Ask a CRO team what’s on its testing roadmap and you hear about headline variants, button colors, and checkout field order. Ask where page speed fits and you get a shrug. Speed is “an engineering thing,” handled separately if handled at all. Teams keep that separation out of habit, and it leaves a real lever untested.

Why speed lands under “engineering” instead of “CRO”

Most CRO tools center on visible, testable variants: two headlines, two button colors, split traffic, measure the winner. Speed doesn’t fit that workflow. It resists isolation as a clean A/B variant, it touches infrastructure instead of page content, and a different team with a different backlog owns it. CRO teams drop it from the conversation, even though the causal chain (faster page, less friction, more completed actions) is what CRO exists to find.

Speed often has a larger, more consistent effect than the copy and layout tests that do make the roadmap. A headline test might move conversion 2-3% on a good day, and the result sometimes fails to replicate. A meaningful LCP improvement in a typical Foresight deployment moves conversion around 29%. A slow page adds friction for every visitor, while a test variant reaches only a fraction of them.

What testing speed looks like

You can’t A/B test speed the way you test a headline, because you wouldn’t want half your visitors to get a slower experience to prove a point. Use the comparison from our piece on turning milliseconds into revenue instead: compare the conversion rate of sessions that loaded fast with sessions that loaded slow, on your own traffic, with no synthetic split. It falls short of a controlled experiment in the strict sense. It does rest on real behavior, and one query often settles whether speed deserves priority.

Where predictive preloading fits into a CRO practice

Your team probably knows speed matters. “Make the whole site faster” has no obvious owner and no clean test, so it loses to changes that have both.

Foresight fills that gap. It adds a preloading layer that improves perceived speed on the pages your visitors already navigate through, and you measure it with the same before/after rigor as any other test, with no infrastructure tickets from your CRO team. It is a conversion lever in the performance layer instead of the content layer.

If your CRO roadmap has no line item for speed, an audit will show whether it should.