Qualitative research has changed a lot over the past 10 -15 years – digital projects are now commonplace.
So it comes as a surprise that at certain conferences and in some benchmarking studies one gets the impression that qual has remained resistant to ongoing digital change and transformation – a laggard segment that needs pulling into the 21st century, it seems.
The authors of the latest GRIT study profess amazement at the finding that the focus group is still by far (58%) the most applied qualitative method worldwide, according to the report.
Their sardonic comment: “it would be as if the dominant quant method today still were surveys delivered by mail!”
The question “Why?” remains important, so the thinking seems to go; only qualitative research can really tackle that…. – but aren’t there chat bots that can do the whole thing much faster, cheaper, with larger samples and hardly any loss of quality? Come on guys!!
The word “dinosaur” hangs in the air – some stakeholders seem to be want to hook qualitative up to the much larger, presumably more lucrative area of analytics, big data and more.
What works “well” with quant – the use of automation and artificial intelligence – should also apply to qual, which should be scalable.
How to react to such messages?
Sure, there is a definite need for qual to continue on its path of digital transformation: agility is a must in today’s insights world; delivering insights late is totally without value.
And yes, there is certainly room for automation and AI in some larger qual projects – but applied with precision so that the software investment pays off. Some IT providers – discuss.io for example in the 2018 GRIT Report – write enthusiastically about the new options open to qual, using AI to sort and filter unstructured text and video data, saving time on the path-to-topline.
What’s less clear is the impact on pricing and cost structures.
There are, however, counter-arguments.
– 1. Qualitative research shouldn’t be mixed up with the evaluation of open-ended questions in a quant survey
– 2. Insights aren’t generated at the press of a button.
– 3. Insights generation (as yet) has very little to do with machines and robots, especially in qualitative research.
The third of these points is backed up by evidence presented in two award-winning studies on how people can best co-operate with robots, as presented at two recent Esomar conferences: one by Samantha Bond/ SKIM at the Global Qualitative 2017, which was awarded the Best Paper/ Peter Cooper Excellence Award; the second by Hyve/Beiersdorf, which was presented at Esomar Fusion 2018/ Big Data + Global Qualitative, also receiving the Best Paper Award.
Both are robust studies, with different research approaches, both take a close look at the man-machine dynamic – the SKIM study analyzed data from well over 100 video clips of people consuming food; with Hyve/Beiersdorf the focus was on the evaluation of larger online data sets. A netnography analysis was executed in a split-design, one cell purely managed by humans, the other fully automated.
The studies (in brief) came to the same conclusion: AI is only of limited help in qualitative research, even when larger amounts of data are involved. Machine intelligence certainly helps in sorting and filtering unstructured data which saves time; but it’s not useful in generating insights with a clear recommendation on what-to-do-next. The Hyve/Beiersdorf authors conclude:
“when conducting qualitative analysis, there is currently limited value in automated tools without human involvement”.
The business of interpretation, evaluating and fusing data sources, building on existing knowledge, drawing strategic conclusions – AI isn’t there yet.
The 2016 Unilever definition of an insight – “to inspire and provoke to enable transformational action” – is a task that requires human intelligence being brought to bear.
It should also be stressed that the very nature of qual, its raison d’etre, is to explore and understand, taking into account context and culture. The next step can be quantification and validation – or a re-visiting of qual for a focused deep-dive if needed. Qualitative research isn’t about numbers, larger sample sizes – it’s about exploring the power of n=1.
Of course qualitative folk need to surf the digital wave just like any other industry, identifying new digital opportunities, and yes, continue to evolve the hybrid “qual-quant” organizational model – exploring ways of delivering agile and in-depth insights more efficiently. But we shouldn’t get dazzled and dazed by every new-and-shiny, especially when validation claims are often over-stated.
Qualitative research skills are still in demand, however old-school that may sound ;) – 34% of all projects worldwide are qual, according to the results of the 2018 GRIT study. I’d go further – in an age of data overload, we need “people experts” with consultancy skills from the social sciences, from psychology, sociology, as well as the humanities, to partner with data-scientists to help make sense of myriad data streams: getting beyond correlation and squaring up to causality.
So to all industry associations, conference organizers, and other research influencers, here’s my take-out and call-for-action: qualitative research (however digitalized) is different and needs to be treated as such. It has clear consultative potential. Listening to skilled qual practitioners is almost always worthwhile, occasionally transformational; however small the supporting marketing budgets, there is value in amplification for a broad range of MR clients and marketing folk.
So here’s to flourishing qual futures – hybrid, interdisciplinary, insight-driven, digitally informed, human.
Scalability doesn’t have to be the Nr.1 priority.
P.S. For those interested, pop in to the GOR 2019 in Cologne in a few weeks’ time; we’re chairing a session entitled “Digitalization in Qualitative Research: opportunities, limitations” with three cool talks (SKIM Germany; Ericsson ConsumerLab; Happy Thinking People), examining different facets of the digitized world of qual.
P.P.S. This article was originally published in German at online portal marktforschung.de. The stage-one translation into English was via deepl, taking only a few minutes. The necessary and considerable re-work took well over an hour. Is that efficient? You be the judge.