Market research glossary

A

A/B testMethods

An A/B test randomly splits users between two versions – a page, an email, an ad – and compares conversion rates. Significance is tested with a two-proportion z-test. Fix the sample size in advance: stopping as soon as a result looks significant ("peeking") inflates false winners. Check your result with the A/B test significance calculator.

Advertising post-testStudy types

An advertising post-test measures a campaign's effect after it has run: unprompted and prompted recall, recognition, brand linkage and shifts in attitudes. To isolate the effect of the advertising, results are compared with a pre-campaign baseline or with a control group that was not exposed to the campaign.

Advertising pre-testStudy types

An advertising pre-test evaluates an ad before it airs – as a finished spot, animatic or storyboard – for attention, recall, correct brand attribution, message take-out and emotional response. It allows changes while they are still easy to make. Methods and benchmarks are covered in our ad testing report.

Anonymisation / pseudonymisationQuality & law

Anonymisation irreversibly removes the link between data and an individual, so the data are no longer personal data under the GDPR. Pseudonymisation replaces identifiers with codes while the key is kept separately; the data remain personal data. A common mistake is to treat a dataset with names removed as anonymous.

Attention checkQuality & law

An attention check tests whether respondents are reading – for example, a grid row that says "Select 'Disagree' for this row". A failed check is flagged and judged together with other indicators. Too many checks, or trick questions, irritate conscientious respondents and can raise drop-out.

B

B2B researchStrategy & analysis

B2B research studies companies and institutions through the people who make or influence purchase decisions – owners, managers, procurement and technical specialists. Populations are small and access is difficult, so telephone and in-depth interviews and smaller samples are common. Verifying each respondent's role matters more than in consumer work.

BenchmarkingStrategy & analysis

Benchmarking compares your metrics or processes with those of competitors, an industry average or best practice. The comparison is meaningful only when definitions and methods match: an NPS from an online survey and one from telephone interviews are not directly comparable.

Brand awareness (aided, unaided, top of mind)Metrics

Brand awareness is the share of people who know a brand. Unaided (spontaneous) awareness is measured without prompting; aided awareness means recognising the brand from a list or logo; top of mind is the first brand named unprompted. Aided awareness is always at least as high as unaided.

Brand equityMetrics

Brand equity is the added value a brand name gives a product compared with an unbranded equivalent: the ability to command a higher price, loyalty and easier acceptance of new products. It is measured through awareness, associations, perceived quality and price premium, often with conjoint analysis.

Brand funnelMetrics

The brand funnel tracks the stages awareness → familiarity → consideration → purchase → preference or loyalty. Conversion between adjacent stages shows where a brand loses people: high awareness with low consideration points to an image problem rather than a visibility problem.

See also: Brand awareness (aided, unaided, top of mind)

Brand trackingStudy types

Brand tracking is the regular measurement of awareness, consideration, image and usage for a brand and its competitors, using the same method in every wave. It shows whether campaigns and competitor moves are shifting positions. Waves must be large enough to separate real change from sampling noise. See our brand health tracking report.

Buyer personaStrategy & analysis

A buyer persona is a semi-fictional profile of a typical customer in a segment: goals, problems, choice criteria and information channels. It is useful when built on interview and survey data; a persona written from assumptions mainly confirms what the team already believed.

C

CAGR (compound annual growth rate)Metrics

CAGR is the constant yearly rate at which a starting value would grow to reach the end value: CAGR = (end value ÷ start value)^(1/n) − 1, where n is the number of years. It smooths volatility and so hides years of decline. Always state the start and end years.

CAPI (computer-assisted personal interviewing)Methods

CAPI is face-to-face interviewing in which the interviewer uses a tablet or laptop running the programmed questionnaire. It suits hard-to-reach groups, product demonstrations and on-site work in shops, at events or in homes. It needs more time and logistics than online or telephone methods, and interviewer effects must be monitored.

CATI (computer-assisted telephone interviewing)Methods

CATI is telephone interviewing in which the interviewer reads questions from a screen and keys answers into software that handles routing and quotas. It works well for B2B, for older respondents and for short questionnaires. Long phone interviews lead to break-offs, and scales with many points are hard to follow by ear.

Causal research / experimentStudy types

Causal research tests whether changing one factor – price, message, pack design – causes a change in behaviour. Participants are assigned at random to groups that differ only in that factor, so any difference in outcome can be attributed to it. An A/B test is the simplest form; in-store tests need matched control stores.

See also: A/B test

CAWI (computer-assisted web interviewing)Methods

CAWI is a self-completion web survey answered on a computer or phone, usually after an invitation from an online panel or a customer list. It is fast and supports images, video and complex logic. It reaches only people who are online and needs quality controls for speeders, straightliners and inattentive respondents.

