.

Which research-role title families were represented in the accepted public-vacancy corpus?
Which research methods, analytical tasks, outputs and communication duties were explicitly visible in the saved snippets, and in how many accepted vacancies?
What end-to-end professional workflow can be supported by the combination of visible duties without inferring duties that snippets did not show?
Which parts of that workflow should form the core of a beginner-accessible professional curriculum?
What role boundaries separate market research from advertising execution, generic data analysis, product ownership and other adjacent work?
What ethical, privacy, rights and responsible-AI controls are necessary when teaching or performing the workflow?

The questions are descriptive and curriculum-oriented. They do not ask which skill causes employment, which method produces better commercial outcomes, or how often a duty occurs in the global profession. The sampling design cannot answer those causal or population-level questions.

3. Method

3.1 Study design and evidence cut-off

This was a cross-sectional analysis of saved public-search results. The evidence cut-off was 1 September 2026. Thirteen searches were designed to expose research-heavy title and duty combinations, including market research analysts, research executives, consumer or customer insights practitioners, qualitative and quantitative researchers, survey research roles, and roles whose snippets explicitly connected research methods to analysis or reporting.

Twelve saved searches returned 50 results each and one targeted search returned 35, producing a raw frame of 635 records. All results resolved to public LinkedIn job URLs. The thirteen searches provided different search briefs, not independent publisher families. This distinction is important: repeated searches can improve coverage of title and wording variants, but they do not create publisher diversity. The study is therefore a single-platform public-snippet study with multiple discovery briefs.

“Current” in the report title means that a vacancy result appeared in the saved public-search frame on the retrieval date and that no saved occurrence in its dedupe group exposed a stale, ambiguously stale, closed or no-longer-accepting signal under the defined rules. It does not mean that every employer page was opened and independently confirmed to be accepting applications. Nine accepted groups displayed an explicit recent age marker. The remaining 96 displayed no posting-age marker in the saved evidence. No-age visibility is recorded as unknown, not as proof of recency.

3.2 Sampling frame

The unit initially sampled was a public search result. The unit analysed was a unique vacancy group after deduplication. The raw 635 results contained repeated URLs or job IDs because the same vacancy could appear under more than one search brief. Numeric LinkedIn job ID was the primary deduplication key. When no job ID was available, a normalized canonical URL with query and fragment removed was used.

Deduplication reduced 635 results to 569 unique groups; 66 additional occurrences remain in the rejected ledger with pointers to their representatives. Currentness was checked across every occurrence, and any stale, closed or no-longer-accepting signal rejected the group. The representative was selected deterministically by coded-duty count, research-signal count, cleaned-description length and stable source-name tie-breaker. Text from occurrences was never combined, and a cleaner duplicate could not hide adverse status evidence.

3.3 Inclusion criteria

A unique vacancy group was accepted only when every required condition was supported by the saved result.

First, the title had to belong to an accepted direct or adjacent family. Direct patterns covered market research or intelligence and customer, consumer, shopper, audience or experience research and insights. Every bounded adjacent method, research-executive, insights, research-analyst, associate or manager title additionally required explicit market, customer, consumer, shopper, audience, brand, product, commercial, industry, competitor or commercial-insights context in the role's title-plus-cleaned-snippet evidence.

Second, English and foreign-language LinkedIn interface and notification text, People also viewed, related-title strings and job-card tails were removed before coding. The role's title-plus-cleaned-snippet evidence then had to expose a research-domain or collection-method signal. A direct Market Research, Consumer Insights or Customer Insights title supplied role-level research-domain evidence, but never supplied the required output.

Third, the visible description had to contain a substantive output signal. The accepted output families included analysis, synthesis, insight generation, findings, reporting, presentation, storytelling, recommendations, interpretation or decision support. Bare adjectival analytical and title-restatement analytics were explicitly insufficient. This condition prevented either a title or a personal-trait adjective from establishing that the result exposed research work leading to a professional output.

Fourth, excluded work families covered advertising, paid media, campaigns, SEO, content, communications, lead generation, investment, academic, policy, economic, health, child-protection and scientific research. Advertising, campaign, content, communications and lead-generation execution patterns were tested against the cleaned role description as well as the title. Generic analytics titles needed a clear research or insights role signal. Adjacent method roles could inform workflow evidence but could not convert unrelated research domains into market-research demand. Twelve records identified across the v2 and v3 independent reviews as contaminated, weak-evidence or out of scope were forced exclusions and independently checked as absent from the accepted ledger.

Fifth, currentness was group-wide. Any occurrence marked older than 60 days, ambiguously stale, closed, expired or no longer accepting caused rejection. A group with no visible marker remained status-unknown rather than employer-confirmed open.

3.4 Exclusion and disposition

The repaired rules produced 105 accepted vacancies and 530 rejected records, including 66 duplicate occurrences. Overlapping rejection reasons included no substantive cleaned output signal (367), no accepted role context (201), title outside the accepted families (69), duplicate job ID or canonical URL (66), no title-plus-cleaned-snippet research signal (65), a stale or ambiguously stale group occurrence (13), twelve forced independent-review exclusions, sensitive-domain research (five) and closed status (three). Campaign/content/creative-execution scope, content or communications execution, advertising-title and campaign-title exclusions appeared twice each; campaign execution, ecommerce product-sourcing research, economic research, investment/equity research, academic research and generic analytics appeared once each. Reason counts are not mutually exclusive.

