Undoubtedly one of the biggest concerns raised about online research in South Africa (and other developing countries) is sample representativity. I recently stumbled across this blog post on insights 2.0Online market research in China and developing countries , which is a comment on this same issue in China and other developing countries. As a company that works in the online research space we felt it necessary to address the issue here and respond to this most common argument against online research in South Africa.

The main argument against online research in South Africa can be summarised as follows: because internet penetration amongst the general population in the country is low online research will not be able to deliver a representative sample of consumers in the general population, thus online marketing research in South Africa is untrustworthy. Accompanying this argument is usually the latest estimate of the internet penetration in the country (the percentage of the total population that has access to the internet). Current estimates place this figure at around 10.5% (http://www.internetworldstats.com/af/za.htm). So, it’s simple then – online research in South Africa is invalid since we can only reach one in every ten consumers via the internet, right? Not quite, the argument is a non sequitur – the conclusion (that online market research in South Africa is  untrustworthy) does not follow from the premises.  This argument has one simple but serious flaw which makes it untenable.

The single biggest problem with the argument is the assumption that market research studies as a rule target the general population for research. As we know, this quite simply is not the case. Most often companies are interested in a subset of the general population as their target markets. As researchers we engage in purposive sampling, and we have specific target groups in mind when we sample for research studies. Depending on the objectives we might be interested in: white males over the age of 40 who earn more than R50,000.00 per month;  black females between the ages of  30-50 who are loyal buyers of a specific brand of shampoo; 20-30 year old Apple iPod owners etc. You get the idea: a large part of marketing research is conducted with a very specific target market in mind that is a subset of the general population. The implication for online researchers is this – we need to look at the internet penetration amongst these specific target markets to assess the viability of an online approach. It seems reasonable that an online method will not yield an acceptable sample amongst rural consumers in socially disadvantaged areas. At the same time various other market segments are well represented online and can deliver acceptable samples. Furthermore, these online consumers are predominantly in the higher income groups meaning that they do possess significant spending power and thus are of interest to a variety of brands. Using the standard metric of internet penetration amongst the general population will not suffice, we need to adjust that mindset and look at the internet penetration amongst specific target groups on a case-by-case basis.

Additionally, for certain types of online research representativeness is far less of a concern; for example amongst certain target segments where customer or employee databases are provided or when the brand conducts its operations online – close to 100% representativity can be achieved. Hybrid data collection approaches (online and offline) and post weighting schemes can also address issues where certain target groups might not be reached online only.

The concern about representativeness is a necessary one and astute researchers will always be wary of limitations in their approaches. However, instinctive dismissal of online research in South Africa as untrustworthy based on sample representativity is unwarranted. If research buyers were offered reliable statistics on internet penetration amongst their target markets as opposed to the general population metric we can move toward a better understanding of the feasibility of conducting online research in a variety of settings. At Columinate this is very much part of our process  – a thorough evaluation of the feasibility of using an online method by looking at the internet penetration amongst the target market.

Any thoughts? What have your experiences been with online research and representativity? We welcome your views.