Online survey samples

Survey participants are called a “sample” because they represent the population under study, which is rarely consulted as a whole.

Several methods can be used to build a sample

  • Random selection from a database
  • Random selection from a database with application of simple or crossed quotas
    – Representative sample of the target population
    – Judgmental sampling; example: 300 individuals for each purchasing type; in other words, 200 individuals per unit.
    For testing visual concepts, there are basically three approaches:
  • Pure monadic: you will show each test visual to 300 individuals. If you have five visuals, you will have a total sample of 300 x 5 = 1,500 individuals. The advantage is that each person sees only one and is not influenced by the others.
  • Sequential monadic: you will show the five visuals to 300 individuals who will rank them at the end of the questionnaire. Advantages: 300 people are enough, and you will have a ranking at the end without having to make calculations.
  • Partial sequential monadic: you will show two out of five concepts to each respondent in order to limit the number of visuals shown to each respondent.
Read more
Online survey samples

Number of answers: sample size

The number of responses is tied to a notion of statistical reliability that refers to the calculation of confidence intervals (statistical law).

To simplify, the greater the size of your sample, the greater the reliability of answers. However, going beyond certain size thresholds will not bring more information.

B2C – consumer surveys

  • Representative field, France: traditionally carried out on 1,000 individuals with socio-demographic quotas related to gender, age, profession and region.
  • Product purchasers: 600 or 900 responses to be able to interpret results for a subpopulation. Minimum subpopulation size: 100 responses.

    B2B – employee and professional surveys

  • Perception of a trend by employees: 300 individuals to be able to interpret to a lower level with at least 100 responses per profile.
  • Immediate evaluation of a training course: you must have at least three graded sessions in order to validate the grade given to the trainer (around 50 responses).
  • Later evaluation of a training course: this is a quiz, and 50 responses are enough to ensure the validation of what has been acquired by homogeneous groups. If you want a perception by trainees of the application of new knowledge obtained in a training course held one or two months previously, you will need at least 100 responses.
  • Tip

    Be careful to avoid mixing sample terms. It may be a question of the contact
    base used at the start or of the results base. The structure, in particular
    the socio-demographic structure, will be different.

    Precision of customer satisfaction studies

    Basic rule: Precision is inversely proportional to four times the sample size.

    For a twofold increase in precision, the sample must be multiplied by four!

    Thus, if 20% of your customers are dissatisfied on a particular question, a sample of 100 individuals will guarantee a confidence interval of 8 points. If this eight-point interval seems insufficient, you should multiply the sample by four (400 individuals) to reduce this interval by half (to 4 points).

    Each reduction in the confidence interval will follow the same rule, as shown in the example below calculated for a 20% dissatisfaction rate:



    In practice, you will want to optimize your budget and limit the sample size to the strict minimum.

    Precision and satisfaction

    Experience has shown that in a customer satisfaction study, opinions are widely scattered and the rate of dissatisfied customers is relatively low (<20%). This distribution enables limiting sample size to the strict minimum. In addition, if 50% of your customers were shown to be dissatisfied about a given point, you should worry less about whether the "true" response rate is 45% or 55% and more about urgently mobilizing to save the brand.

    Precision and budget

    It is also important to ask yourself how precision operates. If a study of 200 customers gives you five obvious priority actions to carry out, would a study of 800 customers, for four times the budget, change your action plan?
    If you work from your database or website, the response rate will be directly impacted by the following elements:

  • Questionnaire length
  • Clumsiness: repeated matrices and questions
  • Graphic personalization of the invitation email and questionnaire
  • Presence or absence of compensation: drawing, points account, charitable contribution

    Also, if your questionnaire abandonment rate is too high, this means that respondents are put off by certain elements in your questionnaire.

  • Variation of response rates by target

    Not surprisingly, it is more difficult to get a response from a CEO than from an administrative employee. This will have an impact on the basic structure of the response rate. Therefore, it is important to anticipate this natural deformation by over-representing individuals who will respond less frequently in the starting base.

    B2C studies

  • Over-represent seniors and young people, farmers and workers.

    B2B studies

  • Over-represent managers, IT workers and service personnel.
  • graphe


    When you build your starting base, “deform” it according to the response rates
    that are normally observed or supposed to occur. In this way,
    you will have a homogeneous result base without having to adjust it.



    Discover the solution to make the right decisions


    Innovative solutions
    for successful projects



    Take two minutes to discover our tools and think up a project



    And discover the difference made by our collecting and reporting tools

    2001-2018 © AreYouNet - An ODiTY PLATFORM. All rights reserved.
    Design & developed by ODiTY
    webcall back
    We call you back