feat: A/B/C test for homepage 2026 redesign

[pettinarip]

Summary

• Adds A/B/C testing infrastructure for the homepage 2026 redesign
• Variant A: Original homepage (control)
• Variant B: Redesigned homepage with persona modal flow
• Variant C: Redesigned homepage with direct action buttons (no modal)
• Uses dynamic site URL detection for proper tracking in branch/preview deploys
• Polishes hero CTA buttons to match Figma design specifications

Test plan

• Verify A/B/C test panel appears in development/preview environments
• Test variant switching works correctly via debug panel
• Confirm Matomo tracking fires for each variant
• Check dynamic URL detection works on Netlify preview deploys
• Verify hero CTA button styling matches Figma design

[netlify]

✅ Deploy Preview for ethereumorg ready!

Name	Link
🔨 Latest commit	a31cb08
🔍 Latest deploy log	https://app.netlify.com/projects/ethereumorg/deploys/69933ccdaa5d532c47970453
😎 Deploy Preview	https://deploy-preview-17294.ethereum.it
📱 Preview on mobile	Toggle QR Code... Use your smartphone camera to open QR code link.
Lighthouse	7 paths audited Performance: 49 (🔴 down 6 from production) Accessibility: 94 (no change from production) Best Practices: 100 (no change from production) SEO: 100 (no change from production) PWA: 59 (no change from production) View the detailed breakdown and full score reports

---

To edit notification comments on pull requests, go to your Netlify project configuration.

[konopkja]

SimulatorSection: eventCategory prop is accepted but never used

In src/components/Homepage/SimulatorSection/index.tsx, the eventCategory prop is defined in the type but never destructured or passed to child components:

type SimulatorSectionProps = {
  className?: string
  eventCategory?: string  // ← defined here
}

const SimulatorSection = ({ className }: SimulatorSectionProps) => {
//                          ^ eventCategory not destructured

This means the simulator's internal interactions (step navigation buttons, finalCtaLink at the end of the walkthrough) have no event tracking from the homepage context. The TrackedSection wrapper catches the section_view, but individual clicks inside the simulator are invisible to analytics.

Not a blocker for launch, but worth fixing so we can measure simulator engagement across the A/B/C variations.

[konopkja]

SEO Blocker: Bots must always receive the Original variant

@pettinarip — heads up, this needs to be addressed before we activate the experiment.

Problem

The current assignment logic in src/lib/ab-testing/server.ts hashes Googlebot like any regular visitor. Since Googlebot has a consistent IP + User-Agent, it will deterministically land on one specific variant for the entire test.

The redesigned homepage has substantially different content — different headings, different sections (KPI, Savings Carousel, Feature Cards vs the current What is Ethereum, Activity, Builders, etc.). If Googlebot is assigned to a redesign variant:

• Google re-indexes the homepage with completely different content
• Search snippets change
• Temporary ranking fluctuation during test
• Content changes again when the test ends → double disruption

Fix

Add bot detection at the top of getABTestAssignment() in src/lib/ab-testing/server.ts. This is not cloaking — Google's A/B testing guidelines explicitly allow serving the original to bots during experiments.

const BOT_PATTERN = /googlebot|bingbot|yandex|baiduspider|duckduckbot|slurp/i

export const getABTestAssignment = async (
  testKey: string
): Promise<ABTestAssignment | null> => {
  const configs = await getABTestConfigs()
  const testConfig = configs[testKey]

  if (!testConfig || !testConfig.enabled) return null

  const headers = await import("next/headers").then((m) => m.headers())
  const userAgent = headers.get("user-agent") || ""

  // Always serve Original to search engine bots to prevent
  // indexing fluctuation during A/B tests. Not cloaking per Google guidelines.
  if (BOT_PATTERN.test(userAgent)) return null

  // ... rest of fingerprint + assignment logic unchanged ...
}

Returning null triggers the existing fallback in ABTestWrapper which renders variants[0] (Original). No other changes needed.

