Balancing Data and Science in Your Audience Targeting

By Victor Alonso
Consultant, Research and Development
Media Center of Excellence, IRI

The application of data science over the past 15 years has given marketers the incredible opportunity to be much more data-driven to gain the insights they need to plan, target and execute media campaigns.  The data science evolution has also enabled marketers to fill in gaps where sufficient data hasn’t been available. You can see this on a regular basis across an entire media campaign execution lifecycle. Data science has an incredibly important role to play in many areas, such as adjusting for bias. However, all of the continuous enhancements to data collection, storage and processing mean that marketers can now build data assets at scale requiring less science in some cases and allowing the data to speak for itself. This requires marketers to understand the fine balance of when and how to use data and science appropriately.   

Let’s look at an audience targeting example where this plays out:

Due to the short purchase cycle in the CPG industry, the best predictor of a future purchase is a past purchase. As a result, data providers have been in the market with purchase-based audiences for some time. But, when taking a closer look, you’ll see that most of these audiences 1) are made up of a smaller seed set of households that have actually purchased, and 2) include common attributes among those households that have been established and then run through a statistical model where, based on the identified attributes, the model then scores each household in the U.S.  So while this audience may be marketed as “purchase-based,” only a fraction of the households that make up the audience are actually known to be purchasers. Making things a bit murkier are situations when there is not a lot of transparency around how recent the data is that’s being used to build the modeled audience or how much actual purchase data was used to build out the audience. If you are comfortable with the data and modeling process used, using a modeled or propensity-based audience is a great way to give you the reach you are looking for. Depending on the campaign objective, modeled audiences are a great tool with a great use case.

Today, there are additional tools available to compliment a modeled approach. With access to a robust loyalty card dataset, purchase-based audiences can be delivered in a 100% deterministic fashion and at scale. This ensures that ALL the households in that particular audience have been verified to exhibit a specific purchase behavior. Using an audience of verified purchasers guarantees that you are targeting households that actually buy the advertised product, brand or category. This can make a tremendous difference in your targeting efforts and your campaign outcomes.

Also, because of the nature of deterministic audiences, marketers can open the door to completely new possibilities. For example, it’s just as important to consider the recency of the consumer’s last purchase of your brand as it is to consider their lifestyle segment. Think about the opportunities for a weekly purchase-based audience to reach consumers that are in various stages of their purchase cycle. For those who recently purchased, a message thanking them and reinforcing the intelligence of their purchase builds equity, while an expiring coupon may upset the customer. However, for those in the last quarter of the purchase cycle, a coupon may be most relevant. For customers that skipped one purchase cycle, a win-back message would be most relevant. Then there are those who have missed multiple purchase cycles – a different offer or message can be served to them. Verified / 100% deterministic audiences are another way to enable more targeting precision than a modeled audience because of the fundamental nature of how verified audiences are created. 

Data science will always play a valuable role but when data at scale is available, marketers can leverage both 100% deterministic and probabilistic audiences to help them target audiences more precisely and achieve better campaign results.

For more information on improving your audience targeting, contact me at

SHARE: LinkedIn - Twitter - Facebook