• by Jeremy Hlavacek , October 3, 2013

It’s hard to pinpoint exactly when the term “big data” was invented, but about 18 to 24 months ago the “big data industry” started receiving a lot of attention and was trumpeted as the next big thing. Headlines blared, projections were made, and excitement bubbled over. Recently however, there has been some retrenchment and second-guessing the big data space, with articles talking about “big-data disappointment.”This seems to be right on schedule with most shiny-new-object hype cycles, so perhaps the truth lies somewhere in between irrational exuberance and utter disappointment.

First off, referring to big data as an “industry” can be confusing, because although there are companies that specialize in operational functions around data (storage, processing, visualization, etc.), almost every business has some sort of data/analytics practice or business intelligence unit.  Therefore measuring the output of the big data “industry” can be difficult. A better approach might be to think of big data as a market shift that requires rethinking how large data sets apply to core businesses.

So what does this market shift mean for online publishing and media? Is it time to give up on big data? On the contrary, now that the hype is wearing off, it’s the ideal time to double down on data strategy and dig into what “big data” means on a practical basis. With that in mind, here are a few areas for publishers to think about as they consider how to best monetize big data assets with clients.

• Real Time and Recent. How recent is your data? Is it being used in real time? Data can go stale quickly, especially when being used in fast-moving verticals like retail and CPG. A consumer in market for diapers/baby products is going to make a purchase pretty quickly -- probably in 24-72 hours or less. This is why retargeting is arguably the biggest winner in the big data game thus far. The power of the real-time media ecosystem is being able to use data to reach consumers at the precise moment in time when they want to hear from an advertiser. How good does that cold beer ad look on a hot summer day? If you’ve just purchased a car, seeing a low rate from an insurer might be a welcome message.

• Rare or Unique. One of the biggest issues with “big data” is that there is too much focus on the word “big.”  Data sets need to be broad in order to apply to the scale of large advertising campaigns, but truly valuable data should also be rare and not easy to replicate. When publishers look at their data assets, they should be aware not only of how much scale they have, but also of how many unique attributes they have compared to competitors. Are you looking to sell auto-intender segments? That’s fine, but be prepared to compete with a lot of others who can claim the same thing. A better approach might be to pick something endemic to your brand -- ideally, a large data set that no one else has.

• Realistic Application and Expectations. Once the competence in data has been built, there’s a temptation to go overboard and become “data-happy.” It’s easy to do. Clients will suggest using a behavioral data set and a demographic data set and a geo-targeting data set all on one campaign to create the perfect dream segment. There’s only one problem. It also might create a unicorn: the perfect target that doesn’t exist and certainly doesn’t scale. A better approach is to test and learn with regularity and explore reasonable hypotheses. And don’t forget that marketing and economic principles still apply. Big clients will always want scale and efficiency, so sometimes using big-data sets just might not make sense.