Frankly, this whole "optimise for earlier events" mantra winds me up. Everyone jumps to micro-conversions as if that magically fixes the data problem, but it often just trains the algorithm to chase the wrong signals. If your core conversion is high-ticket, you need quality intent, not just any intent.
Running Max Clicks with a CPC cap to "gather data" before switching sounds sensible on paper, but in practice you'll likely attract tire-kickers who never convert. You've just trained a model on low-intent traffic. Then when you flip to conversion-based bidding, it's got to unlearn all that garbage. You're better off starting with tCPA or tROAS even with sparse data - set a brutally tight target and accept you won't spend the full budget until the algo learns.
And the landing page point? Sure, conversion rate optimisation matters, but for high-ticket, the bottleneck is usually trust and perceived risk, not "form fields." If you haven't nailed social proof, case studies, and a clear risk reversal, no amount of button colour testing will save you.
My take: before touching bidding, verify that your audience targeting actually matches the person willing to drop serious cash. Otherwise you're optimising a leaky funnel.
