The survey has been defined as a systematic search for information, generally based on probability sampling, to question a group of people about research questions and variables.

The survey is presented as a tool that will allow statements to be made about sets, not about individuals in particular. It serves as an "approximation to the knowledge of a social or collective fact", not to decide whether something is true or false. The extensive nature of the survey is intended not to delve into peculiarities, but to privilege the uniformity that will account for a particular trend. The objectivist criterion predominates.

The questionnaires are the result of a process of operationalization of concepts that operates with precision.

This technique has benefits and some weaknesses, which we point out in this table:

Source: Own elaboration

Basically, there are two types of survey: the structured survey, which is based on a questionnaire with closed questions, and in which participants are asked to answer by choosing from the options offered, and the unstructured survey, which poses questions that can be answered openly. The latter poses more problems than solutions, since the questions are much easier to prepare than in the structured survey, but do not allow the coding involved in the closed questions that we will see below.

There are other types of surveys, which are shown in this table:

In this table, as we can see, we distinguish personalized or collective surveys, although it can be said that in reality the first typology responds not so much to surveys of a single person as to surveys adapted to different types of informants, by segmenting or stratifying the universe to carry out a sampling system of this type. For example, the survey questionnaire will be adapted to men and women, if that is a criterion for stratification, or it will be done by age segments, or even according to some questions, posed as a gateway or filter to others; if the answer to those questions is negative, a subsequent block of related questions is skipped.

The survey design goes through the following phases:

The technique for designing the questions will follow these steps: first, there is a series of registration questions, in which personal data of the respondent is collected: age, profession, whatever is considered necessary. The questions asked thereafter must be necessary, without being superfluous or redundant. Respondents' time - and their patience - are limited, so it is best that the time a survey lasts no longer than half an hour. Questions should be short and unambiguous; only one question is asked at a time. They should be clear, so that there is usually only one answer option.

The questions should be discrete, usually ordered from the general to the particular and in a hierarchical order, because sometimes one depends on the other. Finally, the questions should be formulated in a logical and intelligible manner. The wording of the questionnaire, which is a key aspect for the success of the research, should be clear, without ambiguities or technicalities, and should be perfectly understandable to the respondents (scientific language and jargon are not applicable unless the survey is intended for researchers or experts in the field). 

Filter questions discriminate individuals, either to choose who will respond to the survey, or to create "pathways" within the survey. For example, this question is a filter question ("Information consumption. Do you do it through social networks?"). The next question can only be answered if you have answered yes to this question ("If yes, please state the social networks you use for information").

The control questions have the function of checking the veracity of the answers. They are usually non-consecutive and refer to objectifiable facts, habits and customs, while consistency questions are used to check the respondent's coherence, inquiring whether he/she does not present major contradictions in terms of opinions and beliefs. Nor are they usually consecutive, for obvious reasons.

The formulation of the questions will be done formally using one of the following modalities:

Dichotomous questions are binary and exclusive, allowing only one or another answer, or sometimes three (positive, negative, neutral or don't know/no answer). Multiple-choice questions allow a choice between several answers, either exclusive (only one answer is allowed) or combined (more than one answer is possible). Others ask for a graded response (a Likert scale or a semantic differential scale). Other types of questions are those in which an attempt is made to force an answer ("If you had to choose one of these options, which one would you choose?") and, finally, there are open questions.

Since it is not always the person who designs the survey who carries it out, it is necessary, in addition to the questionnaire, to draw up instructions for its completion. The questionnaire will begin with a brief introduction explaining the subject of the research, the instructions for filling it in and thanking the participants and guaranteeing confidentiality. It should be borne in mind that very long questionnaires can be more easily abandoned. 

Next, the usefulness of the questions is tested and the questionnaire is adjusted. A pilot test is launched with a small number of informants, and if the test is passed satisfactorily, the survey is carried out in its entirety.

The answers have to be coded and tabulated, so they must be assigned values or weights, because they will then be processed with quantitative computer programs, such as SPSS, which only work with this type of numerical data. The answers sought should be comparable, operational. The variables must be defined conceptually and operationally. It is not always necessary to define them (age, sex), but many require a conceptual and operational definition to be able to work with them (stereotypes, values). The variables must be measurable, exhaustive and exclusive.

In turn, the operationalization process consists of specifying the research problems in precise questions (dimensions), following strict criteria of objectivity, proportionality and precision. This and the work with variables are the basic elements of the survey.

Operationalization is an empirical translation: that is, it consists of making the theoretical constructs that represent the concepts observable. Communication research usually works with concepts that are not easily observable directly, and which require the researcher to make them observable, a procedure that is carried out by means of indicators.

Theoretical concepts, when observed and measured, are broken down into their empirical dimensions and indicators. We must ask about disaggregated indicators (training, level of studies, income, type of work, income, parents' level of education, parents' income) that allow us to construct the "index". The indicators must be quantitative (or scalar ordinal) in order to construct the index. In other words, each of the indicators is quantified and the resulting measure of the index is obtained; the social status of a person or group of people is known quantitatively.

Finally, the scales make it possible to establish differences of degree between individuals with respect to a value or a variable, which also make it possible to detail the results of the survey in greater detail. An intermediate point (which can be considered neutral or indifferent) can be obtained by means of the odd scales. The even scales do not allow this, so the respondent must express himself in one way or another.  The technique of scales makes it possible to observe the attitudes of the respondents. There are three main scales in social science practice and communication research: the Likert, Thurstone and Guttman scales.

Likert scale: Likert created it in 1930. Renis Likert simplified Louis Leon Thurstone's scale (early 20th century). It is an ordinal measurement system. The concept to be studied is broken down by means of a series of statements to which the respondent must agree or disagree, by means of four or five options. All possible dimensions of each concept should be included. Each answer deserves a score that will serve to measure the attitude of the respondents.

Semantic differential scale

Pairs of opposing adjectives are used to measure attitudes in relation to an object or fact.

Guttman scalogram:

The scalogram method is used to measure subjects, stimuli and attitudes.  Responses are scalable, so that if the respondent accepts the items above the one he/she has chosen. (e.g., income level, or level of education).