We'd obviously love every dollar of paid search we spend to convert, but that's not realistic and at the same time we need to explore adding new keywords to the mix. What % of the overall portfolio of spend that doesn't convert is typical/healthy? How is this different for text ads vs. PLAs?
PLAs have the advantage of being able to change the taxonomy of the campaign based on available data. As long as you're hitting your ROAS goal, and are adjusting your product group tree to make sound decisions, this should be fairly self-correcting.
Text ads, however, are a bit of a different beast on this front.
The most difficult scenario here is a very large catalog of products that individually have very little traffic, but add up to a lot of business. If you're trying to target them individually with text ads (say, in some automated way, for example), then you're going to have to use huge amounts of time to make good decisions, and if 80% of those bids end up not paying off, you're in trouble. Therefore, I recommend starting with the highest converting ads (low-funnel, high popularity, etc.). Get these producing a reliable ROAS.
From there, start moving toward the mid-tail, keeping an eye on your account's overall Ad spend.
How well your primary ads do versus your targets will determine how much you have to experiment with. For example, if your highest converting products are running at 10% under your COS limit once optimized, you have that 10% to work with. If your highest converting ads are running right at your COS target, however, then you won't be able to hit your COS target with anything that converts any lower.
Then, if you have a specific opportunity you want to pursue, without going over budget, you have to cool off your bids on your core ads a shred, giving up market share on those, to afford to explore for new potential core ads.
Have questions? Drop me a line via the Contact form--I love to talk shop.