The strict output requirement reduced the risk of accepting a vacancy solely because a search engine returned a research-like title. It also creates selection effects. Roles whose snippets focused on recruitment, operations or methods but did not expose an analytical output could be excluded even when the full vacancy would have qualified. The accepted corpus therefore favours snippets that made both research context and output visible.

3.5 Coding method

Accepted snippets were coded through deterministic, case-insensitive phrase patterns. Output acceptance excludes adjectival analytical and title-restatement analytics; the data-analysis duty code requires an analysis or interpretation action or object. Fourteen duty families were defined before the final summary, and all fourteen had at least one visible match:

research brief and design;
sampling and recruitment;
surveys and questionnaires;
interviews and focus groups;
fieldwork and project delivery;
qualitative analysis;
quantitative analysis;
data analysis and interpretation;
synthesis and insight generation;
reporting and presentation;
recommendations and decision support;
secondary research and market intelligence;
stakeholder and client communication; and
quality and research operations.

A vacancy could receive several duty codes because professional work is multi-stage. Counts therefore overlap. A phrase match established only that the duty signal was visible in the cleaned saved snippet. It did not establish proficiency level, time spent, organizational ownership or the absence of other duties. Seven accepted snippets passed the broader research-domain and substantive-output rules but did not match one of the fourteen narrower duty expressions. They remain in the accepted denominator because inclusion and duty coding served different purposes.

Supporting excerpts were selected deterministically from the cleaned evidence near the first duty or output signal and limited to no more than 20 words. Full vacancy descriptions were not retained in the analytical corpus. Every accepted record preserved its public source URL, canonical URL, title, available employer and location text, jurisdiction inferred from the LinkedIn country host, source search, rank, retrieval date, short supporting excerpt, code list, role-context rule, group-wide status result, every occurrence's status, duplicate occurrences and standard limitations.

3.6 Quality assurance and reproduction

The deterministic quality gate passed. It confirmed all 13 inputs and exact counts, 635 raw records, 105 accepted vacancies, unique URLs and IDs, provenance, retrieval date, excerpt length, role context, research and substantive output evidence, all twelve reviewed exclusions, group-wide currentness, clean excerpts, complete accounting, and no full descriptions. Checks cover Expand and the known Prof Commercial Reporting and Analysis GB residue.

The accepted corpus, rejected ledger, sampling-frame statement, task-frequency summary and quality result form one reproducible evidence bundle. Reproduction uses the same thirteen saved inputs, their recorded SHA-256 hashes and deterministic build rules. A later live search could produce a different frame because vacancies and indexing change; it would be a new study version, not a silent update to this one.

3.7 Rights and ethical handling of vacancy evidence

The study retained factual metadata, public URLs, derived codes and short necessary excerpts rather than bulk page text. Search snippets were treated as discovery evidence, not as a licence to republish full job descriptions. The analysis reports aggregated counts and original interpretation. Employer names were preserved only when directly visible, and no applicant or employee personal data was collected.

Public availability was not treated as proof of unrestricted reuse. The study did not bypass authentication, paywalls, access controls, robots restrictions or rate limits. It did not submit applications or create third-party accounts. The same principle applies to future research teaching: access to a source does not itself establish permission to reuse personal, confidential or copyrighted material.

4. Results

4.1 Corpus composition

The 105 accepted vacancies were distributed across 13 populated title families. Market research was largest with 55 (52.4%), followed by customer insights with 18 (17.1%) and consumer insights with 16 (15.2%). Market insights contributed four, market intelligence three, customer research two, and audience research, insights practitioner, product research, research analyst, research executive, research manager and survey insights one each.

Fifty records used a title family other than market research. Ninety-nine passed a direct market or insights title rule, four passed the bounded method-or-insights specialist rule, and two adjacent research titles required explicit commercial context. The adjacent records inform transferable methods; they do not prove that every qualitative or survey researcher works in commercial market research.

4.2 Jurisdiction and source visibility

Jurisdiction came from the LinkedIn country host, not a verified work location. Thirty records had no country-specific host. Singapore accounted for 10; Vietnam nine; Australia eight; the United Kingdom seven; and the United Arab Emirates six. Bahrain, Ireland and Israel contributed three each; Canada, Denmark and India two each; and 20 other visible country hosts one each.

The spread prevents a one-country reading but is not a representative global sample. Search ranking, platform coverage, language, titles and briefs shaped visibility. Singapore, Vietnam, Australia, the United Kingdom, the United Arab Emirates and Bahrain are concentrated; the United States was not reliably inferable for many results. The 30 records without visible jurisdiction are a material limitation.

All records came from LinkedIn public results. Selected representatives came from survey design/reporting (22), direct insights managers (16), consumer/customer insights (14), market-research explicit output (13), U.S. market research (nine), insights-analyst clean v2 (eight), insights explicit output (seven), market-research analyst clean v2 (six), global insights (five), research executive and specialist roles (two each), and methods explicit output (one). The 35-result targeted search added none after cleaning. These are briefs, not evidence families.

