Big data needs to be more than just expensive noise

By James Robinson , written on May 16, 2014

From The News Desk

For all the vaunted potential of big data, the consumer benefits aren't fully apparent yet. Netflix still recommends me movies I don’t want to watch, Spotify still nudges me towards bands I don’t particularly like and most of the companies targeting ads at me on Facebook are still barking up the wrong tree.

But as Russell Marsh, Global Chief Data Officer for IPG Mediabrands -- one of the world’s largest media agencies with control of $36 billion in media assets -- explains to me, the tools are there but people are just doing it wrong.

“Spotify and Netflix don’t model their data. They have a captive market already. They’re not paid to be good at this,” Marsh says.

Marsh cites a ream of case studies to how big data (or connected data, as he chooses to refer it) has already hit large in advertising. These all sit directly at the border of amazing and creepy. He says that recently he's worked with a client which sells antihistamines, looking at weather and pollen data in order to plan their ad spending. Data showed that people stricken with hay fever buy pills after their allergies flare up, and advertising has real impact only when run at the perfect moment in a change of season. “We saw that there was no point buying media five days before people would need the pills,” he says.

Recently, Marsh says he found himself in a meeting with EE, a UK telecommunications company who were showing off their data, tracking the paths of telecommuters in and out of London each day, by accordance of which cell tower they are bouncing off. For him, there's great potential from that to use that information and optimize where IPG Mediabrands will buy its outdoor media spots.

The capabilities have got so great in recent years, that Marsh says he’s developed a few tenets, a set of tough questions to ask himself, to make sure that he’s not overstepping the boundaries of taste. “First of all you think technically, can we do it? Then it is whether commercially if it makes sense and is legal," Marsh says.

“But the fourth thing you have to ask is morally, should we do it? Are we overstepping? Would this damage our brand? Ultimately, I could take data from certain places, find out sexual orientation, target different groups maybe doing things they shouldn’t be. At Mediabrands that fourth question isn’t asked that often, but I’ve definitely been involved in conversations like that.”

Advertising like this, using available data to its best potential, is difficult. As Marsh explains, as  you chase the advertising holy grail of supplying the right message, at the right place and at the right time to the most receptive customer, the segment of people you’re targeting gets smaller and smaller. SKY UK recently began working with AdSmart, which pre-downloads 2,000 separate ads to a cable box and then individually targets ads at that one particular household. The smaller each targeted audience gets, the more audiences you have to target at the same time, to reach the same amount of people. “Which then opens up a really interesting question, to what happens with the creative when you have 20,000 segments targeted. You can’t have 20,000 bits of creative,” he says.

This style of advertising, with predictive analytics and smaller and smaller targets, pushes traditional advertising closer to being a CRM system. “It is about targeting moments, where you can start identifying behaviors,” Marsh says. He says that when you find out a woman is pregnant, you know that household will be in the market for push chairs and car seats. You can buy databases with information about who has moved house in a neighborhood, who will all be receptive to offers for new electronics and house repainting.

It is all the beginning of a wonderful, terrifying journey. As Marsh points out, there was no Instagram or WhatsApp five years ago. Ten years ago, Facebook was just getting started. He says that the amount of available data has doubled in the last five years and could increase ten time over in the next five years. The challenge will be to not become overwhelmed.

“Having all of this data is really interesting. But working out what is valuable is the hard part. Because the rest is just expensive noise,” Marsh says.

A big, boring piece of the puzzle that needs to fall in place for media agencies to really start having fun is data standardization. Individual companies are often siloed into multiple parts, with data trapped inside different corners. Companies can be called different things in different documents. There are no normalized data keys, so as far as IPG Mediabrands goes, often interpreting a client’s data is like translating different languages from the same source.

Our experiences with this brand of advertising will grow. It is going to be fascinating to see if people push back and what will be considered going too far. Marsh says that at a large media agency level, IPG Mediabrands doesn't see personally identifiable data. But when you start seeing Christmas tree ads during the precise weekend of the year you buy one for your family, will that be any less disturbing?