Market research can experience its own zombie-like virus through sample contamination. What is sample contamination? Much like it sounds, it's sample of people who do not represent the traits of the intended population surveyed. Market research involves two groups of people - a sample and a population.
Let's say a company wants to launch a new widget product. The company wants to find out how appealing the new widget is to people across the country and understand their likelihood to purchase. Ideally the company would like to complete a telephone survey with every single resident in the country to fully understand the product's potential. Interviewing every single resident would be too expensive, take too long, and is impractical for a business. Therefore in many market research studies, surveying an entire population is not feasible.
Therefore samples (or smaller portions of populations) are created in market research so insight can be gained and estimations can be made so data from the sample can be cast to the population. For instance, the survey sample indicates 75% of people in our sample are likely to purchase the product so the margin of error predicts anywhere from 72% to 78% of the population would be likely to purchase the product. Understandably, it is critical that the sample be representative of the population. Much like a virus, sample contamination prevents the collection of healthy and reliable data.
Due to the prevalence of convenience samples, panels, and river sampling where anyone can opt-in to take a survey and in many cases share the survey link with others, ensuring your sample of completes matches your population is of critical importance. If you plan on using social media to publicize your survey, contamination can grow quickly because friends can share with other friends, and so on. If your friends share common demographic or behavioral traits, your survey sample can become contaminated quickly. For example, if the widget feasibility survey is shared on Facebook, chances are you will not get a strong response from the elderly where usage of Facebook is more limited than younger populations. If your new widget product is targeted to males 65+, chances are your Facebook completes and your sample will not match that. As a result, your data will not be accurate.
Sample contamination can be limited through random sampling efforts. For instance if the widget company wanted to conduct a customer satisfaction survey with users of its prior versions it could provide a market research with a random pull of 1,000 numbers from its 100,000 customers. The random pull will ensure that the telephone survey will sample proportionately. However, since telephone surveys are becoming less popular and as the industry shifts to online and mobile platforms, sample contamination has become a larger concern.
Sample contamination can also occur through participants stuffing the ballot box by entering multiple surveys thinking they will increase their chance of winning sweepstakes and other forms of survey sabotage. It's always important to keep a close eye on your data and run through routine and regular quality checks. Weighting survey data so your sample matches your population is always a last resort as well.