CES (customer effort score)Metrics

CES measures how easy it was for a customer to resolve an issue or complete a task – for example, agreement with "The company made it easy to handle my issue" on a seven-point scale. It is most useful for service and online processes, where effort predicts repeat contacts and abandonment.

Churn rateMetrics

Churn rate is the share of customers lost during a period: customers lost ÷ customers at the start of the period. Monthly and annual churn do not convert linearly: 2% a month gives 1 − 0.98¹² ≈ 21.5% a year, not 24%. Research is used to find out why customers leave.

Closed questionMethods

A closed question offers predefined answers: single or multiple choice, a scale or a ranking. Answers are quick to process and compare across groups. Options must be exhaustive and mutually exclusive, with "Other" and "Don't know" where needed, so that no respondent is forced into an answer that does not fit.

Cluster analysisSampling & statistics

Cluster analysis groups respondents so that people within a group are similar and the groups differ from one another. K-means and hierarchical methods are the most common. The number of clusters is chosen on statistical criteria and on whether the resulting segments make sense and can be acted on.

See also: Market segmentation

Cluster samplingSampling & statistics

Cluster sampling randomly selects groups – towns, schools, stores – and then interviews units within them. Fieldwork becomes more efficient because interviews are concentrated. People in the same cluster tend to resemble each other, so precision is lower than for a simple random sample of the same size; this loss is called the design effect.

Coding of open answersMethods

Coding groups verbatim answers into categories – the code frame – so they can be counted and compared. The frame is built from a subset of answers and extended as new themes appear. AI tools speed up the work, but a researcher should check the output, especially for irony and answers that mix several points.

Competitive intelligenceStrategy & analysis

Competitive intelligence is the legal and ethical collection and analysis of information about competitors: products, published price lists, campaigns, job adverts, accounts filed at Companies House or other business registries. It excludes industrial espionage, inducing employees to breach confidentiality and any attempt to obtain information that is not legitimately available.

Concept testStudy types

A concept test shows a description of a new product or service to the target group before it is developed and measures appeal, clarity, perceived newness and purchase intent. In a monadic design each respondent sees only one concept, which gives a cleaner comparison between options. See our concept and product testing report.

Confidence intervalSampling & statistics

A confidence interval is the range around a survey estimate that, at the chosen confidence level, is expected to contain the true value: 42% with a ±3.1-point margin gives 38.9–45.1%. A common mistake is to assume that overlapping intervals for two groups prove there is no significant difference; that needs a direct test.

Confidence levelSampling & statistics

The confidence level states how often the method would produce an interval containing the true value if the survey were repeated many times. The convention is 95% (z = 1.96); 90% (z = 1.645) and 99% (z = 2.576) are also used. A higher level requires a larger sample or gives a wider interval.

Conjoint analysisMethods

Conjoint analysis asks respondents to choose between product profiles that combine different features and prices. A statistical model then estimates the value (utility) of each feature level and willingness to pay, and simulates market share for alternative offers. Choice-based conjoint (CBC) is the most common form.

See also: Willingness to pay

Convenience sampleSampling & statistics

A convenience sample includes whoever is easiest to reach: website visitors, social media followers, colleagues. It is fine for testing a questionnaire or gathering first impressions. It should not be used to estimate market shares or penetration, because participants almost always differ from the wider population.

CorrelationSampling & statistics

Correlation measures the strength and direction of a linear relationship between two variables; Pearson's r runs from −1 to +1. Correlation does not prove causation: both variables may depend on a third. A value close to zero does not rule out a strong non-linear relationship.

Cross-tabulationSampling & statistics

A cross-tabulation shows how one variable is distributed across the categories of another – satisfaction by region or age group, for example. It is the standard format of survey data tables. When reading one, check the base of each column: a percentage based on 40 people is far less reliable than one based on 400.

CSAT (customer satisfaction score)Metrics

CSAT measures satisfaction with a product or a specific interaction, usually on a 1–5 scale, and is most often reported as the share answering 4 or 5. It works best straight after a purchase or service contact. How it fits with NPS and CES is explained in our customer satisfaction research report.

Custom (ad hoc) researchStudy types

Custom research is designed around one client's question: the target group, questionnaire, sample and analysis follow from the decision ahead. The client owns the results. It takes longer than buying a published report, but it produces evidence your competitors do not have and can be scoped to exactly the markets you sell in.

Customer journeyStudy types

The customer journey is the sequence of stages and touchpoints a customer goes through, from recognising a need to searching, buying, using and buying again. A journey map shows where customers drop out or struggle, and which touchpoints carry the most weight. Research usually combines interviews with survey and behavioural data.

Customer lifetime value (CLV)Metrics

Customer lifetime value is the expected total margin from a customer over the whole relationship. A simple version is average margin per period × expected number of periods, with expected lifetime roughly 1 ÷ churn rate. It is used to decide how much can sensibly be invested in acquiring a customer.

