OPINION23 November 2017
While all the attention is on the data scientists, it’s market research where the really interesting work is says Ryan Howard.
Within a global data skills shortage, market research is not doing well. Recruiting data-savvy folk into the industry is a struggle.
While there is no shortage of data science candidates aspiring to what the Harvard Business Review called ‘The Sexiest Job of the 21st Century’, the demand for these skills has continued to surge. Those who otherwise would have found a fitting home in research have options, fuelling fierce competition. IBM now predicts that the demand for data scientists will soar by a further 28% within the next three years.
Making hay from a career path that does not lend itself to formal qualification are a plethora of ‘boot camps’. With enrolment in mind, they focus squarely upon machine learning. Artificial Intelligence is cool while munging and exploration are relegated to introductory footnotes.
Presented with artificial or crispy clean data, the underlying message is that automating prediction is the highest calling of a data scientist. This has resulted in a new type of candidate with CVs stuffed with names of algorithms but squeamish at the sight of real data.
Market research data is very real. Always messy, sometimes contradictory. Primarily concerned with a deeper explanation of consumer behaviour, we rarely have access to neatly pre-categorised datasets, the format required for training predictive algorithms. It’s not our bag. As such, we fall foul of skewed aspirations.
Market research needs to cut through with a compelling case if we hope to attract top talent, if any at all. What is it and to whom should we be speaking?
Behind the surging demand are forward thinking companies who also suffer hype. With inflated expectations and roles narrow in scope, they seek to leverage their solitary legacy database. Lo, two years down the road, their data scientist achieves little beyond intellectual curiosity.
The result is both unfulfilling for the incumbent and commercially unsustainable. The reality is that very few business problems arrive with enough pre-categorised data, replete with actionable, stable relationships, where automated prediction instructs how the core business operates.
For the time being, Amazons and Deliveroos are snowflakes. This is why the talking heads that heralded the universal application of machine learning as a panacea have gone to ground like homesick moles. The tone of the conversation has also switched, with commentators even challenging predictive algorithms’ role in society with their tendency to perpetuate unfair discrimination.
All this will improve gradually as data matures to drive a wider economy, but it really does not matter. Machine learning is destined to become a low-cost commodity, unworthy of attention and readily outsourced. The clue is in the name.
What once took a savant-like grasp of matrix algebra is now exquisitely wrapped up into four lines of copy and paste code. A fifth line optimises your settings to improve accuracy. Just Google it or dive into one of the readily available point and click tools. This means that anyone can wax lyrical about predictive analytics after a Saturday afternoon of YouTube. Still sound like the makings of the sexiest career?
On the other hand, we have market research, which does not measure the quality of a candidate by the number of algorithms they cite. Instead, it offers landscapes full of challenge and range, where novel and knotty questions are posed every day.
It busies itself by integrating all manner of data type and source, running the gamut of analyses, developing huge tool-kits including machine learning. This – set against a backdrop of an industry which exists to reward intellectual curiosity – is where the ability to interrogate data is lauded above all else. Yes, our case is as strong as it is simple. This was the type of sexy that Harvard meant, and our data revolution is just beginning.
Ryan Howard is director advanced analytics at Simpson Carpenter