docs: A/B test nav CTA copy (Comfy Cloud vs Try Cloud for free)

[comfyui-wiki]

What & why

Requested in #website-and-docs (thread): test whether homepage-style copy on the docs navbar CTA drives more clicks and signups.

PostHog data from the thread: the docs CTA ("Comfy Cloud" → comfy.org/cloud?utm_source=docs) gets ~1.2% click rate (7,637 clicks from 6,294 people against 540k unique visitors, 30d), but 27% of clickers create an account. The funnel suggests people aren't clicking because the CTA doesn't read as an offer, not because the destination fails.

What changed

nav-cta-ab-test.js (repo root; Mintlify inlines every root-level .js file on all pages, same mechanism as syft.js):

Self-contained variant assignment (no PostHog feature flag), per the thread discussion with Christian (inline Math.random() option) and given that docs' PostHog runs through Mintlify's proxy with an unverified flag-evaluation path:

• Each visitor is randomly assigned control ("Comfy Cloud") or test ("Try Cloud for free"), 50/50.
• The assignment is sticky per browser via localStorage, so the label does not flip between page loads (the naive inline version would re-randomize on every page).
• Events for analysis:
  • docs_nav_cta_exposure {variant, label} — once per page load
  • docs_nav_cta_clicked {variant, label, href} — on every CTA click (captures the actually-assigned variant)
• The href is identical in both variants to isolate the copy variable.
• A MutationObserver re-applies the label when the SPA re-renders the navbar; storage-blocked browsers fall back to a per-load random draw.

Trade-off (deliberate)

This does NOT use the PostHog experiment (#448700, flag docs-cta-copy, left in draft and unused): no experiment dashboard readout (significance testing, guardrails), no dashboard kill switch, and signup attribution is manual (join docs_nav_cta_* events to $signup by person). Analysis is done by filtering the two events above by variant. This keeps the hook ~100 lines, removes the flag-launch step and the proxy dependency, and the test is live immediately after deploy.

How to launch

Merge this PR → Mintlify deploys → the test is live. No PostHog dashboard step needed.

Verification

Tested in a real browser against the live rendered nav DOM (desktop #topbar-cta-button link + mobile nav link), with a mocked PostHog and controlled Math.random():

• test arm swaps both labels to "Try Cloud for free"; control arm keeps "Comfy Cloud".
• Sticky: reload with a different random draw keeps the stored variant.
• docs_nav_cta_exposure fires with the assigned variant; docs_nav_cta_clicked carries {variant, label, href}.
• Labels re-apply to late-rendered links (SPA navigation).

[mintlify]

Preview deployment for your docs. Learn more about Mintlify Previews.

Project	Status	Preview	Updated (UTC)
comfy	🟢 Ready	View Preview	Aug 26, 2026, 2:18 AM

💡 Tip: Enable Workflows to automatically generate PRs for you.

[coderabbitai]

Important

Draft PR not reviewed

Draft PRs are not automatically reviewed by default.

• Trigger a manual review

To automatically review draft PRs, update your CodeRabbit configuration:

reviews:
  auto_review:
    drafts: true

---

Thanks for using CodeRabbit! It's free for OSS, and your support helps us grow. If you like it, consider giving us a shout-out.

❤️ Share

• X
• Mastodon
• Reddit
• LinkedIn

_{Comment @coderabbitai help to get the list of available commands.}