The rise of "big data" has been one of the key research trends of recent years.
On the whole, the trend has benefitted marketers, allowing them greater opportunities to track consumers in real time and develop more precisely targeted communications. But with this new power comes new challenges. Refining the massive volume of data available into useable insights remains difficult; it can be hard to get the c-suite interested in the data; and the regulatory future is also deeply uncertain.
It was with the aim of helping marketers meet these challenges that Warc organised its Datacentric conference, held in London in December 2011.
The question of how to apply data-driven insights to day-to-day business functions was tackled by many of the day's presentations.
To Dan Hagen, head of planning at Carat, trackable data has helped to raise the status of insight professionals. "The boffins will inherit the earth," he said. "We're going to get a larger volume of data. And we need to be equipped to deal with it."
One of the main benefits has been that big data allows for more precise targeting. To Hagen, over the next five years a major trend will be marketers moving from targeting the "average" consumer to targeting "unique" individuals.
There are four ways in which this is happening. The first is behavioural: marketers are getting an idea of what people are doing through web cookies, analytics and behavioural targeting software. The second involves using the client's own customer data, or proprietary data sets for agencies. Thirdly, marketers are increasingly monitoring social networks as a cheaper alternative to focus groups. Finally, they are using data to track how these individuals influence others.
But challenges for the future remain, with Hagen identifying three big obstacles to progress.
First, there is simply "too much data"; data derived from different sources might be contradictory, indicating that customers feel different things. Second, the need to react in real time, while exciting for agencies, is "terrifying" for clients wary of legal transgressions. The third problem is internal: the data must be visually interesting to executives if it is to be acted on.
If the c-suite thinks that data-driven insights are "boring", the insights are less likely to be used. Therefore, the data must be presented in a visually interesting way.
Lee Feinberg, senior manager for decision planning and visualization at Nokia USA, agreed that tweaking the way insights are presented will help make data less "boring". He suggested that there was a lot to be gained by turning data – usually presented via spreadsheets – into interesting images via data visualisation software.
"Companies are inherently decision-driven, but mostly we're talking about being data-driven," Feinberg said. "Many times, our data ends up as just numbers in reports, and that's not the way we can be most effective."
For the execution of the data visualisation, the most important thing is to get "really creative" to make the data stand out. "Polish and present your work in the best possible way," Feinberg added. "The most beautiful products are often the simplest ones."
Engaging internal stakeholders was also the key theme for Ray Poynter, senior vice president at Vision Critical. For Poynter, insight professionals must become more like storytellers to make the data they present more engaging – and more likely to be acted on by clients.
Presentations should be fun and interactive, asking the audience questions and tapping into the gamification trend that has shaped so much of the media world over recent years.
But insight professionals should be careful of using too many numbers. In fact, Poynter advised them to use as few as possible. Above all, "don't assume numeracy amongst clients and colleagues".
Shorful Islam, head of data at DDB UK, said he worked from the standpoint that "data is really about helping you make better decisions and becoming more efficient". For him, businesses should be constructed around data; data should also be shared; and, fundamentally, data should boost the bottom line.
Perhaps most importantly of all, data needs to be flexible, able to be used across the company. "If you're moving towards your organisation buying in the data, then it can't be used by only one team," Islam added. "You must be able to cut it in any way you want."
Building new consumer data systems is complex and costly. Ideally, a firm would build a brand-new integrated system. But in the real world, it might be necessary to receive inputs from existing (legacy) systems, which can sometimes be decades old and expensive to replace. The in-house financial system could be responsible for a firm's sales reporting, for example.
"You might feel this is inefficient, but re-building systems cost money," Islam said.
Moreover, teams within the organisation with their own data sets can often be "ferociously protective" and resist new systems. To remedy this, Islam suggested that the company creates a data guardian or guardians responsible for having the final say on these questions.
