10 Things We Learned About Big Data At The Analytics With Purpose Conference

Over the last two days in San Diego, BrainJuicer has been proud to co-present the first Analytics With Purpose conference with the AMA. The subtitle – The Human Edge Of Data – was a pointer to the fact that we both wanted this to be more than “just another Big Data conference”, but instead present a panorama of the complex relationships between marketing, research, people and large-scale data. In this post conference Chair Tom Ewing looks back on the event and pulls out 10 talking points.

ama panorama

A panorama of the conference by John Szabo

In her opening keynote at Analytics With Purpoise, IBM’s Elana Anderson outlined the essential problem – and opportunity – confronting marketers in an age of big data: the challenge of individual engagement at scale. With such vast and deep knowledge of people’s behaviour, interaction and interests, where are the ways to practically use it? Over the course of two days in San Diego, I can’t pretend we solved the problem but we explored a few pathways and had fun doing so.

The next ten paragraphs are a summary with a difference for a conference with a difference – a series of soundbites which sum up the event and spotlight some of its themes. If it makes you wish you were there – well, that’s the idea, and keep an eye open for sequel events!

“There’s an API for that”: Rachel Delacour of BIME was one of the most optimistic speakers, a big data evangelist who co-founded her own start-up to make large-scale data more useable (she is also the only presenter I’ve seen criticised at a conference for not being salesy enough! Her audience made her demo BIME’s impressive cloud-based analysis package.) One of her key points is that we have entered an age where almost every consumer activity can be tracked ambiently and often on-the-go. It seems we are truly entering a world without questions.

“Create things that yield data”: David Matathia of Hyundai gave a compelling keynote in which he announced the end of the message ad – instead marketers need to build large-scale experiences which can yield data themselves. For instance, this mash up of Hyundai branding and Google street-view data let a virtual Hyundai drive to the consumer’s door – and in the process reveal as much information as a very lengthy survey would.

“Big brand trackers are dead”: The message ad wasn’t the only thing feeling the pinch. A “Problem And Resolution” panel on Day 1 was unanimous that lengthy tracking surveys were the Problem, and the Resolution was an unmarked grave. In an age of large-scale data, letting your budget be eaten by an unwieldy tracker was a fool’s errand. Speakers including Lowes’ Kyle Nel and OdinText’s Tom HC Anderson vied to outbid each other on the minimum viable length of a tracker (Five questions! No, three! No, one!) and it was left to Dan Stradtman of GE, speaking from the floor, to offer a defense of the large scale tracker – but only if it was integrated with big data, not replacing it.

“Blah blah blah, blah blah, blah blah”: This, you might think, is a soundbite from every conference – but in this case it was some “ad copy” tested by Elea Feit from the Wharton Customer Analytics Initiative, alongside more sensible executions. The point was twofold – to show how easily anyone can run A/B tests on their ads, and to introduce the idea of a “negative control” – testing something you know is bad to provide a baseline for your good tests. In this case, though, the “Blah blah blah” execution got the highest click-through rate – not a metric anyone should rely on, of course, but it proved a third point: how boring most targeted ads are!

blah blah

“It’s (not) the math that matters”: So you’ve got your data – what do you do with it? John Walker of Prophet presented the ten myths of big data, and one was “It’s the math that matters”. It’s not. It’s the interpretation and the story you can tell around it – a major theme of the conference.

“If you’re looking for a relationship, you’ll find one”: If you found a 0.97 correlation between two variables, would you be delighted, or suspicious? If you’re Google’s Chris Chapman, you’d be suspicious. By correlating random time series data with Google searches using Google Correlate, he demonstrated that given a big enough data set you will always find some correlations which are astonishingly high and utterly meaningless. You have to go in with prior knowledge and a framework for analysis – you can’t simply trust the numbers.

“Kill your ideas”: A related insight came from Thomas Zoega Ramsoy, a neuromarketer who gave a fascinating presentation debunking some of the more popular neuroscience “insights” – pointing out, for instance, that different popular writers breathlessly attributed repulsion and love to the exact same part of the brain. Ramsoy had some excellent advice for neuromarketers drawn from the scientific method, which applies to researchers too: instead of looking to data to prove your idea, look to data to kill it off. If it survives, you’re probably onto something.

“Why not leave people in?”: Getty Images’ Alex Mitchell put an important question to data communicators – if we’re talking to people, about people, why not center our presentation on people? The rise of big data has gone hand-in-hand with the fashion for infographics – beautiful but often abstract visualisations of large-scale data sets that end up taking much of the human element out. Refocusing data on the human stories behind it, and the human experience of it, was a theme underlying a great deal of the event. (See an experiment we did with Getty Images on this topic here).

“Big data is a culture of theft”: Andrew Greenfield of Vision Critical stepped up on Day 2 to take a more critical look at some of the assumptions underpinning the big data discussion – namely, that the firehose of data is endless and will only continue to grow. Not so, Greenfield said – we’re in the Wild West right now, and a Lawman might turn up at any moment. You have to plan for privacy concerns and potential regulation, because these things will come up – the idea of the transparent oversharer is a myth perpetrated by data special interests, and surveys show consumer concern over privacy is a sleeping giant. Greenfield’s solution? Get people involved with their data – whether as owners or active collaborators.

“With big data, you have to start small”: And finally, a theme which can’t really be attributed to any one speaker, as so many people talked about “small data” and starting small. Whether it’s learning the ropes of experimentation, building up synthesised data sources, or simply starting with individual goals and questions, you can get the most out of big data by ignoring its bigness. And as the terrific closing speaker Tracy Wong of WDCW pointed out, in advertising at least big data is no match yet for the big idea. “Sometimes,” as he said, “the answer will come not from the survey of millions but from the voice of one.”


4 thoughts on “10 Things We Learned About Big Data At The Analytics With Purpose Conference

  1. Pingback: Brainjuicer write-up of the AMA “Analytics With Purpose” Conference | What + Why

  2. Pingback: Our Best Blog Posts Of 2013 | Brian Juicer Blog

Go on, tell us how you feel!

Fill in your details below or click an icon to log in:

WordPress.com Logo

You are commenting using your WordPress.com account. Log Out / Change )

Twitter picture

You are commenting using your Twitter account. Log Out / Change )

Facebook photo

You are commenting using your Facebook account. Log Out / Change )

Google+ photo

You are commenting using your Google+ account. Log Out / Change )

Connecting to %s