A very helpful article. One question nobody seems to be able to answer is how sophisticated the algorithm they use for lookalike audiences. I picture the Facebook data as a graph rather than a normal database. Each node being a person, thing, company, website, group, like, keywords in in comments and so it goes on.
Therefore it is quite easy to compare two people’s graphs using a mathematical function and see how similar they are and then place them in the lookalike audience.
Are Facebook under any obligation to disclose how/what data the algorithm uses?