Finally, Islam suggested that there needed to be a cultural shift in the way data was perceived. "One of the core reasons data doesn't get adopted across organisations is that it isn't valued as an asset," he said. "There's a tendency to say, we'll keep three months of data and delete the rest."
The complexity of integrating data into organisations was also the guiding theme of a presentation from Gavin Meggs, strategic insight director of Sky IQ, Sky's consumer insights division.
Like Islam, he advocated taking a long-term view and not throwing away old data sets. "You need to evolve to become a learning organisation," Meggs said. And to become a learning organisation, you need to have a memory."
At the same time, it is important to use the data as a way of informing future communications and business strategy, rather than just as a way of analysing the past. Meggs added: "You need to ask, why did this happen, and how can we modify our behaviour."
For Malcolm Murdoch, director of digital data and performance at Mindshare, a major advantage of using online data is that it can provide real-time feedback on campaign success.
First, setting clear targets for the campaign in the initial brief is key. From there, to gauge the true impact of each media channel used, agencies should consider using statistical modelling to link measurable outcomes (impacts, calls, sales) to results.
Murdoch suggested that measurement need to be less "active" and more "passive", more "continuous" and less "discrete". In short, measuring real-time online mentions, or using weekly data such as footfall figures is going to be a lot more useful than the traditional brand tracker or focus groups.
This shift has massive implications for media agencies. With the development of better large-scale, passive measurement metrics, advertisers are going to be able to buy by audience – consumers with shared characteristics – rather than the traditional media buy based on placement. These buys are "cheaper and work better," Murdoch added: a view that is backed up by Mindshare data on brand uplifts and cost per acquisitions.
The shift also has implications for budget-setting. To Murdoch, "the most important thing is to keep money aside for optimisation. "Once you've got the metric so that you can change the campaign on a weekly basis, you need to keep budget free," he added.
One of the main ways companies measure trackable online data is Google Analytics, a free tool that provides online statistics. In her presentation, Louisa Middleton, product marketing manager at Google UK, gave tips on using tools of this kind.
Simply put, consumer purchses can be broken down into the "who", "what", "when", "where" and "why". Website analytics is strongest around the when, the where and the what, while other sources like POS data and shopper panels can provide insights around the who and the why. In other words, analytics, while useful, should only be used in conjunction with other research techniques.
Within the online world, clickstream data – recording the parts of the screen a computer user clicks on while web browsing – has become an essential part of the researcher's repertoire. This is because customers often voraciously research their options online before buying: for example, if they are going to go on holiday, they might check websites for air travel and ground transport, as well as looking up tour operators.
The picture gets still more complex when trying to relate the online to the offline world. Recent work by Deutsche Bank and GfK has suggested that almost half of people researched products online before purchasing offline, while just 10% did both researching and purchasing online.
When measuring the relative importance of online touchpoints to the customer's decision to buy, it is important to avoid misattribution and the trap of "last click wins". Google UK research cited by Middleton shows that 43% of online conversions involve "multi-click" journeys.
Ultimately, the path to purchase remains difficult to decode, and this process has become more difficult with the development of online touchpoints. But though the terrain is difficult to navigate, this does not mean that insight professionals should stop trying. Middleton advocates a five-step plan for using online data: investigate the data, develop hypotheses, test these hypotheses through experiments and modelling, feed results back into attribution models and optimise the data.
The Datacentric conference also included several case studies of how large companies, assisted by their agencies, are currently using data in campaigns.
In a joint presentation, Francesco D'Orazio, head of social media at Face Group and Jake Steadman, a brand manager at O2 UK, discussed results of a research project studying O2's 58,000 Twitter followers by looking at the audience data and their social graph (who they were connected to), then creating 3D maps of connections to identify sub-communities and shared interests.
The client, a UK mobile network provider, has been assiduously building its social media presence over recent months. O2 has also gone further than most in integrating data, having created a standalone business intelligence division earlier this year. "It's one team, under one roof, and that is going to help us put customer data even more at the heart of what we do," Steadman said.
