Don't waste money on bad retargeting data

First-party data has never been as important for marketers as it is today. Hence the ubiquitous drive to further develop the technologies that make this data usable. The data includes online signals from tags on a website or app SDK, but also data from CRM systems or offline marketing lists that are activated by connecting to a DMP. Unfortunately, many marketers still often use first-party data for overly naive, manual retargeting, leaving the full potential of their data untapped.

Marketers today rely heavily on retargeting - it delivers good results even with less sophisticated usage. In my opinion, however, too much money is spent on retargeting potential customers who have no clear interest or those who show the behavior of an intention to buy but in reality have no intention to buy. This refers to users who have only opened a website by mistake, for example, or who have only visited a page very briefly.

I think there are big differences in the quality of retargeting data and have found that the best retargeting inventory achieves a 10,000-fold higher conversion rate and is hundreds of times more effective in terms of cost-per-action than the worst. But most companies buy their data "flying blind", paying the same price for all qualities.

While retargeting performs well for most campaigns, advanced machine learning models based on artificial intelligence and self-learning algorithms can be at least as effective as the best retargeting. I can see this from the overlap that occurs when comparing the performance of retargeting and prospecting based traffic. This is really exciting: it means marketers can get very good retargeting-like performance and also reach new users instead of just targeting existing ones.

Of course, the best results are achieved with a combination of first-party data and detailed third-party data. However, it's not practical to do this in the usual way of manually creating composite audiences using Boolean combinations. I believe that marketers need a technology partner - one that can use machine learning models to automatically identify the most powerful combinations of data and determine their relative value.

Machine learning can be used to uncover connections in the data streams that are not apparent at first glance: Users who have just booked a trip, for example, are more likely to buy certain products in the 4 to 6 weeks before the trip. This was the result of a study by Rocket Fuel based on the evaluation of campaigns from the hotel and accommodation sector.

It has therefore proven to be very effective if the self-learning algorithms include this behavior as an attribute in their own campaigns for such products. This enables retailers to achieve higher campaign performance.

Many DSPs build models to identify and expand target groups that have previously worked well and use these models to divide people into target group segments. However, this step is only the first in a series of measures that can help you get the most out of your target groups. The same customer is sometimes more and sometimes less valuable for a campaign depending on context - depending on numerous contextual factors such as activity, currently visited website or app, time of day, weather and countless others. By augmenting this contextual data with relevant behavioral, demographic and first-party data in a single model, you can effectively assign a value to each potential moment of interaction. This means that each moment can draw from the learnings of the previous moment. In my view, this is the only approach to successfully harnessing the combined power of a DSP and a DMP - as well as the combined power of first-party data and third-party data.

If you haven't taken the plunge yet, do yourself a favor and avoid costly experimentation. Take the direct shortcut and combine your first-party data with machine learning, behavioral information, demographic data, contextual data and third-party data to get the best possible performance from your campaign.

About the author

Oliver Hülse is responsible for Rocket Fuel's business activities in Germany, Austria and Switzerland. He is responsible for customer relations with leading advertisers and agencies. He is also building up the Hamburg team and is currently recruiting for the areas of sales and account management. Oliver Hülse comes from the Adconion Media Group, where he was most recently Managing Director of the German GmbH. The 41-year-old from Hamburg began his career thirteen years ago as Sales Manager at the Milchstrasse publishing group and later at Tomorrow Internet. He then spent several years as VP International Operations at Interactive Media CCSP.

Rocket Fuel is a provider of a programmatic marketing platform with moment scoring technology. Its customers include brands and agencies that can use Rocket Fuel's two pillars - Data Management Platform (DMP) and Demand Side Platform (DSP) - to optimize performance, awareness and lift across all marketing objectives, channels and devices.