D

DashboardDigital & AI

A dashboard is an interactive display of key metrics from a survey or tracker, refreshed with each new wave and filterable by segment and period. A good dashboard shows the base behind every figure and flags significant changes, so that random fluctuation is not mistaken for a trend.

Data cleaningQuality & law

Data cleaning is checking and removing or correcting poor-quality responses before analysis: speeders, straightliners, duplicates, logical contradictions and nonsense open answers. Rules should be set and documented in advance, so that data are not "cleaned" towards a preferred result.

Decision-maker (DMU)Strategy & analysis

In B2B, the decision-making unit (DMU) is the group of people who shape a single purchase: users, technical experts, procurement, finance and the final approver. A single "decision-maker" is often a simplification. B2B samples should cover the different roles, because their selection criteria frequently differ.

Descriptive researchStudy types

Descriptive research measures who, what, how many and how often – the share of buyers, purchase frequency, brand awareness. It is usually a quantitative survey of a representative sample. It reveals associations between variables but cannot prove cause and effect; that requires an experiment.

See also: Causal research / experiment

Desk research (secondary research)Study types

Desk research analyses information that already exists: Eurostat and national statistics, company filings, trade association data, published tenders and academic studies. It is usually the quickest first step and shows which gaps need primary data. Its weakness is fit: definitions, periods and geographies rarely match your question exactly. See how we scope desk research projects.

Discussion guideMethods

A discussion guide is the plan for a qualitative session: topics, main questions and probes, stimulus material and a rough time allocation for each part. Unlike a questionnaire it is not read word for word; the moderator adapts order and phrasing to the conversation while still covering every objective.

Double-barrelled questionQuality & law

A double-barrelled question asks about two things at once – "How satisfied are you with the price and quality?" – so the answer cannot be interpreted if the respondent feels differently about each. The fix is to split it into two questions. A tell-tale sign is "and" in the subject of the question.

E

Estimated contract value (public procurement)Quality & law

Under Article 5 of Directive 2014/24/EU, the estimated value of a contract is based on the total amount payable, net of VAT, including any options and renewals. It determines whether EU thresholds apply and which procedure is used. Splitting a contract to avoid the rules is prohibited. Comparable awards and market consultation help justify the figure.

EthnographyStudy types

Ethnography means observing people in their own environment – at home, in shops, at work – often for hours or days. Its strength is exposing the gap between what people say they do and what they actually do. Samples are small, so the output is understanding and hypotheses rather than measurement.

Executive summaryStrategy & analysis

The executive summary is a short section at the start of a report – usually one or two pages – with the main findings, the answers to the research questions and the recommendations. It must make sense on its own. A strong summary leads with the answer, not the methodology.

Exploratory researchStudy types

Exploratory research is used early, when the problem is not yet well defined: the aim is to understand motives, map the issues and generate hypotheses worth testing. It typically relies on desk research, interviews or focus groups. Its findings guide the next step but should not be generalised to the whole population.

Eye trackingMethods

Eye tracking records where, in what order and for how long people look at an ad, pack, shelf or web page. Output is usually shown as heat maps and time to first fixation. It measures attention, not liking or understanding, so it is normally paired with interview questions about what people noticed.

F

Factor analysisSampling & statistics

Factor analysis reduces many correlated statements to a few underlying dimensions – twenty brand-image items to four factors, say. It is used to build scales, to simplify data before segmentation and to check whether a set of questions measures what it is meant to measure.

Feasibility studyStudy types

A feasibility study assesses whether a project – a new product, site or market – is viable: is there enough demand, how strong is the competition, what do regulation and operations require, and do the financial projections hold up? Lenders and investors often expect one alongside a business plan, usually with downside, base and upside scenarios.

See also: Market size

Field research (primary research)Study types

Field research gathers new data directly from people or organisations through surveys, interviews, group discussions or observation. You need it when published sources cannot answer the question, or answer it for the wrong population. "Field" does not imply face to face: online surveys and telephone interviews count as fieldwork too.

See also: Fieldwork

FieldworkMethods

Fieldwork is the data-collection stage: conducting interviews, running online surveys or recruiting participants for groups. Its duration depends on method, sample size, incidence and response rate. Interviewers and incoming data should be checked during fieldwork, not only afterwards. Estimate interview length and field time with the survey length calculator.

See also: Incidence rate

Focus groupMethods

A focus group is a moderated discussion, usually with six to eight participants from the target group, lasting around two hours and recorded for analysis. It is good for exploring motives, vocabulary and reactions to ideas. Focus groups do not measure shares or produce representative figures. Read more about running focus groups.

See also: Moderator

G

Gabor-Granger methodMethods

The Gabor-Granger method asks respondents whether they would buy at a series of prices and builds a demand curve from the answers. At each price, the revenue index equals price multiplied by the share who would buy; the optimum is the highest index. Stated intent overstates real purchasing. Try it with our price sensitivity calculator.

