Recent upgrades to Audience Predictor aim to solve two primary challenges for marketers

Our automated lookalike model, Audience Predictor, has undergone recent upgrades to offer advertisers a more cost-effective and sophisticated way to precisely reach the right audience.

The most impactful advertising starts with your consumers and uses the power of data to deliver better ad experiences across the customer journey. Key to this is the effective application of data to every ad impression. As such, we are excited to announce that we have upgraded Audience Predictor, with the goal of delivering the precision clients need to drive performance.

Audience Predictor will now use a new decisioning logic, called multi-element bidding, which offers advertisers a better way to support targeting precision while also balancing scale. What we are moving towards is an innovative approach to apply and buy data. One which enables advertisers to layer in more relevant data to make smarter bidding decisions, while proportionally paying multiple data providers based on the relevance and value of data creates.

Our aim is to help solve the two primary challenges marketers face when creating audiences:

  1. Choosing data segments without understanding whether they will improve performance. This can result in using data segments that add cost but not value.
  2. Layering multiple data elements onto a single impression can be cost prohibitive and can limit scale. This can result in marketers using less data in decisioning.

For example - It makes sense that a luxury car ad will resonate more with a person who is in-market for a new car “and” currently own a luxury car “and” has a household income of $200K. However, grouping segments with “and” logic adds a cumulative data cost for every data segment you use, it also limits scale as every data element must be present to bid on an impression.

By implementing multi-element bidding within Audience Predictor, we are offering a holistic solution that moves past the limitations of “and/or” logic which can offer a more cost-effective way to add targeting precision by prioritizing bidding on impressions with the most relevant data segments present.

How does it work?
Previously, Audience Predictor would identify data segments that share traits with your existing seed audience, but instead of looking for impressions where multiple relevant data elements were present, only one of the identified segments needed to be present to bid. This was due to limitations in the data buying process.

In the example above, Audience Predictor identifies three relevant segments based on the traits found in the Test Drivers seed audience. Audience Predictor would then bid on any impressions where one of identified segments were present.

With Multi-Element Bidding, Audience Predictor will now evaluate all the available data present for each impression and prioritize spending on impressions that match multiple data elements based on their relevance to the 1st party seed data and value to the campaign goals. This upgraded approach aims to balance targeting precision with scale by using more data in decisioning, while also allowing the flexibility to bid on impressions with any combination of relevant data elements present.

In the example above, Audience Predictor will evaluate all data-segments on each impression and prioritize bidding on impressions where the most relevant data elements are present.

The goal is to provide an audience-targeting strategy where every data segment adds value, and every impression is relevant and contributes to your goals.

Trader Recommendations

With Audience Predictor advertisers can use their data to make better decisions, while also working to achieve the right balance of precision and scale for awareness, consideration, and conversion campaigns.

For example — imagine you are running a full-funnel campaign for an online retailer:

  1. Awareness Audience Predictor strategy: Consider choosing your Universal Pixel’s default mapping as the seed to create a scalable audience that will target users more likely to want to learn about your brand. A good KPI could be tracking site visit CPA in combination with reach.
  2. Consideration Audience Predictor strategy: Consider choosing a product-level Universal Pixel mapping as the seed to create precise audience-targeting strategy that reaches users most likely to spend time navigating the brand’s website and learning more about specific products. A good KPI could be tracking CPA for specific product pages and add-to-cart mappings.
  3. Conversion Audience Predictor strategy: Consider choosing a checkout confirmation Universal Pixel mapping as the seed to create a performance-driven audience-targeting strategy that reaches users more likely to make a purchase. A good KPI could be tracking purchase confirmation CPA and/or ROAS.

These strategies can be easily activated in the Solimar Campaign Wizard or by creating new Audience Predictor ad groups in existing campaigns.

Looking towards the future
Today, Multi-Element Bidding is available in Audience Predictor, and our aim is to expand this solution to the rest of our audiences over time. By creating a model that bids on impressions where the most relevant data elements are present, we are working to help advertisers deliver more relevant messages to consumers, with the goal of better performance and more data being used overall. Advertisers can become more effective, and data providers should be incentivized to create more relevant segments as they are rewarded by the value they create.

Reach out to your Trade Desk account manager to find out more.