Campaign Objective

Angler AI activation offers predictive custom audiences as well as predictive seed for platform Look-a-Like audiences geared towards finding high propensity audience for the advertiser.

Predictive Custom Audience

Predictive custom audiences can be generated and distributed to ad platforms with these 3 simple steps:

Create seed : identify the ideal current or past customers that describe the campaign objective
Build model: using the seed, trigger a bespoke AI model development
Audience distribution: when the AI-model is done, identify top slices of audiences and distribute to desired ad platform as custom audience. Angler does distribution for you and keep refreshing audiences post-distribution as underlying data conditions change.

Create Seed

Seed can be constructed with Boolean expressions (AND, OR) of brand's 1st party (orders, customers, products etc.), zero party (survey, quizzes etc.), as well as Angler attributes about those customers, such as demographics (household income, age, gender, household composition), purchase insights (over or under index of spending with a set of brands). In the below example, the seed is constructed with the following conditions:

customers acquired during BFCM/holiday periods of FY 2021 or 2022
received at least 15% discount on their 1st order
having household income at least $100K/year

Once the seed is defined, give the seed (list) a name and save it. Now you are ready to build AI-models for high propensity audience using this seed.

Build Model

Select Models from the left hand pane, and then + Create AI Model to start the workflow.

Select model type as Prospecting and then hit Next.

In the next screen set model objective. For high propensity audiences one of the three objectives may be selected:

Orders : where objective is to minimize cost per order (CPO) generated from high propensity audiences. The audience will include high propensity current customers who are likely to purchase again;
New Customers: where the objective is to acquire new customers only, and optimizing new customer CAC;
Order Value: where the objective is to find high propensity customers expecting to make higher than the average first order value (AOV)

In the next screen, select the right seed that you would like the model to use. In our workflow here, we are selecting the seed that we've defined above, called BFCM21_22_new_cust_disc.

Confirm configurations in this final screen and hit Start button. This will trigger the workflow of AI model feature extraction, training and scoring.

The model run takes between 4-6 hours, based on the size of the seed file, once the scoring of entire 251M+ graph is complete, you will receive an automated email from Angler. Come back to the models section and you will find completed model cards.

Model status page showing all 4 steps of AI-model building are completed with the confirmation of the seed that being used for the model. This page also provides status update when the model is being built, with the current processing step and an estimated time remaining.

Lastly, AI-model resolution is displayed in the view results section. In the chart below coverage of all new customers acquired from the seed list that covered from the 5% increments of propensity scores are displayed. In this particular model, about 28% of seeds are covered by top 5% of prospecting audiences, each 1% audience slice equates to 2.51M US 18+ adults.

Audience Distribution

Now you are in the last section of distributing custom audiences to ad platform using the AI-model we've just built. Select Audiences panel and then click on + Create an audience

Select AI Extension option, and fill out Audience details and Name in Destination. Please note that name in destination is how the audience slices will show up at your destination (e.g. Meta Audience section). Due to string length restrictions in ad platforms, limit name in destination within 25 characters.

Select AI model that you would like to use for distribution, in our workflow we're selecting the BFCM21_22_new_cust_disc model for distribution.

In the selection criteria section make the following selections:

size of the audiences: each 1% slice is 2.51M adults (18+) pre-match. You can select upto top 10%. In the example below we are selecting 95+, i.e. top 5% propensity audiences
the checkbox, when ticked, exports bottom 20% of audiences (low propensity audiences) for negative targeting or suppression from broad targeting.

In the destination section, decide destination (in this example, we are selecting Meta), and distribution of audiences with experiment (80%) and holdout (20%), and suppression (bottom 20%). You are also able to change the slices in the custom experiment option.

Distribution will start and a status bar will appear in the blank space below.

When distribution is complete, status will be updated appropriately. At this point, Angler has successfully sent custom audiences to your destination and such audience is ready for activation.

Audiences will show up in Meta's audience section, as shown below.

Platform Look-a-Like (LAL) with Angler Predictive Seed

You can also leverage Angler platform to create, distribute and refresh predictive seed to ad platforms, currently available for Meta, Google Ads, and Tiktok.

Creation of Seed

The seed can be of two types:

simple seed: where the seed is created by Boolean expression (AND, OR) of client's first part, zero party data, in addition to Angler's demographics segments and Angler's modeled traits (e.g. customer pLTV >= $100)
value based seed: where Angler appends future value of every seed for platform's value-based LAL functionality. The value we append is the 90-day repurchase revenue of the customer in the next 90-days. This allows ad platforms to find prospects who have higher expected future value.

customers acquired during BFCM/holiday periods of FY 2021 or 2022
received at least 15% discount on their 1st order
having household income at least $100K/year

Distribution of Seed

Distribution of seed starts by clicking + New Distribution button

Select connection and ad account where you would like Angler to distribution the seed. For the value-based LAL audiences select the predictive LTV scoring model, in case you have multiple active models. We recommend enhanced matching option for pre-boarding (improved matching) at certain destination like Tiktok. Distribution starts when you enter Start.

Upon successful distribution status is updated as below.

For every seed that Angler distributes, there are 3 platform LAL audiences are programmatically created by default, at 1%, 3% and 5% respectively.