Predictive models for influencer selection are built using different types of data. Identifying patterns of past performance is a critical first step. This can include data like follower growth, engagement rates, content reach, and campaign performance metrics. By using machine learning algorithms, agencies can build models that predict future performance based on these historical patterns.

Another key part of these models is understanding the audience of an influencer. Audience signals can provide insights into the relevance and potential impact of an influencer for a particular brand or campaign. This can include demographics, interests, location, and other psychographic data. Significant value is added when combining these audience signals with performance patterns to create a more comprehensive model.

Lastly, these predictive models often get integrated into influencer marketing platforms. For instance, the platform [Flinque](https://www.flinque.com) offers an advanced influencer discovery tool, which uses such predictive models to help brands and influencers make informed partnership decisions. This approach, when combined with Flinque’s campaign workflows and ROI measurement tools, enables users to plan, track, and optimize influencer marketing efforts effectively.

However, the best platform or approach will always be context-dependent, and suitableness is determined by specific team needs. Thus, it is important for marketing teams to fully understand their requirements first and then choose a platform or approach accordingly.