- What were the qualities that contributed to such a successful model?
- How can more marketers take advantage of data onboarding to choose the best audience data and seed segments?
Criteria for Seed Selection
Marketers frequently use offline data criteria to define high-performance seed audiences for look-alike modeling.
Three top criteria for seed selection are:
- Average Order Value
- Purchase Frequency
- Lifetime Value (LTV)
Among LiveRamp’s partners, Quantcast has found high LTV to be a strong performance indicator, but the business type, audience sizes, and order values are also important contributing factors.
According to Andy Huson at Quantcast, “It really depends on the business. High LTV is probably the best indicator, but they may have small segments that have a higher margin at less scale or other segments that are an important fit within their company strategy.”
Similarly, Erica DePalma, VP Digital Marketing at Media Horizons, shares that they found value in using all three selection types for Schwan’s, who defined high- value customers through a combination of high average order values, purchase frequency, and lifetime value.
Many of our partners stress the value of taking time with the client to determine campaign specifics and context.
Quantcast recommends that clients upload new data every month to ensure cookies stay fresh, and data sets stay relevant.
Fresh data leads to better match rates and better performing campaigns. Quantcast also works with clients to ensure activation of their data aligns with their goals as lookalike modeling is just one use case of onboarding data from LiveRamp.
DePalma shares that Media Horizons works with clients “to ensure that only cleansed data with proper segmentation is used.”
Similarly, Donna Hamilton, VP Sales and Business Development at Alliant, shares that they “assign their own data scientists to work with clients to identify the business objective for each campaign and to help determine the audience that should be used for model development.”
LiveRamp can onboard models from any data provider and has an extensive set of partnerships in place.
LiveRamp can also onboard seed data for online look-alike modeling into any major DMP as well as a broad set of online data specialists, including AddThis, Connexity, Dstillery, Mentad, NetSeer, Quantcast, and Rocketfuel.
To navigate this sea of options, look for areas in which look-alike modeling vendors have a particular strength or specialty.
Many 3rd party data vendors are focused on a certain vertical or channel for which they perform best.
“Quantcast, for example, may be a good fit for modeling for online purchasing data because they have comprehensive, real-time visibility into consumer behavior on web and mobile channels,” Huson shares.
Vendors such as Datamyx and Equifax IXI have focused on sending LiveRamp financial performance data for anonymized modeling.
According to Donna Hamilton, “subscription services and insurance are two areas where Alliant Online Audiences have been particularly effective.”
The Proof is in the Pudding
Doing all of the above has proven results.
At Schwan’s, DePalma says, “New customer orders, as a percentage of total orders, increased over 100% for nearly all display programs. Return on Ad Spend also doubled from 39% to 78%.”
In addition to this, Schwan’s custom audience campaign on social media “showed new customer acquisition costs that were 48% lower than interest/behavioral targeting.”
Similarly, Hamilton shares that Alliant found success with a top multinational CPG on their delivery subscribed service, who used onboarding for modeling resulting in a 36% lower CPA than projected, and 14% less CPA than Target Access Digital efforts.
Amazing Results Come through Extensive Preparation
Data onboarding, targeting, and look-alike modeling don’t happen automatically.
With the right seed criteria, necessary data preparation, and the right vendor for your use case, you can expect real returns.
Interested in learning more?
Learn more about the process by downloading our look-alike modeling product sheet. (Download PDF)