AB Split Test Version 2.6: Hide and Move Elements, High-Speed Full-Page Tests, and Full Developer Control

🆕 Hide and Move Elements in Magic Mode

The problem: Magic Mode let you swap text and images, but layout was off limits. If you wanted to test moving a testimonial above the fold, or hiding a secondary CTA to reduce friction, you had to rebuild the page in your page builder, create a duplicate, and set up a full-page test.

The solution: Magic Mode now lets you hide any element or move it to a different position on the page, with a live drag preview. No code. No page builder.

Tests you can now run directly in Magic Mode:

• Move a testimonial above the fold to see if social proof earlier in the page lifts conversions
• Hide a secondary CTA to test whether it is creating friction or helping
• Test whether a form performs better above or below a feature list
• Relocate a hero sub-headline below a CTA block without touching any code

---

🆕 Full-Page High-Speed Mode

The problem: Full-page split tests redirect visitors to the correct variation URL. On slower connections or cached pages, there was a brief flash of the original page before the redirect fired.

The solution: High-Speed Mode fires the redirect the instant the page begins loading, rather than waiting for the full page to parse. Visitors land directly on the right variation with no flicker.

No setup required. It applies automatically to all full-page tests once you update.

---

🆕 Mark Complete and Restart

The problem: Ending a running test or reopening a completed one meant navigating into the test settings, making changes, and saving. A few too many clicks for something you do regularly.

The solution: Both actions are now available directly from the test panel.

• Mark Complete ends the test immediately. You can optionally forward all traffic to the winner at the same time.
• Restart reopens a completed test with its existing data intact, so results continue accumulating without starting from zero.

---

🆕 Every Plugin Setting via REST, WP-CLI, and MCP

The problem: Plugin settings could only be changed inside the WordPress dashboard. For developers, agencies, and anyone building automated workflows, that was a blocker.

The solution: Every plugin setting can now be read and updated programmatically. Secret keys are kept write-only to protect sensitive values.

What this makes possible:

• Connect Claude or another AI assistant directly to your AB Split Test settings panel via Model Context Protocol
• Fetch performance statistics, update configurations, or trigger weekly report routines from a terminal
• Automate test creation and settings management across multiple client sites without opening WordPress

We introduced MCP support in v2.5.0. Version 2.6 extends that access to settings, not just test management.

---

🆕 Meta Pixel and Matomo Tracking

The problem: Variation views and conversion events were not flowing into Meta Pixel or Matomo, so users running Meta ads or using Matomo as their primary analytics platform had a blind spot.

The solution: Both platforms now receive variation and conversion data automatically, alongside the other analytics platforms AB Split Test already supports (GA4, GTM, Plausible, Fathom, Mixpanel, Usermaven, Umami, Cabin Analytics).

No custom code required.

---

🆕 CallTrackingMetrics Support

The problem: For businesses where the primary conversion is a phone call, not a form submission or purchase, there was no clean way to attribute calls to the right test variation.

The solution: Phone calls tracked through CallTrackingMetrics can now count as conversions and be attributed to the correct variation automatically.

Useful for:

• Service businesses where calls are the primary conversion goal
• Agencies needing proper ROI reporting on phone-first funnels
• High-ticket offers where visitors call before they buy

---

Faster, More Reliable Tracking

Tracking has been moved to a lightweight first-party endpoint with a queue. Beacons fire faster, concurrent visitors on high-traffic tests can no longer drop or double-count events, and results still appear in real time when you open a report.

Two accuracy fixes shipped alongside this:

Full-page control visits were being undercounted. First-time visitors who stayed on the control page were not being counted at all, which was skewing results toward variations. Fixed.

WooCommerce conversion leak. A checkout edge case was recording conversions under the wrong or blank variation for tests the buyer was not enrolled in. Fixed.