What's already fine

• ✅ Canonical URL is the same for all variants (https://ethereum.org/en/)
• ✅ No separate URLs per variant
• ✅ hreflang tags unchanged
• ✅ JSON-LD rendered outside ABTestWrapper (same for all)

This is the only remaining SEO concern — everything else looks good.

[pettinarip]

@konopkja good catch, pushed the patch. Also extend it to catch social engines to deliver a consistent OG preview when sharing the link.

[konopkja]

⚠️ getAccountHolders() failure can crash the homepage for all variants

The new getAccountHolders() call was added to the main Promise.all in page.tsx, and if it fails (Dune API timeout, rate limit, outage), the page throws:

if (!accountHolders || "error" in accountHolders) {
  throw new Error("Failed to fetch account holders data")
}

This is on the critical path for every homepage visitor, including Variant A (Original) which doesn't use this data at all. A Dune API outage during the test would take down the entire homepage — not just the redesign variants.

Suggestion: Degrade gracefully instead of throwing. The KPI counters in the redesign can show a fallback (e.g. 0 or a static cached value) while the Original variant continues working unaffected:

// Instead of throwing:
const accountHoldersValue =
  accountHolders && !("error" in accountHolders) ? accountHolders.value : 0
const transactionsToday =
  growThePieData && "value" in growThePieData.txCount
    ? growThePieData.txCount.value
    : 0

This way, if Dune is down, the redesign variants show 0 for the KPI counters (or you could hide them) but the page still renders for all three variants.

[wackerow]

Security Review

PR: feat: A/B/C test for homepage 2026 redesign (#17294)
Reviewed: 2026-02-17
Risk: LOW

Findings

[LOW] Matomo analytics enabled on preview deploys

File: src/components/Matomo.tsx:10-18, app/api/ab-config/route.ts:36
Issue: The IS_PREVIEW_DEPLOY guard was removed from Matomo initialization and the AB config route. Preview and branch deploys will now send analytics events to the production Matomo instance, potentially polluting production data with test traffic.
Recommendation: Confirm this is intentional for A/B testing on preview deploys. If preview analytics should be isolated, consider a separate Matomo site ID for preview environments.

---

Areas Reviewed

• API Route Security
• SSRF & External Fetching
• XSS & Content Injection
• Security Headers & Transport
• Secrets & Configuration

Not applicable: GitHub Workflow Security, Dependencies & Supply Chain

Limitations

• Cannot verify runtime values of Netlify environment variables (DEPLOY_PRIME_URL, DEPLOY_URL, URL) used in the new SITE_URL fallback chain.
• The home-redesign base branch was not fully reviewed — only changes introduced by this PR were evaluated.

---

🤖 Review performed by Claude Opus 4.6 (claude-opus-4-6)

[pettinarip]

Security Review

PR: feat: A/B/C test for homepage 2026 redesign (#17294) Reviewed: 2026-02-17 Risk: LOW

Findings

[LOW] Matomo analytics enabled on preview deploys

File: src/components/Matomo.tsx:10-18, app/api/ab-config/route.ts:36 Issue: The IS_PREVIEW_DEPLOY guard was removed from Matomo initialization and the AB config route. Preview and branch deploys will now send analytics events to the production Matomo instance, potentially polluting production data with test traffic. Recommendation: Confirm this is intentional for A/B testing on preview deploys. If preview analytics should be isolated, consider a separate Matomo site ID for preview environments.

Areas Reviewed

• API Route Security
• SSRF & External Fetching
• XSS & Content Injection
• Security Headers & Transport
• Secrets & Configuration

Not applicable: GitHub Workflow Security, Dependencies & Supply Chain

Limitations

• Cannot verify runtime values of Netlify environment variables (DEPLOY_PRIME_URL, DEPLOY_URL, URL) used in the new SITE_URL fallback chain.
• The home-redesign base branch was not fully reviewed — only changes introduced by this PR were evaluated.

🤖 Review performed by Claude Opus 4.6 (claude-opus-4-6)

we are using a different Matomo site id for non-prod envs.