Nine accepted groups exposed an explicit recent posting-age marker. The remaining 96 contained no visible age marker. The study therefore cannot claim that all 105 employer postings were independently open on the retrieval date. It can claim that 105 unique results appeared in the saved search frame, met the repaired substantive rules, and had no stale, ambiguously stale, closed or no-longer-accepting signal in any saved occurrence.

4.3 Visible duty frequencies

Table 1 reports the deterministic duty counts. The denominator is 105 accepted unique vacancies. Because one snippet may contain several duty signals, rows overlap and should not be summed.

Visible duty signal	Count	Share of accepted corpus
Surveys and questionnaires	51	48.6%
Reporting and presentation	48	45.7%
Synthesis and insight generation	18	17.1%
Data analysis and interpretation	15	14.3%
Quantitative analysis	15	14.3%
Recommendations and decision support	13	12.4%
Qualitative analysis	12	11.4%
Stakeholder and client communication	12	11.4%
Fieldwork and project delivery	8	7.6%
Interviews and focus groups	7	6.7%
Quality and research operations	3	2.9%
Sampling and recruitment	2	1.9%
Secondary research and market intelligence	2	1.9%
Research brief and design	1	1.0%

Survey and questionnaire language was the most visible duty family, followed by reporting and presentation. This does not prove that either duty dominates all market-research work. It shows that survey and output wording were frequently exposed by the selected searches and public snippets. These responsibilities are easy for a vacancy writer or index to name compactly: questionnaire design, survey programming, survey data, reports or presentations can appear in a short result. By contrast, a complete sampling decision, ethics review or stakeholder intake process may require more text than a snippet provides.

Reporting and presentation appeared in 48 records. That count supports the interpretation that research jobs are not completed when data collection ends. Findings must be organized and communicated. Reporting may mean a written report, presentation deck, story flow or client deliverable; the coding family groups these expressions because all represent transfer of interpreted evidence to another person. The matcher explicitly excludes grammatical reporting to the phrases so a supervisory relationship cannot be mistaken for a research deliverable.

Data analysis and interpretation and quantitative analysis each appeared in 15 snippets. The categories are not identical. Quantitative analysis captured explicit quantitative, statistical, regression, conjoint or forecasting language. Data analysis and interpretation required an analysis or interpretation action or object and excluded bare analytics or adjectival analytical. A snippet could match both, so the counts cannot be added as different vacancies or treated as a complete measure of analytical work.

Qualitative analysis was visible in 12 snippets, while interviews or focus groups appeared in seven. This difference is plausible within the coding system: a role can mention qualitative research or thematic work without naming an interview or focus group in the snippet. It would be incorrect to conclude that five vacancies performed qualitative analysis without interviews; the source does not show the complete method mix. The safe conclusion is that qualitative analytical language was visible more often than explicit interview or focus-group language.

Fieldwork or project delivery appeared in eight snippets. This signal matters because research quality depends on execution as well as design. Recruitment progress, fieldwork monitoring, changes to the instrument, response problems, handoffs and schedule control can affect the evidence that reaches analysis. A course that teaches only question writing and statistics would omit this operational middle.

Synthesis or insight-generation language appeared in 18 snippets. Recommendation or decision-support language appeared in 13, and stakeholder or client communication in 12. These are lower counts than reporting, but they identify a decision-facing layer of the role. Reporting can describe facts and analysis; synthesis relates evidence across sources; decision support connects the result to a choice. The distinctions are useful for curriculum even though the snippets cannot establish how employers separate them in practice.

Quality or research operations appeared in three snippets; sampling or recruitment and secondary research or market intelligence appeared in two each; and research brief/design appeared once. These low visible counts must not be interpreted as evidence that the activities are unimportant. The inclusion design required a visible output, while short snippets often highlight deliverables or technical methods rather than upstream planning and controls. Sampling is logically necessary for primary research even when the word does not appear in a snippet. The corpus supports teaching it because survey, interview and fieldwork tasks cannot be performed responsibly without defining who is represented and how participants enter the study. That is a workflow dependency, not an inferred count of unseen vacancy duties.

4.4 From duty signals to an end-to-end workflow

The corpus supports a five-stage interpretation of market research work.

The first stage is decision and study framing. Research-brief/design wording appeared in one snippet, while stakeholder or client communication appeared in 12. These low counts are snippet-visibility results, not evidence that professional studies usually begin without a question. Every accepted role connected a research domain to an output, and producing an appropriate output requires knowing the question and audience. The defensible curriculum implication is to teach how to turn a business uncertainty into a research brief and questions while identifying this mainly as workflow-dependency reasoning rather than a high-frequency vacancy claim.

The second stage is evidence and participant design. Sampling and recruitment appeared explicitly in two snippets, surveys in 51, interviews or focus groups in seven and secondary research in two. The combined pattern shows several evidence routes. It does not justify treating each role as mixed-method. Instead, a professional foundation should help a learner choose among secondary, qualitative, quantitative and combined designs; define the target population; build a usable sampling frame; and state limitations appropriate to the chosen route.