Unsurprisingly, the people who chose to follow O2 were highly interconnected. Of a slice of 1,000 followers, there were 600,000 interconnections. The group was revealed to be highly London-centric; in terms of lifestage, there was a predominance of students.
In an attempt to uncover the group's influencers, D'Orazio noted that the Klout scores of O2 followers tends to be between 20-40; the more influential 50-70s are much rarer. Moreover, the research showed that around 60% of the audience did not tweet much; a hard core of 10% produced most of the content.
Time-wise, the major spike in tweets was between 8pm and 9pm.
In terms of the actual content of all the followers' tweets, synergies were uncovered between the brand and the audience. "We learned that most of the people were talking about competitions, technologies, and they were quite into what was coming up next – the future," D'Orazio said. This is, naturally, a good fit for a company that wants to provide users with new ways of using technology.
Real-time commentary on TV programmes was also very popular, with getglue.com – "a Foursquare for TV shows" unexpectedly popping up in the top 20 most-visited domains list.
In all, the brand received 3.1m mentions in a month from followers. Revealingly, three in four people tweeting @O2 didn't follow the brand, suggesting room for growing follower numbers in future.
Across all categories, dating websites have some of the richest online data sets. Indeed, in his presentation, Ottokar Rosenberger, UK country manager at eHarmony, one of the largest dating sites, claimed that his employer "collect[s] more data about our users than any other product in the world".
eHarmony uses an alternative approach to traditional search-based dating sites. Users cannot browse the site for lists of potential partners. Instead, eHarmony instead asks them to fill out a lengthly questionnaire when they join the site, thereby creating a very rich set of data that is highly revealing of the user's personality. "Data drives our customer experience, not the other way round," Rosenberger added.
In practice, the affinity matching model employed by eHarmony functions like an advanced model of Amazon recommendations, using a combination of questionnaire data and the user's own activity on the website to guess which other users they would be likely to talk to.
Real-time optimisation – one of the key advantages of using online data rather than traditional market research – is built in to eHarmony's processes. "We have algorithms that continually work out which users are receptive to which messages, and adjust our messaging accordingly," Rosenberger said. "There's a lot of complex modelling going on behind the scenes."
For example, it might be a good idea to match an extroverted "initiator" with a more introverted "respondent"; putting two of the first group together could result in the potential couple talking over each other, while putting two of the latter group together would make it less likely that a conversation would be struck up at all.
Rosenberger suggested that the smart way the firm uses data helps it find compatible couples, building brand awareness. eHarmony has certainly found success with its strategy: the firm already has 2 million users in the UK, having launched in 2008.
The advances in "big data" described by the presenters at the conference have the potential to reshape the ad industry completely. And yet, one of the major benefits of using real-time, trackable online data – serving more precisely-targeted ads – is under threat.
Barry Ryan, MRS standards and policy manager pointed out to delegates that regulators, concerned about user privacy, were threatening to crack down on the online behavioural advertising sector.
One of the major threats is the European Union's e-privacy directive, which is set to become law in all 27 member states. Once enforced, website owners might be required to provide specific "opt ins" to users to receive cookies: text-based files that follow their web browsing activity.
Ryan pointed out that the issue of web privacy is fraught, particularly as people have been sharing personal information on the web for years. "In many respects, the cat is out of the bag," Ryan added.
One of the main problems facing the industry is that the new laws are being drafted by people with a limited understanding of the technology. Worse, Ryan said, "they just have an instinctive dislike for it."
So the legislative environment is likely to get considerably tougher. In the EU, Ryan said that there would likely be a period of around two years before the directive is enforced, meaning that there is still capacity for the laws to be interpreted in a way that still allows the cookies to be used largely as before.
Ryan advised brands to attempt to reshape the conversation around targeted ads. "The way you manage this is to build trust with customers," he said.
"Say to them, 'here's my value to you'."
About the Author
Joseph Clift is a Web Producer for Warc.