GDPRQuality & law

The GDPR (Regulation (EU) 2016/679) governs the processing of personal data in the EU. In research it means a lawful basis (often consent), collecting only the data you need, limited retention and respecting respondents' rights of access, rectification and erasure. The UK applies its own version, the UK GDPR, alongside the Data Protection Act 2018.

I

ICC/ESOMAR International CodeQuality & law

The ICC/ESOMAR International Code on Market, Opinion and Social Research and Data Analytics sets the industry's ethical rules: transparency towards participants, data protection, special care with children and vulnerable people, and a clear separation between research and selling. It is issued jointly by the International Chamber of Commerce and ESOMAR.

In-depth interview (IDI)Methods

An in-depth interview is a one-to-one conversation between researcher and participant, typically 45–60 minutes, following a guide of open questions. It suits sensitive topics, complex decisions and B2B respondents who are hard to bring together in a group. See how we run in-depth interviews.

Incidence rateSampling & statistics

The incidence rate is the share of the population – or of people screened – who qualify for the survey. If only 10% qualify, about 3,000 people must be screened to reach 300 completes. Low incidence lengthens fieldwork and is one of the main drivers of project scope. Do not confuse it with response rate.

See also: Response rate

Informed consentQuality & law

Informed consent is a participant's voluntary agreement given after learning the study's purpose, who is running it, how the data will be used, whether audio or video will be recorded and that they can withdraw at any time. Children need parental consent. The ICC/ESOMAR Code requires it, as does the GDPR where consent is the lawful basis.

InsightStrategy & analysis

An insight is a non-obvious understanding of customer motives or behaviour that explains the data and suggests what to do. "Some customers abandon checkout" is a finding; "they abandon it because delivery charges appear only at the final step" is an insight. Insights usually come from combining quantitative and qualitative evidence.

ISO 20252Quality & law

ISO 20252 is the international standard setting requirements for processes and quality management in market, opinion and social research, from the brief and respondent recruitment to data processing and reporting. The current edition dates from 2019. Certification is voluntary; the standard is also a useful framework for internal quality control.

K

KPIMetrics

A KPI (key performance indicator) is a measurable indicator tied to a specific objective, with a target and a time frame – NPS, penetration, conversion rate or unaided awareness, for example. Keep the list short, and give each KPI an owner and a documented method of measurement so that periods can be compared fairly.

L

Leading questionQuality & law

A leading question steers respondents towards an expected answer through its wording – "Don't you agree that…?" or "How satisfied are you with our excellent service?". Such questions inflate agreement. The fix is neutral wording and a balanced scale with equal numbers of positive and negative points.

Likert scaleMethods

A Likert scale measures attitudes through statements that respondents agree or disagree with on a symmetrical scale, usually of five or seven points. The data are ordinal, so as well as means it is good practice to report the full distribution or the top-two-box share. Strictly speaking, a single statement is a Likert-type item.

M

Margin of errorSampling & statistics

The margin of error is half the width of a confidence interval. At 95% confidence, e = 1.96 × √(p(1 − p)/n); for n = 1,000 and p = 0.5 it is about ±3.1 percentage points. It covers random sampling error only, not errors from question wording or coverage. Calculate it with the margin of error calculator.

Market researchStudy types

Market research is the systematic collection and analysis of information about a market: its size, demand, customers, competitors, prices and sales channels. It supports decisions such as entering a country, launching a product or reallocating marketing budget. A typical scope and table of contents are shown on our market research report page.

See also: Marketing research

Market segmentationStrategy & analysis

Market segmentation divides a market into groups of customers with similar needs or behaviour who respond differently to marketing. Criteria can be demographic, geographic, behavioural, attitudinal or needs-based. A useful segment is measurable, large enough, reachable and clearly distinct. See how a segmentation study is designed.

Market shareMetrics

Market share is a company's or brand's sales as a percentage of total market sales for the same period. Value and volume shares differ when brands sell at different price levels. Agree the market boundaries before comparing: a share of "coffee" and a share of "coffee pods" are different numbers.

Market sizeMetrics

Market size is total sales in a category for a defined geography and period, in value (revenue) or volume (units, litres, tonnes). The answer depends on the definition: which products, channels and customer types (B2B, B2C) are included. Approaches are compared in our market sizing report.

Marketing mix (4P / 7P)Strategy & analysis

The marketing mix is the set of decisions a company uses to influence demand: product, price, place (distribution) and promotion – the 4Ps. For services, people, process and physical evidence are added to make 7Ps. Variants with 5 or 8 Ps exist, but the underlying idea is the same.

