Really interesting, as always, Paul! One thing I’d add is that excluding the lookalike audience (a very large audience) from the broad audience may throw off the results. If you only ran the broad audience, allowing it to target the people who also exist in the lookalike audience, how would it perform? So, I’d be curious in either just the broad audience or using a true A/B split test so that you don’t remove what may be the most valuable subset from the broad audience.
I’ve actually never targeted broad audiences, but like you, I’ve also heard of advertisers having success with it. I’ve also started using interest expansion lately, and I’m getting lots of good results there for the top of the funnel.
]]>