Making Data Work

Conventional wisdom holds that using third-party data for ad targeting is not effective, especially for direct-response-oriented campaigns; that despite all the fanfare about audience-based buying, micro-segmentation and intent-based-targeting, third-party data just doesn’t live up to the promise.  The reasons offered by doubters are varied – the data are too expensive, the data are bogus, inventory is all that matters, and so on – but the conclusion drawn is always the same: third-party data doesn’t work.

Actually, it does.  Third-party data can be an effective and efficient way to drive customer acquisition.   The challenge in making third-party data effective is in knowing what data to target.  With over 100,000 targetable data points available across dozens of data vendors, finding the right segments to target is a daunting task.  Most marketers take an intuitive approach and pick segments that seem aligned to their customer base.  But there is no guarantee that segments that feel like they should work will in fact drive results.  Additionally, segments chosen through intuition don't account for overlap with a marketer’s current customers.  This can lead to wasted impressions leading to wasted impressions and inflated costs, as marketers are better served targeting current customers via their own retargeting campaigns.



Rather than buy data based on intuition, marketers should buy based on quantifiable fact.  The state of RTB technology today is such that marketers can forecast which data segments are best aligned with their current customers before they ever spend a dime.  These proactive audience forecasts allow visibility into the data segments that are most aligned with a marketer’s customer base, and have highest likelihood of performance.  Many times, the results are surprising and counterintuitive – like the women’s apparel company that found a niche target in motorcycle-driving pet owners, and the mom’s services company that found Volvo and Audi wagon drivers were their most likely customers.  None of these segments were easily intuited, but they all worked.

Done right, in fact, proactive forecasting of third-party data can lead to exceptional performance.  We found that, on average, top performing third-party data segments perform 40X better than untargeted media.  And the best third-party segments produce results comparable to remarketing.   There are very few channels that produce comparable ROI.  Results will get better, too, as data providers offer up new levels of granularity and control. 

Products such as Exelate’s MAX modeling deconstruct the standard, pre-packaged data segments and hone in on the specific data points most relevant to a given advertiser.  Customized approaches like these are just now emerging, and they hold great promise in reducing false positives and further enhancing data targeting effectiveness.   

To be clear, there are no silver bullets in our world.  But the prevailing wisdom on third-party data targeting is dated.  If you aren’t already engaged in identifying the third-party data segments that can drive your business, now’s the time to start.

3 comments about "Making Data Work".
Check to receive email when comments are posted.
  1. William Hodges from Tiny Circle LLC, April 3, 2013 at 5:16 p.m.

    The wisdom isn't dated, but what people are saying is that it's a skill set with no easy answers. On top of that, it's an expensive skill set to learn, especially for buyers that have to target for multiple profiles and products. The solutions out there are modern gold mining techniques drawing further and further away from real marketing skills. Think we're missing a lot of power of real advertising.

  2. Peter Gasparini from uKnow, April 3, 2013 at 11:30 p.m.

    Nice succinct article Matt! Congratulations to you, Art and Accordant Team for building an effective, competitive trading desk.

  3. Eric Bosco from ChoiceStream, April 5, 2013 at 11:33 a.m.

    Great article Matt. We've found the same exact thing here at ChoiceStream. 3rd party data works great, but you need to have the algorithms and machine learning needed to mine it.

Next story loading loading..