Marketing researchStudy types

Marketing research covers every marketing decision – product, price, promotion and distribution – rather than the market alone. Many people use the two terms interchangeably. Where a distinction is drawn, market research describes the environment, while marketing research also evaluates the company's own actions, such as whether a campaign changed how the brand is perceived.

See also: Marketing mix (4P / 7P)

MaxDiffMethods

MaxDiff (best–worst scaling) shows respondents a series of small sets of items – features, claims, benefits – and asks them to pick the most and least important in each set. The result is a ranking with clear distances between items. It avoids the weakness of rating scales, on which nearly everything scores as important.

MedianSampling & statistics

The median is the middle value when observations are ranked: half lie below it, half above. Because extreme values do not pull it, it is a better summary than the mean for income, spending and completion times. A large gap between mean and median signals a skewed distribution.

ModeratorMethods

The moderator leads a focus group or interview using the discussion guide, probes for detail and makes sure every participant is heard without dominant voices taking over. A good moderator stays neutral – never defends the product or hints at the "right" answer – and departs from the guide when an important theme emerges.

Mystery shoppingStudy types

Mystery shopping uses trained assessors who pose as ordinary customers and score the service against a checklist: greeting, product knowledge, waiting time, compliance with procedures. It measures whether standards are delivered, not how satisfied real customers are. Scenarios and scoring are explained on our mystery shopping page.

N

NetnographyDigital & AI

Netnography adapts ethnography to online communities: the researcher observes and analyses how people discuss a category, problem or brand in forums, groups and comment threads. It calls for ethical judgement – public discussions differ from closed groups, where permission from members or moderators is needed.

Non-response biasSampling & statistics

Non-response bias arises when people who do not take part differ systematically from those who do on the topics measured – dissatisfied customers, for example, may ignore a satisfaction survey. It is reduced with reminders, multiple contact modes, weighting and by comparing early with late responders.

NPS (Net Promoter Score)Metrics

Net Promoter Score comes from the question "How likely are you to recommend … to a friend or colleague?" on a 0–10 scale. Promoters score 9–10, detractors 0–6; NPS = % promoters − % detractors, ranging from −100 to +100. Compare scores only within the same sector and method. Calculate yours with the NPS calculator.

O

Omnibus surveyStudy types

An omnibus is a regular, nationally representative survey in which several clients place their own questions in a shared questionnaire. It suits a handful of closed questions that need a quick, representative answer – prompted awareness of a new brand, say. It is unsuitable for long question sets, complex routing or narrow target groups, because space and interview time per client are limited.

Online panelMethods

An online panel is a database of pre-recruited people who have agreed to take part in surveys and whose profiles – age, region, purchases – are already known. It enables fast fieldwork and precise quotas. Panels are not probability samples, so quality depends on how members are recruited, verified and de-duplicated.

Open-ended questionMethods

An open-ended question lets respondents answer in their own words, without preset options. It captures customers' vocabulary and unexpected themes but requires coding and analysis. In online surveys, a few well-placed open questions are more useful than many that people answer with a single word.

See also: Coding of open answers

P

p-valueSampling & statistics

The p-value is the probability of obtaining a result at least as extreme as the one observed if there were in fact no difference (the null hypothesis). It is not the probability that the hypothesis is true. The usual threshold is 0.05; across dozens of comparisons in one table, some "significant" results will be chance.

Packaging testStudy types

A packaging test checks whether a pack stands out on the shelf, how quickly shoppers find it, what it communicates and whether they like it. It is often run on a virtual shelf among competing products, sometimes with eye tracking. Liking alone is a weak criterion if the pack disappears among its competitors.

Penetration rateMetrics

Penetration is the share of a target population – households, consumers, companies – that bought or used a category or brand at least once in a given period. Sales can be broken down into penetration × purchase frequency × average spend per purchase, which shows where growth is coming from.

Perceptual mapStrategy & analysis

A perceptual map plots brands on two dimensions according to how customers see them – "traditional–modern" and "mainstream–premium", for example. It is built from image ratings using factor or correspondence analysis. It shows which brands are seen as close substitutes and where unoccupied positions may exist.

PESTEL analysisStrategy & analysis

A PESTEL analysis reviews the macro environment across six groups of factors: political, economic, social, technological, environmental and legal. It is mainly used for market entry and long-range planning. Factors should be rated by impact and likelihood rather than simply listed.

See also: SWOT analysis

Pilot test (soft launch)Methods

A pilot test, or soft launch, releases the survey to a small part of the sample before full fieldwork to check routing, timing and whether questions are understood. Pilot data are reviewed by hand. If questions change as a result, pilot interviews are usually excluded from the final dataset.

Population (universe)Sampling & statistics

The population, or universe, is the whole group you want to draw conclusions about – for example, "UK adults who bought a car in the past two years" or "manufacturers with 10 or more employees in Germany". Define it precisely before sampling: who is in, who is out and as of which date.

