Representative sample
For most marketing and social research projects the target audience of interest is too large to attempt surveying all of its members, either in terms of time or costs it would require. A well-designed sample that reflects the main characteristics of this target audience and can represent the whole audience is used in marketing research. It is important to select a representative sample to ensure that the research results can be applicable to a chosen target audience and will produce reliable data. The following analysis describes the main principles of the sample representivity to guide you in choosing the appropriate data collection method. To compare the advantages of telephone and internet survey sampling, a case study drawn from real-life research project is included.

A representative sample according to ESOMAR is the sample that contains units in the same proportion as the population of interest (target group), for example, 55% female / 45% male, in the age of 25-45 years.

The main principles of a representative sample:
1. Random sampling: sampling method when all representatives of the population have an equal opportunity to be selected in the survey sample. Both - traditional and internet surveys suffer from 2 types of limitations using random sampling method:
  • Coverage error - when all members of the population do not have an equal opportunity of being included in the sample, due to a limited access to a communication channel being used in the data collection (for example, home phone landline, internet, etc.).
  • Non-response error – The selected respondents are not available for surveying (for example, not at home or mobile phone is switched off) or refuse to participate in the survey.
2. Sufficient sample size. The sample size determines the probability of a statistical error of the survey results. Statistically speaking, larger sample size ensures higher confidence level and smaller margin of error when applying research results to the population. Most commonly used statistical intervals to determine the sample size is: 95% confidence level and +/-5% margin of error (example: if 80% of the respondents in a survey answered that they use smartphone, then with a 95% confidence we can conclude, that the use of smartphones in the target group is between 75%-85%).

3. Appropriate quotas. The number of respondents in different socio-demographic groups is proportional to the segmentation criteria of the target market.
The target
audience
Has access
to the medium
Has access
to the medium and
volunteers to participate
Reached sample
size and quotas







In the analysis below we have summarized the process of collecting respondents on the customer satisfaction survey for the company X. Two data collection methods have been used: mobile phone interviews and internet survey. For the mobile phone interviews the phone numbers of potential respondents have been selected using the random sampling method from the customer database available in the organization X. For the internet survey the potential respondents have been selected from the SolidData Online panel. The objective for each data collection method was to collect 500 respondents. Taking into account the average response rate, initially the sample of 1500 potential respondents were selected for each survey method. Summary of the sampling for each data collection method:

Sampling process Telephone
interviews
Online
survey
Has access to the medium:
Has access to the medium and volunteers to participate:
(number of respondents N)
100%
26.1%
(from 1500 potential respondents who were contacted to participate in the survey, fully completed surveys with N = 391 respondents)
77%
35.3%
(from 1500 potential respondents who received an invitation to participate, fully completed the survey N = 529 respondents)
Respondents in the quotas by gender and age: Male
Female
Age groups
10-19 =
20-29 =
30-39 =
40-49 =
50-59 =
60-78 =
47%
53%

32%
38%
14%
9%
5%
2%
46%
54%

29%
40%
12%
11%
6%
2%

Conclusions
Although access to internet as a medium is lower in the given population, telephone research suffers more from non-sampling error than online research. More responses from the same number of potential respondents are collected in online compared to the telephone research. High percentage of mobile phone users could not be reached, which puts a rate of reach for mobile telephone and for online medium in equal positions.

Comparing the quotas reached through telephone and online surveys, both research methods delivered similar split of the respondents within the age and gender groups.



Research analysis provides clear statistical and financial evidences that online research delivers reliable results with higher efficiency in terms of cost and time. Prior making the decision about using online data collection, the sample representation to the population and internet penetration level in the specific target group should be evaluated. While web has become one of the main communication channels in many target audiences, some target groups still have low access to internet and could not be suitable for online surveys. SolidData has the data about internet penetration and potential response rates available for various target groups. Please contact us to receive more detailed information.

References:
  • Henning, J. (2011, May 25). The true face of quality. Available: http://www.research-live.com/magazine/the-true-face-of-quality/4004916.article
  • ESOMAR World research Codes & guidelines, Conducting Market and opinion research using the Internet. (2005, August). Retrieved from ESOMAR [on-line], Available: http://www.esomar.org/uploads/pdf/ESOMAR_Codes&Guideline-Conducting_research_using_Internet.pdf
  • Miller, T. W. (2001). Can we trust the data in online research. Marketing research, Summer 2001, 26-32.

Copyright © Solid Data SIA
Online research panel in the Baltic States