Market and opinion research glossary

Comparing two versions of a page, ad or message with comparable audiences to see which performs better.

Comparing metrics against external references, such as competitors, industry averages or best practices. It shows whether a result is good or bad relative to the market.

Example Comparing a company’s NPS with that of other companies in the same industry.

The socioeconomic classification standard set by ABEP (Critério Brasil). It divides the population into classes A, B1, B2, C1, C2 and DE based on household goods, the head of household’s education and access to public services.

Research conducted after a campaign runs, to measure recall, comprehension and impact, often by comparing people who saw it with people who didn’t.

Evaluating an ad or campaign before it runs to check comprehension, messaging, appeal and potential misreadings, while there’s still time to adjust.

A face-to-face interview with the questionnaire on a tablet. It logs the time and length of each interview, and data comes in consolidated, with no manual data entry.

A phone interview conducted by a trained interviewer using an electronic questionnaire, with supervision and call monitoring. Reaches different regions quickly, with quota control.

Example Opinion and satisfaction surveys with audiences spread across the country.

A questionnaire that participants complete online themselves, sent to a customer base or recruited audiences. Supports images and video, at lower cost and with faster turnaround.

Measures how much effort customers had to put in to resolve an issue. The less effort, the more likely they are to stay.

Example “How easy was it to resolve your request today?”, right after a service interaction.

Sorting open-ended answers into categories so they can be counted and compared like closed-ended answers.

Example Turning hundreds of free-text reasons into 10 categories, such as price, service and delivery time.

Evaluating an idea for a product, service or campaign before developing it, to measure interest, comprehension and purchase intent.

The range where the true value in the population most likely falls: the sample result plus or minus the margin of error.

Example 68% with a ±2.8-point margin: the true value most likely falls between 65.2% and 70.8%.

How sure we can be that the margin of error holds. At 95%, if the survey were repeated 100 times, the result would fall within the margin in 95 of them.

Satisfaction with a specific product, service or interaction, usually on a 1 to 5 scale. The score is usually the percentage of 4 and 5 ratings.

Example “How satisfied are you with the delivery of your order?”

Identifies which attributes have the most impact on an overall score, such as satisfaction or likelihood to recommend. Combined with each attribute’s performance, it shows where to start improving.

Example Finding out that delivery time affects satisfaction more than price does.

Spending time in people’s own environment, at home, at work or at the point of sale, to understand habits and contexts they may not be aware of or able to describe.

Verification by an independent body that a company meets the requirements of a standard. It means quality doesn’t rest on the research firm’s word alone.

Example Checon is certified under ISO 20252, with external audits.

Records where a person looks, for how long and in what order, on a screen, package, shelf or ad. Produces heatmaps and gaze paths.

Example Finding out that nobody notices a page’s main button.

A personal interview in homes, high-traffic locations, events or service points, with or without a tablet. Reaches people who aren’t reachable by phone or online and lets you show products and materials.

Identifies basic emotions from facial expressions, moment by moment, as people view a video, ad or website.

The percentage of positive answers on a scale, such as “strongly agree” plus “agree.” It sums up the scale in a single number that’s easy to track.

Example 36% strongly agree and 32% agree: 68% favorability.

A group discussion, usually with 6 to 10 participants, led by a moderator, that brings out shared perceptions and disagreements. It can be held in person or online, with live streaming.

Example Testing concepts and messaging, and exploring new topics.

Measures skin conductance, which varies with the intensity of emotional arousal, second by second. It shows how strong a reaction is, not whether it’s positive or negative.

Measures how quickly people link a brand to specific attributes, revealing automatic associations they may not need to, or be able to, put into words.

A chart that maps how much each attribute matters against how well it’s rated. It shows what to fix first, what to maintain and where investment won’t pay off.

A one-on-one conversation, guided by a discussion guide, that explores experiences, motivations and decisions without group influence. Well suited to sensitive topics and hard-to-reach audiences.

Example Hearing from managers about how they choose a supplier.

A finding that explains a behavior and points to an action. It goes beyond the data to say why something happens and what to do about it.

Example Customers weren’t leaving over price, but over slow service.

The international standard for market, opinion and social research. It sets requirements for every stage, from project design and fieldwork to quality checks, analysis and reporting.

An automated phone interview in which participants answer by voice or keypad. Suited to high volumes and short questionnaires.

Example Transactional NPS right after a service interaction.