The third stage is instrument and fieldwork execution. Questionnaire work was widely visible, interviews and focus groups were present, and fieldwork or project delivery appeared in eight snippets. This stage includes writing neutral questions, piloting, recruitment control, participant communication, moderation, documentation and monitoring. It also requires operational judgment: when a quota is filling unevenly, when a question is confusing, when response quality is weak and when a fieldwork change must be recorded rather than hidden.

The fourth stage is analysis and synthesis. Quantitative, qualitative and general data-analysis signals were all material. Analysis must follow the design. Interview material may be coded into themes while preserving counterexamples. Survey data may be summarized and compared with clear bases and proportionate uncertainty. Neither method should be asked to prove what it cannot. Synthesis then compares sources, identifies agreement and contradiction, and distinguishes an observed finding from the researcher’s interpretation.

The fifth stage is insight communication and decision handoff. Reporting or presentation was visible in 48 snippets, and smaller groups explicitly named synthesis, stakeholder communication or recommendations. The professional output is therefore not a table of frequencies alone. It is a structured account of the question, method, sample, findings, limitations, interpretation and implications. A decision owner should be able to see which statements are directly supported, what remains uncertain and what next step is proportionate.

This five-stage sequence is an interpretation of dependencies among visible tasks, not a claim that all 105 vacancies described the same process. Specialist roles may enter at one stage. A qualitative researcher may concentrate on recruitment, moderation and thematic analysis. A survey researcher may own questionnaire and sampling work. An insights manager may focus on synthesis and stakeholder decisions. A beginner course can still teach the shared workflow so learners understand their own task and the inputs and handoffs around it.

5. Professional role boundaries

5.1 Research is not advertising execution

The sampling frame intentionally excluded advertising, paid-media, campaign, content and search-engine-optimization titles. The boundary is functional rather than organizational. A researcher may investigate brand perceptions, advertising concepts or customer response, but the researcher does not thereby own media buying, campaign deployment, creative production or acquisition targets. The research artifact documents evidence and its limits; the advertising artifact specifies or executes promotional activity.

This distinction protects both curriculum coherence and professional judgment. If survey design is taught mainly as a prelude to campaign optimization, learners may miss research questions that serve product, customer experience, pricing, service or market-entry decisions. If an insight report is evaluated by whether a campaign performed well, the study may be pressured to validate a preferred action. A research-first course should instead evaluate whether the question, sample, instrument, analysis and interpretation support the stated conclusion.

5.2 Research is not generic data analysis

Market researchers analyse data, and 15 snippets explicitly exposed the tightened data-analysis or interpretation code. The existing discipline of data analysis nevertheless has a wider tool and data focus. A market research workflow begins earlier: it decides what should be asked, who should be represented, how responses should be collected and how method choices shape interpretation. It also includes interviews, qualitative coding, questionnaire design, recruitment and participant handling.

The appropriate boundary is not that market researchers avoid spreadsheets, statistics or visual evidence. It is that tools serve a research design. A general data-analysis course may start with an authorized dataset and a business question; market-research practice may be responsible for creating the evidence through a sample and an instrument. Curriculum should teach enough analysis to interpret research evidence while avoiding a second full programme in SQL, programming, business intelligence or data engineering.

5.3 Research informs but does not own every decision

Research can support product, commercial, service and strategy teams. The researcher defines and explains what the study found and what it may imply. The named decision owner remains responsible for the action, budget, legal basis, operational change or market commitment. This separation is especially important when evidence is incomplete or groups are affected differently.

The same boundary applies to specialist decisions. A researcher should not decide that personal-data processing is lawful, that a health claim is valid, that a financial offer is suitable or that employee monitoring is permissible. Those questions require qualified owners and applicable local procedures. Good research makes the unresolved question visible and routes it; it does not hide it inside a polished recommendation.

6. Curriculum implications

6.1 A workflow-led architecture

The findings support four connected learning blocks. The first should cover decision framing, research questions, secondary evidence, method selection and ethical data handling. The second should cover target populations, sampling frames, recruitment, screeners, interview guides, interview conduct and qualitative coding. The third should cover survey blueprints, question and scale design, sampling and sample-size decisions, piloting, fieldwork quality and survey analysis. The fourth should cover triangulation, segment interpretation, insight development, visual evidence, report writing, presentation and handoff.

This sequence follows professional dependencies. Learners should not calculate a sample before defining the population and decision. They should not write questions before mapping what must be measured. They should not code interview material before understanding how it was collected. They should not present an insight before checking the source, base, contradiction and limitation. A workflow-led architecture helps a beginner know what to do first, what input to request and what output to hand off.

6.2 Sampling and recruitment deserve explicit treatment

Sampling or recruitment language appeared in only two snippets, yet it requires dedicated teaching. This is not an attempt to inflate a low-frequency signal. Sampling is a necessary condition for interpreting both qualitative and survey evidence. Without a defined target population and participant path, the learner cannot explain who is represented, who may be missing or whether a percentage supports a wider claim.

