Paralysis By Analysis: Is Too Much Data A Bad Thing?

anthony-katsur-ddt“Data Driven Thinking” is written by members of the media community and contains fresh ideas on the digital revolution in media.

Today’s column is written by Anthony Katsur, CEO of Maxifier.

Want to target your online ad to men living in Wisconsin who like to read about politics, are researching a vacation in Hawaii, have an interest in motorcycles, and listen to classical music? Now you can! Fantastic, right?

This is hyperbole, of course, but it illustrates how the advent of digital has heralded hyperfocused targeting, segmentation, placement, and performance to the degree that niche audiences can be defined and targeted with tactical precision. Data-management platforms enable marketers to collect a wealth of behavioral information, such as sites or pages visited, searches made, content viewed, frequency of consuming specific types of content, and time spent on certain actions, and combine it with third-party data sources to provide a data-rich audience segment.

But is all this data actually improving advertising? Are we driving the key metrics that are important to the marketer? Are we improving efficacy, or are we just finding efficiencies in reaching a cookie pool? Our focus on data has attracted and supported direct response more than any other channel, but is it also limiting us from unlocking brands?

I think our obsession with data may, in some ways, be limiting our growth instead of boosting it. As an industry, we’re drowning in a sea of data because we believe that more is better. But, too often, it feels as though we’re throwing data at our clients simply to see what will stick. Instead, we should simplify our targeting and audience segmentation and support data with relevant attributions and metrics.

If we look at the offline data giants such as Experian and Acxiom, we can see that these companies have achieved success by sticking to the guiding principle that “less can be more.” In spite of their huge volumes of data and many years of experience developing accurate and relevant segmentation, they offer only a small number of audience segments. Experian’s Mosaic USA, for example, breaks the population into 71 segments, which fit into 19 broader groups. While it could create many more segments, it recognizes that giving marketers more to choose from wouldn’t win it more business. On the contrary, choice can be demotivating. Just think of large versus small restaurant menus, for example. More choice means more confusion and more procrastination, whereas fewer choices -- as long as they’re good ones -- can make finding the right fit easy and quick.

Great content and context can often be more than enough, and I would argue that this combination is vastly undervalued and overlooked. All too often, online, the cookie has become the sole arbiter of audience qualification, unlike in TV, where content is still just as important and effective as an audience proxy. There’s a reason male-oriented advertising appears during sports games and female-targeted advertising runs during “The Bachelorette.”

Online Lessons From The Offline World

We can learn lessons from the offline world about how online targeting should evolve and be packaged into simple, yet scalable, audience segments. Marketers want to target segments that are selectable and scalable while ensuring their desired audiences can be reached across all campaign mediums. If we can address these needs to help drive online targeting, perhaps we can make the Web a more relevant environment for brands, further encouraging more digital advertising.

There’s no easy path, but I would argue that the amount of data currently employed almost feels like data for data’s sake. While traditional channels often don’t come close to the volume and recency of digital data, online remains the bastion of the direct-response advertiser while brands continue to invest in traditional channels to achieve their goals. Digital offers great starting data points which should resonate with brands, but this is only part of the story. How, for example, do reach, frequency, and viewability translate into the brand awareness and favorability metrics that brands want to measure online? We have the building blocks, but we need to craft the data and metrics to create a compelling value proposition that advertisers want to buy into.

Don’t get me wrong, I’m not saying all this data is a bad thing. After all, no other marketing channel is as measurable as digital. But there needs to be a balanced approach to its use. In an attempt to quantify every element of the digital-media transaction, we get caught up in the data (take a shot every time you hear “big data” thrown around our industry!), while often ignoring the fundamentals of marketing.

Some of the hesitancy around embracing digital is probably a cultural or comfort issue. The old adage “no one ever got fired for buying IBM” applies; similarly, no one is likely to get fired for buying advertising on “Mad Men.” Traditional channels are not only safe, but they’re packaged with supporting data and relevant metrics that make it easier for brands to work with them. We’re still in the early days of our digital culture, but we need to realize that jargon, complexity, and vast amounts of data can alienate the very brands we want to attract. We need, instead, to take steps to bring information together in a way that is simple, scalable, and translates into the vernacular of the marketer.

The a la carte menu of data we’re providing -- without solid correlation to the success metrics of brand marketers -- is type-casting our entire ecosystem. Our obsession with data, while one of our greatest strengths, is also our Achilles’ heel.

Follow Anthony Katsur (@sleepwhendead) and AdExchanger (@adexchanger) on Twitter.

2 Comments

  1. Disagree. Algorithms are data hungry. This man said it best. "We don’t have better algorithms. We just have more data." - Peter Norvig, Google.

    Reply
  2. Hi Jonathan,

    Thanks for taking the time to read the article, but I think you missed the point. Algorithms are data hungry, no question. However, there is an art and science to how we use it.

    1. We're feeding marketers data, not metrics. Throwing data at marketers to see what sticks is not how we scale the digital marketing for new types of buyers. Crafting solid metrics that matters to a marketer based on good "foundational" data is where we need to head. Data for data's sake doesn't mean anything unless it's tied to a marketer's goal, which are often much more complex than a CTR, engagement rate or % viewable. Those data points might be crafted into a credible and compelling manner to support the end goal of the CMO.

    2. Very closely tied to my first point, how we package data matters. Take again the example of the offline world. Arguably, they also have an almost infinite amount of data points/audience segments/target markets they could bring to marketers, but they don't? Why is that? It's because they've keyed on the data that drives results and back those results up with strongly correlated metrics. However, in the digital world, your choices border on infinite. It's too much and can be more than overwhelming.

    3. Also, there are reasonable limits to data as an input. In the words of a very smart scientist in our space (I won't volunteer his name here), "Don't believe in sampling? The next time you go for a physical, give the doctor all of your blood."

    I think there's a "good enough" we're missing here. I believe that goes along with any great, new discovery. In this case, it's our ability to process, store and model massive amounts of information. The pendulum has swung too far in one direction. It's time to bring it back to center, which starts with the question to the marketer, "What is your goal?" Once we understand that, then we can start optimizing to specific and arguably more complex success metrics supported by the mounds of data our algorithms can devour.

    Reply

Add a comment

XHTML: You can use these tags: <a href="" title=""> <abbr title=""> <acronym title=""> <b> <blockquote cite=""> <cite> <code> <del datetime=""> <em> <i> <q cite=""> <s> <strike> <strong>