As you may notice, it's sometimes not so straightforward to craft a well-defined audience that is both relevant and meets your expectations in terms of insights and audience understanding. That's why we've compiled the key best practices to consider when creating an audience based on criteria (excluding audiences created from custom or lookalike audiences).
​
📌 Please remember, these guidelines are general best practices. The most crucial factor in defining your audience is understanding exactly what you want to analyze and being critical of your approach. Not all audience definitions will adhere to these rules, but they serve as a helpful framework for most cases.

Let's begin 🤓

Ideally we want you to have 2 or 3 levels (or more). The idea is to create several levels that filter the audience as in a funnel, so as to keep only those people who are most relevant to the final audience.

Here's an example of how to define a two-level audience to profile people interested in Kim Kardashian. We can read the following as 👉 People having an interest in Celebrity (entertainment & media) AND Kim Kardashian

Here's a variation: you still have 2 levels, but you add a variable to the first level.
👉 People having an interest in Celebrity (entertainment & media) OR Internet Celebrity (interest) AND Kim Kardashian

Now, let's show you an example of a 3 level audience definition. Here, the idea remains the same, but we're going one level further. If we take a brand as an example, we'd have:

Let's see what it looks like in SOPRISM:

👉 People having an interest in Beverages (food & drinks) AND Soft drinks (nonalcoholic beverage) AND Sprite (soft drink) OR Coca-Cola Zero (soft drink)

For job titles (declarative data), demographics and behaviors, it is not necessary to add a second or third level. Other levels are usually used when you need a specific brand or field to add to these criteria. 👇

Ideally, the audience should be at least 100,000 people. However, there are exceptions where you can launch a project with a smaller audience, keeping in mind that the data may be interpreted differently.

It's also important to note that an audience can be too large. For example, if you target nearly the entire population, excluding only a very specific group (like people interested in horses), you might end up with data that's not very useful, as it effectively represents the general population.

👉 Here's a table summarizing the minimum audience sizes to be respected, depending on your objectives.

Ensure that each level include criterion of comparable size, yet if the interest criteria significantly aligns with the definition, allocate it separately.
​
To give an example, consider the scenario where one level includes Health & wellness (personal care) interest, encompassing millions worldwide, while another level comprises Nursery (children & parenting) potentially involving only around 2M individuals globally. 👉 Placing Nursery alongside the vast category of Health & wellness proves futile due to the significant disparity in scale.
​
Instead, it is more effective to allocate Health & wellness on its own level, given its substantial size. If Nursery hold particular significance, it should be grouped on a separate level with similarly sized interests.

When utilizing exclusions, exercise caution. Interests are intricately linked, forming a web of connections. By excluding a single brand, you risk inadvertently excluding a multitude of related interests. It's advisable to proceed with exclusions only when we are completely confident that the excluded entity is not relevant to our audience.

Sometimes, the specific brand, person or specific criteria you want to target might not be available in META's targeting options. Here are some strategies to help you effectively define your audience in such cases:

To take things a step further, you can also combine these strategies to create an overview of this unknown audience.