Knowledge Base

Sending Audiences to The Trade Desk Using Lookalike Modeling


This article provides a guide for Narrative clients on how to utilize the Trade Desk's (TTD) audience modeling process, specifically focusing on lookalike modeling. This capability is crucial for clients looking to expand their reach by targeting new but similar audiences based on their existing customer data.

Overview of the Audience Modeling Process

The Trade Desk utilizes a structured approach to audience modeling, which involves the following key steps:

1. Seed Audience

The seed is the foundational dataset from which the lookalike (LAL) model is developed. It represents a segment of your current users or customers that exhibit desirable traits or behaviors you wish to replicate in new prospects.


  • The seed must contain at least 10,000 IDs to ensure effective modeling.
  • If the seed contains fewer than 10,000 IDs, a cross-device (XD) enhancement is necessary before modeling to expand the dataset adequately.

2. Cross-Device Application

The Cross-Device method is used to increase the seed size by linking multiple devices and identifiers belonging to the same user. This step is crucial when the initial seed size is insufficient for effective modeling.


  • Apply Cross-Device technology to the seed data to enrich and expand the dataset.
  • This enrichment allows for a more robust model by incorporating a broader spectrum of user behaviors and characteristics.

3. Lookalike (LAL) Modeling

Once the seed is established and enhanced, TTD's Lookalike Modeling process begins. This model uses the characteristics of the seed audience to find new users with similar profiles.

Modeling specifics:

  • The typical modeled audience size is 10-15 times the size of the seed, but it can be adjusted based on specific campaign needs or target audience size requirements.
  • Modeled audiences should be shared with a specific TTD advertiser seat before they are available in TTD's marketplace.

4. Modeled Segments

Post-modeling, two main types of segments are generated:

  • Modeled segment with seed: Includes the original seed audience plus new, similar users identified through the LAL model.
  • Modeled segment without seed: Consists only of new users identified as similar to the seed audience, excluding the original seed data.

Narrative clients can list either audience in TTD's marketplace.


Using The Trade Desk's sophisticated lookalike modeling process allows Narrative clients to effectively broaden their audience base by targeting individuals who resemble their best customers. By following the outlined steps and recommendations, clients can enhance their advertising strategies and achieve better engagement rates and return on investment.

< Back

Hi! I’m Rosetta, your big data assistant. Ask me anything! If you want to talk to one of our wonderful human team members, let me know! I can schedule a call for you.