- Sample
- The subset of people actually surveyed. Properly selected, it lets you speak about the whole universe with a known margin of error.
- Universe
- Everyone you want to draw conclusions about. Defining it badly is the costliest mistake in a study, because it invalidates everything that follows.
- Margin of error
- How far the sample result can deviate from the universe’s true value. 42% with ±3 points means the true value almost certainly lies between 39% and 45%.
- Confidence level
- The share of samples like yours that would contain the true value if you repeated the study many times. The industry standard is 95%.
- CAPI
- Computer Assisted Personal Interviewing: a face-to-face survey with the questionnaire on a tablet.
- CATI
- Computer Assisted Telephone Interviewing: a computer-assisted telephone survey.
- CAWI
- Computer Assisted Web Interview: a self-administered online survey, usually on a panel.
- Quota
- The number of interviews to complete in each subgroup — age, sex, city, level — so the sample resembles the universe.
- Weighting
- A statistical adjustment correcting over- or under-representation of a group in the final sample. Always documented: an undisclosed weighting is a manipulation.
- Technical datasheet
- The block accompanying every publishable result: universe, sampling frame, selection method, fieldwork dates, size, margin of error and confidence level. Without it, a number is not data.
- Verbatim
- The respondent’s literal answer, unsummarised. It is what explains why a score is what it is.
- NPS
- Net Promoter Score: promoters minus detractors on the recommendation question. Useful as a time series; of little use as a standalone number without the verbatims behind it.
- U&A
- Usage and attitudes study: who consumes the category, how often, on what occasion and by what criteria they choose.
- Top of mind
- The first brand someone names unaided when asked about a category. The most demanding measure of awareness.
- Spontaneous and aided awareness
- Spontaneous: the brand is named without appearing on a list. Aided: it is recognised when shown on one. The gap between the two says a great deal about the brand work still pending.
- Blind test
- Product evaluation with the brand hidden, to separate what is liked about the product from what the label contributes.
- Saturation
- The point in a qualitative study at which new interviews stop contributing new themes. It is the criterion for deciding how many to run.
- Screener
- The opening block of questions deciding whether a person fits the target profile. A weak screener ruins the entire sample.
- Socioeconomic level
- A classification of the household by its capacity to meet needs and its standard of living. In Mexico the reference rule is AMAI’s.
- Tracking
- Repeated measurement over time with the same questionnaire and sample design, so waves are comparable with each other.
- Social desirability
- The tendency to answer what sounds good rather than what one thinks. It is countered with questionnaire design, not with more sample.
- Statistical significance
- That a difference between two figures is larger than sampling chance would explain. Without it, a two-point gap may be noise.
