Determining Big Data Best Practices at #SESSF 2013!

Posted in SES San Francisco

Welcome back to aimClear’s continuing coverage of #SESSF 2013! On our last day of the event, we talked data (BIG data!) in the discussion panel, How to Determine Business Objectives with a Data-Driven Approach. The panel members consisted of Chris Kerns, director of analytics at Mass Relevance, Brennan Carlson, senior vice president of product and strategy at Lyris, and Shay O’Reilly, senior business analytics manager, advertising solutions at Adobe – and, all the fun was moderated by Crispin Sheridan, senior director, global search, SAP.

For anyone that has ever felt overwhelmed from dealing with data, this discussion is for you – so, let’s get started!

Common Challenges In A Big Data World
First up, the panel discussed common challenges and misconceptions that go hand-in-hand with mining data. Carlson started out with a good, simple misconception: correlation does not always lead to causation. For example, when Hurricane Sandy hit the east coast, most of the tweets relating to the storm were coming from Manhattan. You might be quick to infer that this data tells us that Manhattan was therefore affected the most by the storm – but we know that was not the case. Rather, the most people with access to Twitter via smartphones were in Manhattan, and not in the areas affected most.

O’Reilly expanded, saying finding and utilizing good data is a huge challenge. If data is driving the big decisions of a company, it better be good data – or else you’re going to have a really bad driver guiding your company through critical decisions. And no one wants a bad driver.

The theme of data guilt is one the biggest challenges that Kerns runs into most. Data guilt is often a result of people and companies having too much data and thinking they have to use it all. Businesses will say, “We have 15 sources of data, so we need to use them all when we’re trying to achieve Y result.” This attitude, Kerns explains, is the complete opposite of how companies should approach their data. Instead, marketers should utilize only the most telling data sources. This is especially important to keep in mind as we move into an age where we continue to receive more and more access to data, especially from social sources. So, fight the guilt!

Handcuffed

Kerns added that marketers also need to get over the fear of reporting failures. Any digital marketer can spin metrics so they look positive for a client, but it’s really not doing anyone any favors. Trusting relationships within companies and between agencies and clients is important for getting the most out of data. Feeling pressured to spin metrics only sweeps problems under the rug. 

KPI Relationships
The panel moved on to cover the challenge of managing KPI relationships. O’Reilly responded first, acknowledging that this topic is a huge issue and raises a difficult question – how do we determine attribution? One way of tackling this challenge is by planning ahead. Looking at current numbers a company has access to and creating hypotheses ahead of time, rather than leaving all of the data analysis for the end, often helps with determining attribution. Carlson added that as digital attribution is on the rise, marketers need to direct extra attention to understanding what is enough information for appropriate insight – and avoiding the data guilt trap of trying to use too much.

Returning back to the idea that digital marketers should be selective with the data they choose to analyze, Kern commented that creating good KPIs starts with the right suite of metrics. Emphasis on the “suite” part, because using one single metric can skew results. For a recent, pop-culture reference, Kern turned to Miley Cyrus’s recent VMA performance (you know the one). The day after her controversial moves hit the airwaves, Cyrus took to Twitter to state that she [insert Tweet] – which to her, and many others, sounds like a successful outcome. But when Kern analyzed the sentiment of the Miley-Cyrus-VMA-related-tweets, it revealed a not-so-positive outcome. Kern found that the positive sentiment in Tweets about Cyrus was dropped by half after all that twerking – yikes!

Sparse Data
After discussing the challenges that can come with too much data, the panelists went on to talk about how marketers can deal with the flip side – too little information. O’Reilly began with thoughts on how the big data world is guiding companies toward personalized customer experiences. But this can be really difficult for businesses who only have a few touch points on their thousands or even millions of customers. How do you personalize an experience then? Because digital marketers don’t have all the information, or all the time in the world, we’re forced to utilize the approximate metrics available and fill in the blanks in order to create more customized experiences.

Carlson added that successfully managing and using sparse data takes a lot of trial and error. A great example of this is modeling, and saying that person A looks like person B, so we should serve them similar experiences. If it works, great. If not, continue with trial and error to fill in the blanks until it does work.

Building upon the idea, Kerns encouraged marketers to not be afraid to build their own data using a combination of human analysis and algorithms. For example, Twitter does this every day and creates their trending topics using a combination of humans and algorithms. The algorithms monitor for, and pull in, popular terms for humans to evaluate and then push out as trending topics.

Choosing the right KPIs for social campaigns
Last, the panel covered the challenge of determining appropriate KPIs for social campaigns. Kerns began by admitting that social campaigns get a bad wrap. Too many marketers are looking to sell things through social, so they start tweeting randomly without a plan. And no matter the company or number of followers, no campaign is successful without a plan.

A good example of a company that ran a successful social campaign is Hollister. The company started with a goal to increase online sales and decided to use a “flock-to-unlock” social campaign strategy. Hollister told its followers that if they received 50,000 Tweets, they would give away a $5 t-shirt to everyone who tweeted. If not, no t-shirt. The campaign was a huge success in several ways. Not only did Hollister receive its 50,000 Tweets and received tons of exposure, but they also got the sales. The Monday after the campaign, sales spiked by 44% compared to the average Monday. Oh, and they achieved 500% ROI – not bad! And it all started with a plan.

Carlson recommends not isolating social metrics, but rather looking at social data alongside other marketing channels for a more accurate picture of an audience.  And, he additionally touched on the fact that too many marketers don’t have a solid grasp on data science and this often lands them in a lot of trouble. To effectively use big data (and determine the right KPIs for a social campaign), digital marketers need to walk before they run – or, understand basic data mining algorithms before they try to do too much with big data and overcomplicate things.

O’Reilly agreed whole-heartedly, exclaiming, “Beware of geeks with data!”

Thanks for a very insightful panel, gentlemen. Stay tuned for even more coverage from SES San Francisco!

 Data image © Maksim Kabakou – Fotolia (dot) com
Handcuffs image © ruigsantos – Fotolia (dot) com