Porter's five forcesStrategy & analysis

Porter's five forces model assesses how attractive an industry is through five pressures: rivalry among existing firms, the threat of new entrants, the threat of substitutes, buyer power and supplier power. It is a staple of industry analysis and helps explain why profitability differs from one sector to another.

PositioningStrategy & analysis

Positioning is the place a brand aims to occupy in target customers' minds relative to competitors: who it is for, which need it meets and why it is different. It is checked with brand-image research and perceptual maps; if customers cannot play back the intended position, it exists only in the strategy deck.

Preliminary market consultation (public procurement)Quality & law

A preliminary market consultation lets a contracting authority, before launching a procurement procedure, seek advice from independent experts, authorities or market participants to plan the procedure and prepare the documents, as Article 40 of Directive 2014/24/EU allows. Article 41 requires measures so that participants gain no unfair advantage. See our public procurement market consultation report.

See also: Estimated contract value (public procurement)

Price elasticity of demandMetrics

Price elasticity of demand is the percentage change in quantity demanded divided by the percentage change in price. With an elasticity of −2, a 10% price rise cuts volume by about 20%. When the absolute value exceeds 1, demand is elastic and a price cut raises revenue; below 1, the reverse holds.

Primary dataStudy types

Primary data is collected specifically for the study in hand, so the definitions, target group and questions match the decision you need to make. The trade-off is time: designing a questionnaire, running fieldwork and processing results take weeks, whereas secondary data is often available the same day.

See also: Secondary data

Purchase intentMetrics

Purchase intent is a respondent's stated likelihood of buying, usually on a five-point scale from "definitely would buy" to "definitely would not buy". It is reported as the top-box or top-two-box share. Stated intent overstates actual purchasing, so results are calibrated or compared with norms from earlier tests.

Q

Qualitative researchMethods

Qualitative research works with a small number of participants and open questions – focus groups, in-depth interviews, ethnography – to understand motives, language and perceptions. It answers "why" and "how". Results should not be reported as percentages, because the sample is not designed to be representative; sizing needs a quantitative stage.

Quantitative researchMethods

Quantitative research collects structured answers from a large sample and reports results as numbers: percentages, means, indices. It answers "how many" and "how often" and, with a suitable sample, supports conclusions about the population within a margin of error. It is weaker at explaining the reasons behind the answers.

See also: Qualitative research

QuestionnaireMethods

A questionnaire is a structured set of questions with fixed wording, order, routing and answer options, administered identically to every respondent. Length affects drop-out and the quality of answers towards the end, so every question should earn its place. Try the draft on a few people before fieldwork starts.

See also: Closed question, Open-ended question

Quota samplingSampling & statistics

Quota sampling recruits respondents until preset quotas are filled – for example by gender, age and region in line with official statistics. It is standard in commercial online and telephone surveys. Because it is not a probability method, the margin of error is only an approximation rather than a strict calculation.

R

Random (probability) samplingSampling & statistics

In random sampling every unit in the population has a known, non-zero chance of selection. Only then can sampling error be calculated rigorously. Complete lists of a population are rare in practice, so commercial research often uses quota samples and applies the same formulas as an approximation.

Regression analysisSampling & statistics

Regression analysis models how a dependent variable – satisfaction, purchase intent, sales – changes with one or more independent variables. In research it is the usual engine of key driver analysis. Coefficients show association rather than proof of cause, and strongly correlated predictors (multicollinearity) make the results unstable.

Research briefStrategy & analysis

A research brief is the client's document describing the business problem, the decision to be made, the target group, the timing and the data already available. A good brief names the decision ("should we launch in Poland?") rather than just the topic ("we want research on Poland"). Unsure where to start? Try the report finder.

Research hypothesisStrategy & analysis

A research hypothesis is a testable statement formulated before the study – for example, "customers over 55 prefer to order by phone". Hypotheses determine what the questionnaire must cover and how the data will be analysed. In statistical testing, they are assessed against a null hypothesis of no difference.

Research proposalStrategy & analysis

A research proposal is the researcher's response to the brief: objectives, method, sample, research questions, timeline, deliverables and terms. A good one explains why the method was chosen and what its limitations are. You can request a proposal for your own project.

Response rateSampling & statistics

The response rate is the share of invited or contacted eligible people who complete the survey. Definitions vary by method – some count partial completes, some exclude unreachable numbers – so always state which one you used. A low response rate does not by itself mean bias, but it raises the risk.

See also: Non-response bias

Retail auditStudy types

A retail audit checks stores for product availability, shelf share, shelf prices, out-of-stocks and point-of-sale material. Results are reported as numeric distribution (the share of stores stocking the product) and weighted distribution (the same, weighted by each store's category sales). The method is set out in our retail audit report.