The curriculum should distinguish a target population, sampling frame, invited sample and achieved sample. It should explain probability and non-probability approaches in plain language. It should show how inclusion criteria and quotas can improve purposeful coverage without turning a convenience sample into a representative one. It should teach that a conventional probability-sample margin of error does not belong on a non-probability sample merely because software can calculate a number.

6.3 Interviews must include conduct, not only guide writing

Seven snippets explicitly mentioned interviews or focus groups, and 12 exposed qualitative analysis. A practical course should teach both collection and interpretation. Guide writing alone does not prepare a learner to open a session, explain participation, ask neutral questions, probe without leading, manage time, document context and close respectfully. Analysis alone can obscure how moderator choices shaped the material.

Required self-study practice can remain individual. Learners can work with an original synthetic transcript, identify weak questions, select better probes, create a fieldwork record and code supplied excerpts. An optional authorized workplace activity may involve a real interview under employer procedures, but course completion should not require another learner, an unconsented participant or real personal data.

6.4 Survey learning should connect design, fieldwork and interpretation

Survey and questionnaire language was the most visible duty family. The course should therefore give it substantial space without becoming a software tutorial. Learners need to translate a research question into a measurement plan, choose an appropriate respondent, write neutral items and response options, order the questionnaire, pilot it, monitor fieldwork, prepare the response data and interpret distributions and cross-tabs.

Each step should make common failure modes visible. A double-barrelled item can produce an answer whose meaning is unclear. An imbalanced scale can push responses. A low base can make a percentage unstable. Uneven nonresponse can weaken comparison. A statistically detectable difference may be too small to matter to the decision. Weighting may adjust known dimensions but cannot repair every selection problem. Teaching these relationships is more valuable than requiring one survey platform.

6.5 Analysis must preserve method limits

The analytical signals support qualitative and quantitative pathways. Qualitative practice should cover an explicit codebook, consistent coding, case comparison, theme development, contradictory evidence and the difference between depth and prevalence. Quantitative practice should cover data preparation, bases, missing responses, descriptive summaries, cross-tabs, uncertainty and practical importance. Advanced branded methods or vendor workflows are not necessary for the foundation.

Mixed-method synthesis should not become a ceremony in which every source is forced to agree. A triangulation matrix can show which question each source addresses, whether sources are independent, where they converge, where they contradict and what additional evidence would change the interpretation. Preserving disagreement is a professional strength when the disagreement reveals segment, context or method differences.

6.6 The insight report should be the principal capstone

Reporting or presentation was visible in 48 snippets, making it an appropriate final professional output. A capstone can provide one fictional company, one decision and one bounded evidence pack containing secondary summaries, interview excerpts and survey data. The learner’s principal deliverable should be a Market Research Insight Report rather than an audit or assembly of every course template.

The report should state the business decision, research questions, methods, sample, evidence cut-off, findings, contradictions, limitations, interpretations, implications and proportionate recommendation. Visuals should serve the decision and expose their bases. The capstone should be judged on support and restraint as well as clarity: every material claim should trace to supplied evidence, and the recommendation should be no stronger than the design permits.

6.7 Professional artifacts

A workplace-ready course should teach reusable artifacts in ordinary professional language: a research decision brief, question map, source review, method plan, ethics and data-handling plan, sampling frame, screener, interview guide, fieldwork record, qualitative codebook, survey blueprint, questionnaire, sample plan, pilot log, analysis workbook, triangulation matrix, insight-development matrix, insight report and decision handoff. Each should solve a recognizable work problem, identify its user, include a usable blank template and show an original completed example with reasoning.

7. Responsible AI in market research

AI can assist research work, but it cannot become evidence. A model can help challenge a research question, identify ambiguity in a questionnaire, suggest alternative codes for review, test whether a claim is supported by a supplied summary or improve the structure of a report. Those uses still require a person to supply authorized inputs, verify every retained statement and own the method and decision.

Several prohibited substitutions follow from that principle. AI must not invent respondents, simulate interviews and present them as collected evidence, create quotations, fill missing survey responses, fabricate market sizes, decide a legal basis, grant consent or claim that a sample is representative. A fluent synthetic answer is not a participant statement. A generated percentage is not a measured result. A plausible citation is not a verified source.

Privacy and confidentiality constrain tool use. Identifiable transcripts, recordings, customer records, contact lists, sensitive attributes and confidential commercial material should not be entered into an unapproved external AI service. Even de-identified text requires a realistic re-identification assessment and employer authorization. Synthetic course cases avoid this problem for required learning activities.

AI-assisted analysis also needs method-specific checks. For a questionnaire, the learner should inspect leading wording, double questions, assumptions, scale balance, response completeness and routing. For qualitative coding, the learner should compare suggestions with the actual text, retain counterexamples and avoid turning repeated language into prevalence. For survey analysis, the learner should rerun every material calculation, verify bases and denominators, and reject invented weights or benchmarks. For an insight report, the learner should trace each claim to the supplied source and label unresolved uncertainty.

The appropriate disclosure records the purpose of AI assistance, the inputs supplied, the suggestions retained, the edits made and the checks performed. It does not imply that a particular model version validates the research. Human oversight is not an approval phrase appended at the end; it is the sequence of source, method and calculation checks that makes the retained work defensible.