Law 13,709/2018. In research, it ensures participants know how their data will be used and consent to that use, and that their answers are used for research purposes only.

An agreement scale, usually with five points, from “strongly disagree” to “strongly agree,” with a neutral point in the middle.

Example “The company delivers what it promises”: strongly agree, agree, neither agree nor disagree, disagree, strongly disagree.

How far the sample result may be from the true value in the population, in either direction, at a given confidence level.

Example 1,225 interviews at 95% confidence give a margin of ±2.8 percentage points.

The professional who leads focus groups and interviews: follows the guide, keeps the conversation going, probes answers and makes sure everyone is heard.

A trained evaluator who poses as a regular customer and uses a standardized script to record how the service went, in person, by phone or through digital channels.

Example Checking whether every location in a chain meets the service standard.

Errors that have nothing to do with sample size: badly worded questions, refusals, poorly trained interviewers, data entry or processing mistakes. They’re kept in check through quality control.

Example A leading question gives a skewed result even with 10,000 interviews.

Measures loyalty with one question: on a scale of 0 to 10, how likely is the customer to recommend the company to a friend or colleague? The score is the percentage of promoters (9 and 10) minus the percentage of detractors (0 to 6), and it ranges from −100 to +100. Reference ranges: critical (−100 to −1), improvement (0 to 49), quality (50 to 74) and excellence (75 to 100).

Example 60% promoters and 15% detractors: an NPS of 45.

We watch how people actually behave where they shop, use a product or receive a service. It shows what really happens, including what participants don’t notice or don’t mention.

A group of participants who engage on a platform over days or weeks, with activities, discussions and tasks set by a moderator. It lets you track opinions over time and co-create solutions.

A database of people who have agreed to be invited to surveys. Since their profiles are already known, recruiting specific audiences is faster.

The full set of people a survey aims to understand: a city’s residents, a company’s customers, an organization’s employees.

The three NPS groups: promoters give 9 or 10, passives give 7 or 8 and detractors give 0 to 6. Passives don’t count in the NPS calculation.

Explores the why: people’s motivations, perceptions and language, through smaller groups and in-depth conversations. It doesn’t measure proportions; it reveals what numbers alone can’t show.

Measures how much: proportions, ratings and differences between groups, using representative samples and a structured questionnaire.

A standard set of questions, asked in the same order and with the same wording for everyone, so answers can be compared and added up.

The sample is built to mirror the makeup of the population on criteria such as sex, age, region and social class. Widely used in market and opinion research.

Example If 52% of the population are women, 52% of the interviews are with women.

A selection method in which everyone in the population has a known chance of being chosen. It’s the theoretical basis for calculating the margin of error.

Example Drawing phone numbers or households from a list that covers the entire population.

Selecting participants who match the profile defined for the study, using a screening questionnaire and checks before the session.

How well a sample mirrors, on a smaller scale, the makeup of the population on the criteria that matter for the study: region, sex, age, social class and so on.

Example Like a well-stirred soup: every spoonful has a bit of every ingredient.

A usability testing technique in which participants first complete the tasks without interruption, then review their own session and explain what they were thinking at each moment.

A group of people chosen to represent the population being studied. When it’s well chosen, you can understand the whole by interviewing only part of it, within a known margin of error.

Example Like tasting soup with a spoon: you don’t need to eat the whole pot to know how it tastes.

The gap between the sample result and the true value in the population that exists only because part of the population was interviewed, not everyone. It’s what the margin of error measures.

Existing information, such as public statistics, industry studies and internal data, gathered and analyzed to answer a question without new fieldwork.

Example Sizing a market with IBGE data before deciding where to open a store.

Dividing the population into groups, called strata, such as regions or city sizes, and drawing the sample within each one. It ensures every important group is represented.

Research repeated in waves, using the same methodology, to track how metrics change over time.

Example Measuring favorability every six months and comparing it with the previous wave.

Remembering a brand or campaign. Unaided, when people name it with no prompting; aided, when they recognize it from a list, image or clip.

Example “Which banks do you know?” (unaided) and “Do you know brand X?” (aided).

Watching real people use a website, app or system to complete tasks, to find out where they get stuck and why.

A statistical adjustment that gives each interview more or less weight so the sample matches the actual makeup of the population, when a group ends up over- or underrepresented.

Example If young people make up 15% of the interviews but 20% of the population, each interview with a young person counts slightly more.