Inferences have uses. I’ll always try to target marketing based on user segments that are deterministic and I’ve read how the data is collected. If I can’t find that, I’ll look for modeled/etc audiences that include inferred data points. It’s likely discord would have targeting/data for self-reported demographics at a smaller scale and higher price point. Then the inferred demographics at a larger scale and lower price point.
That means your strategy is to make some assumptions about where your marketing will be well received, and insist on only showing it there.
This limits you to the quality of your assumptions. It's always going to be better to show your marketing where it actually works than to show it where you believe it works.
Well yeah you start with an educated guess and optimize once it’s live. So budget/etc moves fluidly and if discord audience 1 performs well it gets more budget. DSP I use doesn’t have discord inventory (if it exists?) but pretty much everything is optimized by deep learning outside of moving branding budgets around.
