7 Questions Marketers Should Consider When Weighing the Quality of Their Data

The media landscape has realigned itself with lightning speed to the power of data. After a century of being limited by “brute force media,” marketers quickly glommed onto the vast potential of digital and addressable audiences.

The data “exhaust” from the digital experience has been a game changer for marketers, powered first by sites, then by ad networks and finally by programmatic. Money shifted from reach-based to “people-based” planning, augmented by powerful new data companies, monetizing the categories and groupings of people brands want to reach in ever finer granular detail.

But somewhere along the way, the proposition fractured. We discovered that data itself is not the key to addressable marketing and better business outcomes—quality data is. And the difference between data and quality data, or data with integrity, is difficult to see. To use a buzzword of the day, the market for data is not transparent. To use another, it’s a swamp: an opaque, poorly understood mess. If you want to be a data-driven marketer, you need to make your way through the morass to interrogate your data. So here are seven questions to keep in mind in assessing data integrity:

Is your data fresh or stale? The average life of a cookie is 30 days. About 55 million people change their phone carrier every year, 60 million physically move, 43 percent of customer records are out of date or invalid, and 60 percent of data is incorrect within two years.

Data becomes obsolete quickly, yet many providers continue to use stale data because it provides the illusion of scale. The only data that matters is accurate data. Make sure you understand how data is collected, whether it’s corroborated against authoritative identity standards, and how often obsolete data is purged.

Is your data 3-D or flat? In the world of data, there are six key areas that matter to marketers: demo (age, gender, income); geographic (where they live/roam); attention (what they concentrate on); consumption (what they buy); behavioral (what matters to them); and intentional (what they’re about to do).

Data providers act as if people exist independently in each of these areas, as if any of the above is sufficient to define a person. I’d ask, are you just a demo? Just a measure of attention? Just a signal of intent? No. Real humans are a combination of all of the above. Collectively, consumers are diverse. Individually, they are multifaceted. Flat data (an individual attribute) is just a signal.

How modeled is your data? Here’s a truth: All data is modeled. Here’s another: At some point, models falter. Do you know at which point?

In order to be useful, data needs to have scale. Marketers seek a balance of specificity and reach. It’s important to understand the size of the initial seed audience versus the size of the total audience to develop a degree of confidence in the data you’re using. If it’s significantly modeled, how certain can you be that you’re still reaching your target?

How transparent is the modeling? Do you know your look-alikes? The data market tends to be opaque, and with data, the devil is in the details. If you don’t know how a look-alike audience is formed, you have no idea whether it can be trusted.

For example, most data sets use only digital identifiers and connections. Definitive email-to-cookie linkages generate only a 30 to 50 percent match rate. So the data you’re starting with may be less than half right. Statistical modeling creates hypothetical look-alikes off the total (which is less than half right), exacerbating the issue. If you don’t know the model, you can’t interrogate the veracity of the data set.

This story first appeared in the April 24, 2017, issue of Adweek magazine. Click here to subscribe.

Recommended articles