S

SAM (serviceable available market)Metrics

SAM is the part of TAM you can actually serve with your product, channels, languages and geography. A UK online retailer that does not ship abroad, for example, excludes overseas demand even though it sits inside TAM. SAM is a more useful planning figure than TAM.

SampleSampling & statistics

A sample is the part of the population that is actually surveyed. How it is selected matters more than its size: a large but skewed sample gives a precise estimate of the wrong thing. A representative sample mirrors the population on the characteristics that matter for the topic.

See also: Sample size, Sampling frame

Sample sizeSampling & statistics

Sample size is the number of completed interviews. For a large population, a 95% confidence level and a ±5-point margin of error at p = 0.5, you need about 385 respondents (1.96² × 0.25 / 0.05²). Every subgroup you plan to compare needs a sufficient base of its own. Work it out with the sample size calculator.

Sampling frameSampling & statistics

The sampling frame is the list or source from which respondents are drawn – a business register, customer database, online panel or telephone list. Gaps between frame and population create coverage error: a customer database, for instance, misses everyone who buys only from competitors.

ScreenerMethods

A screener is a short questionnaire used to select participants against set criteria: category use, age, role in the purchase decision. It normally excludes people working in marketing, advertising or for a competitor, and those who have taken part in similar research recently. A weak screener undermines even a well-moderated group.

Search demand analysis (search volume)Digital & AI

Search demand analysis measures how often people search for specific phrases and how that changes by month and region, using Google Trends and keyword tools. It indicates interest and seasonality, not purchases, and tool volumes are estimates. It is a core part of a digital market analysis.

Secondary dataStudy types

Secondary data was collected earlier by someone else for a different purpose – a statistics office, a trade body, a competitor's annual report. Before relying on it, check the definition, reference period, method and coverage. A "small business" means fewer than 50 employees in EU statistics, yet a trade report may define it by turnover instead.

Semantic differentialMethods

A semantic differential is a rating scale between two opposite adjectives – "modern–old-fashioned", "warm–cold" – usually with seven points. It is used to build an image profile of a brand or product and compare it with competitors on the same pairs. The pairs must be genuine opposites, or the answers become hard to interpret.

Sentiment analysisDigital & AI

Sentiment analysis automatically classifies text as positive, negative or neutral. Sarcasm, slang, mixed languages and sector-specific vocabulary reduce accuracy, so the model should be validated against a sample of manually coded texts before results are reported. Accuracy figures from a vendor's demo data rarely carry over to your own.

Share of voiceMetrics

Share of voice is a brand's share of total advertising presence in its category, measured by media spend, impressions or online mentions. Comparing it with market share – excess share of voice – shows whether a brand is investing above or below its current position. Use the same source and definition for every brand.

Share of walletMetrics

Share of wallet is the proportion of a customer's spending in a category that goes to your company. A customer who spends only a small part of their category budget with you represents growth that does not require winning new customers. It is measured by survey or from loyalty-programme data.

Social desirability biasQuality & law

Social desirability bias is respondents' tendency to answer in ways that make them look good: overstating recycling, exercise or reading, and understating alcohol use or debt. It is reduced with self-completion modes such as CAWI, reassurance about anonymity and indirect question wording.

Social listeningDigital & AI

Social listening is monitoring and analysing public mentions of a brand, competitors or topics across social networks, forums, news sites and review platforms. It shows themes, tone and change over time. It does not represent the population – only people who post publicly, who are usually the most engaged.

SOM (serviceable obtainable market)Metrics

SOM is the part of SAM you can realistically win within the planning period, given competition, budget and capacity. Investors scrutinise the assumptions behind it most closely: target share, pace of customer acquisition and retention. A SOM that is simply a round percentage of SAM rarely survives due diligence.

See also: TAM (total addressable market), SAM (serviceable available market)

SpeederQuality & law

A speeder is a respondent who completes a survey too quickly to have read the questions. The threshold is set relative to the median completion time – for instance, under a third of it – and agreed before fieldwork. Removing speeders is a standard step in online survey quality control.

Standard deviationSampling & statistics

Standard deviation measures how widely values spread around the mean, in the same units as the data. In a normal distribution about 68% of values lie within one standard deviation of the mean and about 95% within 1.96. Do not confuse it with the standard error, which is the standard deviation divided by √n.

Statistical significanceSampling & statistics

A difference is statistically significant when it is unlikely to be explained by sampling chance alone – conventionally when the p-value is below 0.05. Significance is not importance: with very large samples a one-point difference can be significant without changing any decision, while small samples can miss differences that matter.

See also: p-value

STP (segmentation, targeting, positioning)Strategy & analysis

STP is the sequence segmentation → targeting (choosing which segments to serve) → positioning the offer for them. The order matters: positioning without a chosen segment usually ends in a message that tries to appeal to everyone and persuades no one in particular.