8. Rights, ethics and participant protection

Market research often processes information about people. The European Commission summarizes GDPR data-protection principles as lawfulness, fairness and transparency; purpose limitation; data minimisation; accuracy; storage limitation; integrity and confidentiality; and accountability. A general professional course should therefore teach privacy and ethics principles without offering jurisdiction-specific legal advice. Core controls include a specific stated purpose, clear participant information, voluntary participation, proportionate withdrawal handling, data minimisation, limited retention, appropriate security, restricted access and accountable deletion. The organization must identify the applicable legal basis and local procedure through its qualified owners; consent should not be presented as the universal legal basis for every study.

Recording interviews requires special care. The participant should know whether audio, video, screen content or notes will be captured, how each will be used, who will receive it and how long it will be kept. Local recording law and organizational policy may impose additional conditions. Required course activities should use supplied synthetic transcripts rather than asking learners to record real people.

Minors, health or financial information, sensitive personal characteristics, covert observation, employee monitoring and consequential individual profiling create higher risks. They are unnecessary for a foundation capstone and should be excluded from required exercises. A professional encountering them should stop the ordinary workflow and involve the relevant privacy, legal, ethics, safeguarding or domain owner.

Professional codes and standards can inform good practice, but rights must be respected. The ICC/ESOMAR Code can be cited and paraphrased as a source of high-level ethical principles; its text, checklists, marks and logos should not be reproduced or used to imply membership or endorsement. ISO 20252 can be referenced at the level of its public scope, but a course should not reproduce protected service requirements or claim compliance or certification. The 2019 edition was in revision at the study date, making fixed compliance language particularly inappropriate.

Commercial trademarks and vendor materials require the same discipline. Bain states that NPS and related marks require attribution and a commercial licence. A course can teach a generic recommendation-intent or loyalty measure without using protected branding. Commercial survey platforms, questionnaire banks, software workflows and training examples should not be copied. Original tool-neutral templates, synthetic datasets and author-created examples allow the curriculum to teach the method without appropriating another provider’s expression.

9. Limitations

This study has substantial and explicit limitations.

First, the source consists of public search snippets rather than independently opened full vacancy pages. The snippets may omit duties, qualifications, employer identity, location, age and context. Duty frequencies are therefore lower-bound visibility counts. Absence of a code means only that the defined phrase was not visible; it does not mean the duty was absent from the job.

Second, every raw result came from LinkedIn public job URLs. Thirteen search briefs improved discovery coverage but did not provide thirteen publisher families. Platform indexing, ranking and normalization may favour certain titles, countries, employers and expressions. The corpus cannot estimate prevalence across all job boards or employers.

Third, the sample was purposive, not probabilistic. Searches were designed to find research-heavy roles and explicit output language. The accepted 105 vacancies are not a random sample of the global profession. Percentages describe the accepted corpus only and have no sampling margin of error for a wider vacancy population.

Fourth, the strict inclusion rule created selection effects. Requiring a visible substantive analysis, synthesis or reporting signal in the cleaned description contributed to 367 rejections. Bare adjectival analytical and title-restatement analytics did not qualify. Some rejected snippets may have described valid research roles in their full text. The report does not infer their hidden content and does not use them to expand the accepted count.

Fifth, recency is only partially visible. Nine accepted groups showed an explicit recent age marker; 96 did not show an age. Every occurrence of an accepted group was checked for stale, ambiguously stale and closed language, but a missing marker is not proof that a posting was open. The title’s word “current” refers to the retrieval frame and the repaired no-adverse-status rule, not to employer-confirmed live status.

Sixth, title and jurisdiction fields may be incomplete or normalized. Jurisdiction was inferred from the LinkedIn host and may not match the work location. Thirty records lacked a visible country-specific host. Employer text was unavailable in many snippets. The study therefore avoids employer-level or country-level conclusions.

Seventh, deterministic phrase coding trades nuance for reproducibility. Synonyms outside the pattern set can be missed, while a matched word can represent different levels of responsibility. The categories overlap and cannot be added into a single total. Seven accepted snippets matched the broader research and substantive-output tests without matching a narrower duty category.

Eighth, six accepted records entered through an explicitly documented adjacent-role route rather than a direct market/customer/consumer/audience/commercial-insights title. Four were bounded method or insights specialists and two were adjacent research titles with commercial context. They strengthen method coverage but must not be interpreted as six additional direct market-research roles.

Ninth, the study describes visible professional requirements but does not measure actual performance, hiring outcomes, salary, course demand, learning effectiveness or employer satisfaction. The BLS context is a combined U.S. occupation and does not resolve these limitations.

Finally, curriculum implications combine empirical visibility with professional dependency reasoning. Sampling is taught not because two visible mentions prove universal prevalence, but because surveys and interviews cannot be interpreted responsibly without a population and participant path. Such dependency-based decisions are identified as interpretations rather than presented as vacancy counts.

10. Conclusion

The accepted corpus presents market research as an evidence-to-decision workflow rather than an advertising function. Survey and questionnaire work and reporting or presentation were the most visible duties. Quantitative analysis, data interpretation, qualitative analysis and fieldwork formed a substantial middle, while smaller groups exposed synthesis, stakeholder communication, recommendations, interviews, design, sampling, secondary research and quality operations.

