Online and offline surveys represent not just two different mediums of conducting market research but often also two different methodologies of administering questionnaires and collecting data in research.
Unlike offline surveys, online surveys are mostly devoid of human interaction and are faceless, automated surveys which are self-administered and self-filled by respondents themselves on a digital screen remotely. Accordingly, the right orientation to conducting online surveys is to approach and design them directly and not in reference to offline surveys.
In designing and conducting online surveys, the quartet of Sampling Methodology, Questionnaire Framing, Participation Incentive, and Data Validation objectives should be defined and outlined very carefully, and kept in mind and monitored equally carefully while implementing/conducting the survey online.
The Right Sampling Approach
Choosing the right sample size and sampling method is most critical in ensuring ‘good quality’ research, irrespective of the medium from where the sample is collected. And to be of good quality, a research study must have a sample composition that is ‘representative’ of the population, a sample size that must be ‘statistically significant’ and a sample collection method that must be free of ‘sampling biases’ as much as possible.
The sample size of any research survey practically depends on the level of population representation one is trying to achieve in the research, and the depth of analysis (segment level analysis) one wishes to do, and not on the medium of conducting the survey. The more finely we want to make the sample represent the population and its various segments, the higher is the sample size we would need to take and achieve.
Sampling biases can creep into any research easily, either by a faulty survey design or by a systematic willingness or unwillingness of some potential respondents to participate more, or less, in the survey. Such biased sampling can end up skewing the data, making them non- or less-representative of the population. Although it is often practically difficult to completely eliminate sampling biases in online research, one should at least attempt to reduce or minimize them as much as possible.
The Right Questionnaire Framing Approach
In online surveys, the questions (and the instructions to fill the survey) need to be phrased keeping in mind the ‘target respondent’ and not the ‘interviewer’, as it is the target respondent who will read and answer the questions completely on his/her own, without any directions or guidance from any interviewer.
Accordingly, the questions need to be in a language that the respondents are comfortable reading, comprehending and responding in, and the choice of words and sentences need to be very simple, designed for the ‘least educated among the respondents’ to be able to understand and answer them.
‘Leading’ questions that (intentionally or unintentionally) induce/force/direct respondents to select or give a specific answer should be completely avoided. Also, the length of the online questionnaire should not be long as the respondent may lose interest, get fatigued and drop off. This may affect the quality of responses and therefore results.
The Right Inducement to Participate
In online surveys, as there is very little human interaction, there are no social courtesies, niceties, or pressures on the respondents to participate. They either participate of their free will or they don’t. So, the critical criteria for participating in online surveys often boils down to the respondent’s ‘motivation’ to participate, which is a combination his/her willingness to make an effort to fill the survey and having the required spare time to do so.
This results in very few people willing to participate in such surveys unless there is some kind of inducement or incentive, whether financial or non-financial. Without such incentives, the respondents may either not fill the survey online at all, or may fill them far too casually than required. So, with something at stake (the promised reward), and something to lose (disqualification from winning that reward if the filled responses are improper or incorrect), the respondents often tend to respond more honestly and sincerely.
Therefore, the incentives offered, and the terms and conditions outlined for potential respondents to participate and complete online surveys, should not only aim to make the survey participation a more ‘fruitful’ experience for the respondent but also to induce more sincere and honest responding.
The Right Data Validation Approach
Data validation has two distinct dimensions that must be taken care of in any research – firstly, the authenticity of the ‘respondent’ and thereafter the authenticity of the ‘response’.
As far as ensuring the ‘authenticity’ of the ‘respondent’ is concerned, it is more about running proper back-checks and validating the respondent profile and not so much about which medium of research is being used per se. In that context, irrespective of the medium of survey, the authenticity of the respondent is always more controllable and identifiable in a ‘panel’ where the respondent has an ongoing relationship with the research or the marketing firm and has an ongoing stake in identifying himself/herself correctly, filling the survey sincerely and honestly to continue that relationship.
So, unless the requirement is of a randomized sampling, using a consumer/research panel could be seen as a more appropriate and authentic source of sample collection than catching respondents from varied and random sources and places.
When it comes to the authenticity of the ‘response’, a highly secure and rigorous data monitoring and validation process needs to be set up even before the survey is fielded. A good, robust data validation framework must ensure that all the right question and answer logics are built within the questionnaire to capture as clean data as possible. You must also incorporate a proper ‘de-duping’ (or de-duplication) process in your survey fielding so that the same person is not able to participate again in the same survey.
Last but not least, you must also rigorously follow some of the post-sample collection data cleaning steps like checking for ‘in-a-jiffy’ responses, ‘straight liners’, ‘outliers’ as well as ‘junk’ responses and clean them out.
