How look-alike audiences are built on social media
Audience look-alike technology is used for targeted advertising. To do this, the software analyses users by comparing them to the target audience’s characteristics. The result is people who are as similar as possible to the representatives of the target audience. And the selection can be made according to various criteria, such as age, place of residence, interests, etc.
An important point when applying the look-alike mechanism is to choose a suitable initial base for analyzing accounts. For this purpose, users are divided according to basic factors:
– presence of interactions with the promotional post;
– presence of clicks on the link;
– activity on the site;
– registration on the site;
– presence of purchase.
For each of these points, a respective account base is created. A look-alike audience is formed based on these accounts. First and foremost, they focus on the users closest to the funnel to take the desired target action. A look-alike audience will be most effective for them.
Neural networks do all of the look-alike search work, but it is essential to understand that training is needed to improve their skills. Algorithms do everything themselves, but they need data for such a process. For example, Facebook and Instagram can be trained on databases of any size to improve targeting. Still, experts recommend that the number of users here should be more than a thousand accounts. However, it is important to understand that the database should include a more significant number, as a thousand people are the value the neural network will produce after processing all the data.
A list of email addresses or phone numbers must be uploaded to the social network’s advertising cabinet to determine a look-alike audience. The algorithms then compare this data with their own information. And if the contacts are in the social network’s database, then the account is added to the new database. The services do not save user information – once processed, they are immediately deleted. In this way, social networks adhere to a policy of protecting confidential data. In addition, the compiled new base cannot be unloaded from the account to minimize the risk of users’ personal information leakage.
After that, they create a look-alike audience, setting a certain similarity percentage with the initial TA. In this case, it is possible to experiment by testing different variants on the same characteristics. The higher the similarity percentage, the smaller the base will be.
A look-alike would be an ideal tool for a business that wants to reach a large audience of users interested in buying but, as yet, needs more consumer data. In such a case, a look-alike audience would increase user reach by targeting specific characteristics of potential customers.