No single frequency table defines the complete role. The stronger result comes from the sequence connecting the signals. A professional begins with a decision and research question, chooses evidence and participants, designs instruments, conducts fieldwork, analyses the resulting material, integrates agreement and contradiction, and communicates findings and limitations. Specialist roles may own only part of that sequence, but beginners benefit from seeing the complete process and its handoffs.

The evidence supports a Professional Certificate in Market Research centred on sampling, interviews, surveys, analysis and insight reporting. It does not support a course dominated by advertising, campaign execution or generic analytics software. A credible curriculum should teach original tool-neutral methods, use synthetic practice data, make privacy and rights boundaries proportionate, and treat AI as a checked assistant rather than a source or respondent.

The report’s conclusions remain bounded to 105 unique public vacancy snippets retrieved on 1 September 2026 and accepted under the repaired direct-role, adjacent-role, substantive-output, cleaned-evidence and group-currentness rules. They should be revised only through a new, documented evidence version. Within that boundary, the corpus provides a defensible basis for a practical research-role curriculum and for one final professional deliverable: a market research insight report whose claims, limitations and recommendations can be examined by a real decision owner.

Continue learning

Apply the evidence from this report through MTF Institute's Professional Certificate in Market Research. The programme turns the identified capabilities into structured theory, guided AI practice and reusable workplace artifacts.

References

U.S. Bureau of Labor Statistics. Market Research Analysts. Occupational Outlook Handbook. https://www.bls.gov/ooh/business-and-financial/market-research-analysts.htm
U.S. Bureau of Labor Statistics. Occupational Projections and Characteristics, 2025–2035. https://www.bls.gov/emp/tables/occupational-projections-and-characteristics.htm
European Commission. Principles of the GDPR. https://commission.europa.eu/law/law-topic/data-protection/information-business-and-organisations/principles-gdpr_en
ICC and ESOMAR. ICC/ESOMAR International Code on Market, Opinion and Social Research and Data Analytics. https://community.esomar.org/uploads/public/knowledge-and-standards/codes-and-guidelines/ICCESOMAR-International-Code_English.pdf
International Organization for Standardization. ISO 20252:2019 — Market, opinion and social research, including insights and data analytics. https://www.iso.org/standard/73671.html
Bain & Company. NPS Trademarks and Licensing. https://prod.netpromotersystem.com/resources/trademarks-and-licensing/
LinkedIn public job results. Record-level public URLs and short supporting excerpts are preserved in the dated MTF Institute research source ledger for the 105-vacancy corpus.

Data availability and reproducibility statement

The verified archival record is published at Zenodo DOI 10.5281/zenodo.22254955; the direct public PDF is available here. The publication date is 2 September 2026 and the technical report number is MTF-CF-RR-2026-09-01-25.

The dated research bundle contains the accepted record ledger, rejected record ledger, sampling-frame statement, deterministic task-frequency summary, build rules and quality result. The retained evidence includes public URLs, factual metadata, derived codes and short necessary excerpts of no more than 20 words. Full vacancy descriptions are not redistributed. Reproduction rebuilds the corpus from the same thirteen saved public-search snapshots, their hash manifest and the same deterministic rules. A new retrieval should be published as a new study version because vacancy and indexing state changes over time.

Appendix A — Data dictionary

A.1 Corpus-level fields

Field	Meaning
`schema`	Versioned structure identifier for the corpus.
`course_key`	Stable internal course identity associated with the evidence bundle.
`retrieval_date`	Date on which the saved public-search frame was retrieved.
`scope`	Plain-language description of the accepted research-role boundary.
`record_count`	Number of accepted unique vacancy groups.
`records`	Array of accepted vacancy records.

A.2 Accepted-record fields

Field	Meaning and interpretation
`vacancy_id`	Stable derived record identifier based on LinkedIn job ID or a canonical-URL hash.
`job_id`	Numeric LinkedIn job identifier when visible.
`role_title`	Cleaned role title derived from the public result.
`title_original`	Original title string retained from the saved result.
`employer`	Employer text when directly visible; null when not reliably exposed.
`location`	Location text when directly visible; null or incomplete when not exposed.
`jurisdiction`	Country inferred from the LinkedIn host; not a verified work location.
`canonical_url`	Normalized public vacancy URL used for deduplication and reference.
`source_url`	Original public URL of the selected representative occurrence.
`source_publisher`	Public result provider; LinkedIn for this corpus.
`source_domain`	Hostname of the selected public result.
`source_snapshot`	Saved search snapshot from which the representative was selected.
`source_search_id`	Retrieval identifier for the saved public-search operation.
`source_rank`	Position of the occurrence within its saved search result.
`retrieval_date`	Evidence cut-off date for the record.
`supporting_excerpt`	Deterministically selected evidence window of no more than 20 words from the cleaned description.
`excerpt_word_count`	Count of words in the retained supporting excerpt.
`coded_duties`	Zero or more deterministic duty-family labels visible in the snippet.
`evidence.title_family`	First matching accepted title family.
`evidence.role_context_rule`	Direct-market, bounded method/insights-specialist or commercial-context adjacent rule that admitted the role.
`evidence.research_domain_signals`	Visible research-domain phrase families.
`evidence.analysis_synthesis_reporting_signals`	Visible output phrase families used for inclusion.
`evidence.posting_age`	Group-wide age/availability result after every saved occurrence is checked, including stale, ambiguous and closed flags.
`limitations`	Standard record-level cautions about snippet, age and normalized metadata.
`source_occurrences`	All saved-search occurrences belonging to the same dedupe group, including each occurrence's posting-status result.
`duplicate_occurrence_count`	Number of additional occurrences beyond the selected representative.
`acceptance_reason`	Plain-language statement that the record met title, research and output rules.

