By Craig Kolb,
Acentric Marketing Research
As published on Bizcommunity, 17 January 2014
The coming year promises to be an exciting one, with many new technologies and promising trends on the horizon. From the integration of big data and samples, to the growth of open panels and the tantalising possibility of automated marketing management; 2014 should be an interesting year.
‘Big data’ and traditional samples will come to be seen as complimentary
Big data pundits have pronounced the end of sampling on more than one occasion. More often than not, they seem to come from an IT or data mining background and haven’t looked beyond – what is for the most part – observational and transactional data.
Big data will never entirely replace samples for the simple reason that marketers still need to ask questions; at which point 100% response rates become impractical.
A poor understanding of sampling theory, adds to the confusion, with the automatic assumption that a full census is always vastly better than a sample, which of course is not true – typically the improvement in accuracy is miniscule – and the gains are confounded by other forms of error inherent in the data being collected.
Further than that, the human executive at the end of the process shouldn’t be forgotten. As humans we sample – we have a limited ability to process and experience data. Big data is fairly useless unless it conforms to our limitations (at least as long as humans are still making the decisions!). While certain types of data may be collected ‘continuously’, analytics and reporting tools do better by delivering it to users in significant chunks.
Bottom line – samples are here to stay, and the impact of big data on traditional surveys will most likely be felt in terms of what questions are asked, rather than in terms of outright replacement – sample-based surveys are in fact increasingly frequent, not less so.
Online-survey panels increasingly important and increasingly diverse
Five years ago, online surveys were rarely used. Today, they are more commonly used than telephonic and face-to-face surveys. Much of the credit must go to online panels, which have made the process of conducting online surveys practical. Online panels consist of large pools of pre-recruited and profiled consumers, providing a ready and immediate resource to researchers. Further than that, panels have improved quality by making incentives a regular feature of the survey process – while traditional approaches typically left consumers unrewarded. Panels have moved online surveys from the ‘hit and miss’ list-based approaches to a viable approach with high response rates. CINT, a Swedish company has moved panels a step further by turning them into an open market place that can benefit from multiple sources, rather than only relying on single-source panels built by research firms.
Rapid online surveys – a damaging trend
Demands for rapid turnaround continue unabated. While online panels have speeded up the process of collecting data, the rest of the research process is for the most part, as slow as ever. Software and processing speeds may have increased, but the human mind is limited in terms of absorption and interpretation of information. Yet, having had a taste of increases in speed in every area of life, many clients are increasingly impatient. The typical solution is to increase the number of invites to panellists – so while the rest of the process timeline remains relatively constant, the data collection timeline shrinks. In fact, certain European panels have millions of panellists and so are able to provide hundreds of responses within an hour or less.
However this a sure-fire way of decreasing the quality of the sample, since early responders have very different profiles to late responders. Stopping a survey after only a few hours means introducing serious biases into the sample. Good survey data collection timelines should be calculated in weeks not hours.
Social media analysis will continue to grow in importance as a supplementary means of gathering qualitative insights into what consumers are thinking. While some have claimed it would replace traditional research, this is unlikely due to the unguided nature of social media research. Further, social media research is prone to the same ‘acting’ and ‘lemming effect’ problems inherent in focus groups. One-on-one interviews produce an altogether different kind of data.
Automated marketing managers
‘Dream teams’, Fortune magazine’s annual list of the best executives may be replaced by ‘dream AIs’.
Given the growing supply of data, the one bottle neck in the system is the human executive. Many careers have been automated out of existence, while others have been enhanced – freeing up time to focus on more important activities. While managers may have once seemed impossible to replace, artificial intelligence is already replacing managers in some areas. An obvious example is Google ads, which have automated bid managers that decide what to pay for advertising, the prominence of advertising, when to advertise and where to advertise – replacing the role played by media planners to some extent. As artificial intelligence improves, and the ability to integrate the views of multiple stakeholders improves, major policy decision may eventually decentralise. ‘The wisdom of crowds’ notion has already been shown to have some merit, and the day may come where major strategic decisions are left to Swiss-style ‘direct democracies’ fed by recommendations from artificial intelligence systems such as Watson.