Performance work competes for engineering time against features that have named owners and revenue targets. Winning that argument is the actual skill.
Ranking impact is a weak argument on its own
Speed is a real but modest ranking factor, and framing a performance project purely as SEO invites a comparison against cheaper SEO work. It usually loses that comparison.
Conversion data is the argument that wins
Segment your own analytics by page load time and show conversion rate against it. On most sites the relationship is steep and undeniable, and it is your own data rather than an industry statistic anyone can dismiss.
Field data beats lab scores
A synthetic score is easy to argue with. Real user measurements from actual visitors on real devices and connections are not, and they usually tell a considerably worse story than the lab number the team has been optimising toward.
Scope it to the templates that matter
"Make the site faster" never gets prioritised. "Reduce largest contentful paint on the product template, which carries sixty per cent of revenue" gets scheduled. Narrow scope, named template, measurable target — that is what survives a planning meeting.