This site uses cookies. By continuing to browse the site you are agreeing to our use of cookies. Find out more here

NEWS15 March 2018

Data science: ‘it’s so hyped', ‘let’s not try to integrate everything all at once’

Data analytics Impact 2018 News

UK – Data science allows researchers to mine a rich seam of information, but its scale, complexity and fragmentation can be overwhelming, according to a panel of marketers and researchers at MRS 2018.

Chaired by Corinne Moy, global director of marketing science at GfK UK, the panel comprised Owen Abbott, head of big data at Office for National Statistics (ONS), Nick Rich, vice-president of global market and consumer insights at InterContinental Hotels, Andrew Geoghegan, Diageo’s global head of consumer planning, and Claire Rainey, head of research change and continuous improvement at Sky.

Moy asked how — with a plethora of data out there, from company-commissioned research to open source info — companies were managing it.

Abbott explained that the ONS was "in the middle of a journey" and how the rich vein of data from open sources could revolutionise aspects of its work.

"We have a strategy called ‘better stats, better decisions'," he said. "Part of that is looking at new and alternative sources of data. A good example is the 10-year census – we're looking at how we can integrate other sources of data into the census to produce data outputs we've not produced before. 

"We're also looking at whether we can replace the census entirely with other integrated sources of data, saving the taxpayer lots of money and also providing up-to-date data and regular stats. A bit of a criticism is that census data is always out-of-date."

However, Diageo’s Geoghegan reckoned data science and its disruptive potential was "so hyped".

"It’s just another tool, like research and behavioural economics," he said. But he admitted that "for us it’s been really successful in applying data analysis to business problems like marketing effectiveness".

Sky’s Rainey agreed on Geoghegan’s point about hype, and suggested a more sober, less frenetic approach to its adoption. She pointed out that legacy data and the historical siloed make-up of businesses meant a lot of work was ahead.

"Going back a few years, everyone started talking about big data and being on a data lake without any oars," she said. "Let’s not try and integrate everything all at once as everything is growing exponentially. Try to do it more bit by bit on an incremental basis."

Geoghegan concurred. "If you try and solve everything, you'll get nowhere," he said.

Meanwhile, Intercontinental Hotels’ Rich highlighted the customer. "You work out where the customer is at the heart of this and arrive at the conclusion that the data journey is the customer journey," he said. "Get out of the mindset that you must bring everything together and stick it in this warehouse."

The panel agreed that while the array of data sets becoming available presented a multitude of opportunities, the need for market researchers was crucial.

"It still requires human beings to join the dots between data sets, which might tell you what but don't tell you why," Geoghegan said. “Look for agencies that can work synergistically with all the data and research available,” he added.

@RESEARCH LIVE

1 Comment

9 months ago

A fascinating debate, that's for sure. Fundamentally, just because we can integrate data together, it doesn't mean we should. It's important to use all of those skills that we've picked up - whether as analysts, researchers, statisticians - that is to think about the purpose of the data fusing that we're attempting. What are the outcomes that we're looking to achieve from bringing some data together? What added value will this combined data bring to the story? Is there causation (or just correlation) between data sets? Without a clear purpose, data science can be a bit like looking for a needle in a haystack, and then what you find may not be all that interesting! The key principles for analysis are not rocket science: What's the problem we're looking to solve? What hypotheses do we wish to test? What data do we have that could be of use? Which gaps do we need to fill? How can we deliver our findings in a clear and compelling manner? Of course, there's some hard number work in the middle - but we should all remember that data science is just another tool in armoury - one to be used alongside market research, desk research and to a certain extent, instinct.

Like Report