A.3 Rejected-record fields

Rejected records use the same core provenance and evidence fields where available. duplicate_of points to the selected vacancy when the record was an additional occurrence. rejection_reasons is an array because one record can fail several rules. Reason counts must therefore not be summed as mutually exclusive categories.

Appendix B — Coding framework

B.1 Title families

The first matching title pattern assigned one of the following families: market_research, market_insights, market_intelligence, consumer_insights, customer_insights, customer_research, audience_research, customer_experience_research, shopper_insights, research_executive, survey_research, survey_insights, qualitative_research, quantitative_research, mixed_methods_research, product_research, experience_research, insights_practitioner, research_analyst, research_associate or research_manager. No accepted records in the final corpus were classified as shopper insights, customer-experience research or research associate, so those labels do not appear in the frequency summary.

B.2 Research-domain signals

Research-domain coding was applied only after LinkedIn UI, notification, People also viewed and related-job tails were removed. It recognized explicit market research; consumer, customer, shopper or audience insights; consumer or customer research; primary research; research projects, studies, methods or methodology; qualitative or quantitative work; surveys or questionnaires; interviews or focus groups; fieldwork; sampling or recruitment; and consumer behaviour. At least one cleaned research-domain or method signal was required for acceptance.

B.3 Output signals

Output coding recognized analysis, synthesis, insight generation or findings, reporting, presentation or storytelling, recommendations or decision support, and interpretation. At least one substantive output signal was required for acceptance. Bare adjectival analytical and title-restatement analytics did not qualify. This gate was broader than the fourteen duty categories and was used to prevent title-only or trait-only admission.

B.4 Duty categories

research_brief_and_design: research design, study design, research questions, briefs, plans, methods or methodology.
sampling_and_recruitment: samples, sampling, quotas, respondent or participant recruitment.
survey_and_questionnaire: surveys, questionnaires, survey programming or polls.
interviews_and_focus_groups: interviews, interviewing, focus groups or moderation.
fieldwork_and_project_delivery: fieldwork, fielding, research projects or studies.
qualitative_analysis: qualitative, thematic, coding, ethnographic or semiotic wording.
quantitative_analysis: quantitative, statistical, regression, conjoint or forecasting wording.
data_analysis_and_interpretation: data or report analysis; analysing data; interpreting data, findings or results; or analysis of data, findings, results, markets or surveys. Bare analytics and adjectival analytical are excluded.
synthesis_and_insight_generation: synthesis, insight generation, findings or themes.
reporting_and_presentation: reports, reporting, presentations, storytelling or story flows.
recommendations_and_decision_support: recommendations, decision support or decision-making language.
secondary_and_market_intelligence: secondary or desk research, competitive or market intelligence, competitor analysis or market trends.
stakeholder_and_client_communication: clients, stakeholders, business questions or client deliverables.
quality_and_research_operations: quality checks, data quality, quality standards, research operations or documentation.

B.5 Coding interpretation rules

Codes indicate visible phrase presence, not verified responsibility breadth or proficiency.
One vacancy may receive several codes; categories overlap.
Absence of a code means the phrase was not visible in the saved snippet.
No missing duty may be inferred from title, employer, sector or another vacancy.
Qualitative wording does not prove interviews were used; survey wording does not prove probability sampling.
Recommendation language does not prove the researcher held decision authority.
Interface commands, related-job titles and another occurrence's cleaner description cannot establish the advertised role's research or output evidence.
Deterministic coding supports reproduction but does not replace a full-text occupational study.

Appendix C — Exact corpus counts

C.1 Disposition

Raw saved public-search results: 635.
Unique dedupe groups: 569.
Accepted unique vacancies: 105.
Rejected records including duplicate occurrences: 530.
Duplicate occurrences rejected: 66.

C.2 Posting-age visibility

Explicit recent marker: 9.
No age marker visible in the saved snippet: 96.

C.3 Title families

Market research: 55.
Customer insights: 18.
Consumer insights: 16.
Market insights: 4.
Market intelligence: 3.
Customer research: 2.
Audience research: 1.
Insights practitioner: 1.
Product research: 1.
Research analyst: 1.
Research executive: 1.
Research manager: 1.
Survey insights: 1.

C.4 Interpretation safeguard

All counts in this appendix describe the accepted bounded corpus. They are neither global prevalence estimates nor counts of live employer-confirmed openings. The BLS employment and openings figures cited in the report describe the combined occupation “Market research analysts and marketing specialists” and are not part of this 105-vacancy corpus.