StraightliningQuality & law

Straightlining is when a respondent gives the same answer to every statement in a grid regardless of content. It usually signals low engagement. Sometimes it is legitimate – the person genuinely holds the same view throughout – so it is assessed alongside other indicators rather than used as the only reason to remove a case.

Stratified samplingSampling & statistics

Stratified sampling divides the population into groups (strata) – by region, company size or customer type – and draws a random sample within each. It improves precision and secures enough interviews in small strata. If strata are deliberately over- or under-sampled, results are weighted back to the true population structure.

SWOT analysisStrategy & analysis

A SWOT analysis sets out strengths and weaknesses (internal) and opportunities and threats (external) for a company or product. It is useful when every point is backed by evidence and leads to an action; a list of generic statements rarely helps. It is often paired with a PESTEL review of the wider environment.

See also: PESTEL analysis

Syndicated research (off-the-shelf report)Study types

Syndicated research is carried out on the research firm's own initiative and sold to many clients, who all receive the same data. It works well for a broad view of a market and for benchmarking against published figures. You cannot change its definitions or methodology, and it will not answer questions specific to your business.

See also: Custom (ad hoc) research

Synthetic respondents (AI)Digital & AI

Synthetic respondents are answers generated by language models that simulate people with a given profile. They can help draft hypotheses or dry-run a questionnaire. They do not replace real data: they reflect training text rather than the current market, and no margin of error can be calculated for them.

T

TAM (total addressable market)Metrics

TAM is total demand for a product if every potential customer in the chosen geography bought from you. It is estimated top-down from statistics or bottom-up: number of potential customers × purchases per year × average purchase value. Estimate TAM, SAM and SOM with our market size calculator.

Target groupStrategy & analysis

A target group is the people or organisations a product, campaign or study is aimed at, described with measurable criteria: age, region, category use, role in the purchase. The narrower the target group, the lower the incidence and the harder the recruitment of respondents.

Text analyticsDigital & AI

Text analytics extracts themes, entities, frequencies and relationships from large volumes of text – open survey answers, reviews, support tickets, call transcripts. It speeds up coding and surfaces topics nobody anticipated. The conclusions still require human interpretation and a check of context.

Topline reportStrategy & analysis

A topline report is a quick overview of the main results on the total sample, produced soon after fieldwork closes and before full analysis. It supports decisions that cannot wait. Figures are provisional and may shift slightly once the data have been cleaned and weighted.

Tracking studyStudy types

A tracking study repeats the same measurement at regular intervals – in waves or continuously – with an unchanged questionnaire and method, so that change over time can be read with confidence. Altering a question's wording, the answer scale or the data-collection mode can break the trend, so any change needs a parallel run of old and new versions.

See also: Brand tracking

U

Usability testStudy types

A usability test observes real users as they attempt specific tasks on a website, app or device: do they succeed, how long does it take, where do they go wrong? Short rounds of around five participants, repeated after each fix, are usually enough to find the main problems; benchmarking success rates needs larger samples.

Usage & attitudes (U&A) studyStudy types

A usage and attitudes study describes how consumers buy and use a category: frequency, channels, brands, occasions, motivations, barriers and attitudes. It is usually a large quantitative survey and serves as the foundation for segmentation and positioning work. Our consumer research page shows a typical scope.

USP (unique selling proposition)Strategy & analysis

A USP is a single, specific advantage that sets a product apart from competitors and matters to customers – the shortest delivery time in the category, for example. Test it with two questions: is it true against the real alternatives, and does the buyer actually care?

V

Value propositionStrategy & analysis

A value proposition describes the benefits a specific segment gets from a product, the problems it solves and why it is a better choice than the alternatives. It is broader than a USP, covering the whole benefit rather than one differentiator. It is tested through interviews and concept tests.

Van Westendorp Price Sensitivity MeterMethods

The Van Westendorp Price Sensitivity Meter asks four questions: at what price would the product be so cheap you would doubt its quality, a bargain, getting expensive, and too expensive to consider? The intersections of the cumulative curves define a range of acceptable prices. It does not measure purchase intent, so it cannot forecast volume. See our pricing research options.

W

Web analyticsDigital & AI

Web analytics collects and analyses visitor behaviour on a website: traffic sources, pages, events and conversions, for instance in Google Analytics 4. It shows what visitors do but not why. Visitors who decline analytics cookies are not tracked, so reported figures understate real traffic.

WeightingSampling & statistics

Weighting adjusts the influence of each response so that the sample's profile matches the population – by gender, age, region or company size. Large weights reduce the effective sample size and widen the margin of error. Weighting cannot compensate for a group that is missing from the data altogether.

Willingness to payMetrics

Willingness to pay (WTP) is the highest price a customer would pay for a product or feature. It is measured directly (Van Westendorp), indirectly through choices between offers (conjoint, Gabor-Granger) or through live price tests. Direct price questions are more prone to strategic, low-balled answers.