Our clients have a lot of data. Geoff Copps, IPG Mediabrands Head of Reseach explains how we help them get the most out of it:
So while data is at the heart of everything I do, I tend to come at it from a slightly different angle to many marketers. When working with data, my primary aim is not so much to apply it in the CRM and executional phases of a campaign (there are other teams within IPG Mediabrands and its London agencies, UM and Initiative, who specialise in these areas).
Rather, I use data for insight and planning purposes: to understand and report customer behaviour, business performance, competitive context and consumer trends.
How best to go about unlocking such insight from data sets? In response to this challenge, I’ve formulated 5 broad strategies that, in the Mediabrands Marketing Sciences team, we apply in our day-to-day working practices:
This is about making the most of the full range of available information. It is not just about volume and comprehensiveness. It means combining client data (transactional, retailer store-level, CRM records), partner data and media data with some lateral thinking. To create fresh insights for clients we have taken advantage of less obvious sources, such as nutritional and environmental datasets.
Academics have long cited ‘discovery’ as the area within which computers are set to have the greatest impact on the future of white-collar work. It is easy to see why: machines work at scale, with precision, without human error, and are adept at pattern spotting. It is these considerable advantages that we are leveraging. We build bespoke ranking tools, customer segmentations, attribution models, and automated processes to aid search and social discovery. And we mustn’t forget the less glamorous but equally important prelude to ‘discovery’ at which machines excel: the heavy-duty tasks of data cleaning and blending.
Much of our data handling work involves optimising processes; less results in a ‘finished product’. Data visualisation is an exception. This is one reason why dashboarding remains a hot topic: automated insight using API feeds or server databases is an increasingly essential part of marketers’ toolkits. With obvious visible results, and with tracking and optimisation benefits for clients, it is not hard to see why.
The ultimate aim of any data project is to deliver improved outcomes for our clients’ businesses. Up to now, the most prominent role for automated processes inside media agencies has been in programmatic ad buying. By interrogating data using the other strategies briefly outlined above, we are able to make recommendations and build audiences for direct-match or lookalike targeting across the broadest range of variables, from lifestyle to location. Mediabrands’ programmatic trading platform, Cadreon, is adept at actioning these data and insight – and the Cadreon team continues to extend its capabilities, beyond display buys into OOH digital, internet radio, and the like.
In a world of automation and under-the-bonnet algorithms, due diligence is more important than ever. Never trust a black box; always ask questions. The most notorious example is probably the claims made for automated sentiment analysis within social listening tools.
What can you believe? A staff at a leading social analytics company recently confided to me the limitations of their ‘natural language processing’ algorithms: despite a claimed ‘85% accuracy’, the tool was almost completely incapable of interpreting a brand mention within a text whose subject could be perceived as inherently serious or potentially negative. No good for healthcare or financial services brands then… Throw in the complications of sarcasm, irony and the rest, and this sort of work clearly has a long way to go.
At Mediabrands we continue to develop our own transparent automated methods (using word proximity, account profile information, etc) to speed up social listening analysis. These methods we use advisedly, of course – for us, full social analysis necessarily involves human input.
So: 5 strategies to guide research practices. But as with all things data it’s difficult to be exhaustive; with enough imaginative and application, the possibilities for insight are virtually boundless.