Surveys are used to collect quantitative information about a population. It may be used to gather opinions or factual information by administering questions to individuals. Surveys can be structured (researcher administered) or as a questionnaire (administered by the respondent). The guidelines to conduct a survey is as follows:
- Define your customer and the intent of survey
- Review existing VOC data
- Review status against goals and commitments from existing VOC data
- Decide what to collect
- Determine target audience
- Stratify and segmen
- Determine sample size
- Select the deployment medium, tools and techniques (Data Collection & Data Processing & Analysis)
- Design the questionnaire
- Pilot the survey
- Modify and optimize the survey
- Collate and analyze results
As a rule of thumb, the sample size for a survey should be at least 5% of entire population or 30 samples per segment, whichever is larger. If more information about the population is available, the sample size can also be calculated statistically using the following formula.
- Nominal scale: E.g. How can our service be improved? (1. Faster 2. More features 3. More outlets)
- Ordinal scale: E.g. Rate your usefulness of the training (1-Low, 2, 3, 4, 5-High)
- Anchored: E.g. Rate your satisfaction with our product (Very dissatisfied, Dissatisfied, Neutral, Satisfied, Very satisfied)
The following are guidelines for the design of a survey to avoid bias:
- Do not ask the obvious (E.g. Would you like a free gift?)
- Do not ask leading questions (Do you find the chairs uncomfortable?)
- Do not ask ambiguous questions (E.g. Do you eat fruits frequently? Definition and frequency is not consistent)
- Do not ask respondent to quantify something which was not measured (E.g. How long did you to be served?)
The way the survey results will be analysed should have been determined during the development of the survey. This will also ensure that the survey will capture the information necessary to make an informed decision. Some of the common ways to analyse includes:
- Simple (E.g. Percentage of respondents who chose a particular response)
- Multi-Vari Analysis (E.g Percentage of respondents who chose a particular response segmented by demographics)
- Modeling (E.g. Regression or correlation of particular response choice with other responses)
The following are some tips for conducting a survey:
- First look for existing historical data before deciding on a survey.
- Do not ask customer the same question is problem has not been resolved since the last survey.
- Use customer interview or focus group to help frame the survey design.
- Decide upfront how survey results will be analysed and determine if information to be gathered is sufficient to fulfill the objectives of the survey.
- Ensure proper stratification and representative sampling.
- Select an appropriate deployment method that is relevant to your respondent.