ImproveAB: A/B Testing

FAQ

Is it really free?

Yes. The plugin is GPL-licensed and free forever, with no pageview or traffic limits. It runs complete A/B and A/B/n tests on its own, as many at once as you like, with every goal type and page targeting.

Does it work with my page builder?

Yes. Element tests target a CSS class, which every major builder lets you add. Build both versions of the element on the same page, then paste the class the plugin gives you onto each version (the first class on the original, the second on the new version). Where the field is:

• Block editor (Gutenberg), Kadence Blocks, GenerateBlocks, Spectra: select the block, open the Settings sidebar, expand Advanced, paste into “Additional CSS class(es)”.
• Elementor 4 (the default for new sites): select the element, paste the class into “Classes” on the Style tab. Elementor 3, and legacy widgets inside Elementor 4: open the Advanced tab, paste into “CSS Classes”. Both can appear on one page, so go by which tab the element offers.
• Bricks: select the element and type the class into the selector input in the left panel, just under the element’s name.
• Etch: select the element and add the class in its class / selector input.
• Divi 5: open the element settings, Advanced, open Attributes, choose “+ Add Attribute” and add one named “class”. Divi 4: Advanced, expand “CSS ID & Classes”, paste into “CSS Class” – those fields are gone in Divi 5.
• Beaver Builder: open the module (or row / column) settings, Advanced tab, paste into “Class” under HTML Element – not the CSS Selectors box.
• Oxygen 6: select the element and add the class in the Classes control of the Properties panel. Oxygen Classic: open “Choose selector…”, then “Add Class”.
• Breakdance: select the element, open Settings, expand Advanced, add the class under Classes.
• WPBakery: edit the element and paste into “Extra class name” (near the bottom of the General tab).
• Avada: open the element options in Avada Builder, Extras tab, paste into “CSS Class”.
• GeneratePress, Astra, Kadence and other block themes: page content is blocks, so use the block editor steps.
• Classic theme or hand-written HTML: add the class to the element’s class attribute.

Builder not listed? Any field named CSS class, Additional CSS class(es), Extra class name or Attributes will do. The same steps are shown inside the plugin, next to the classes, with a Copy button. Point-and-click tests and URL tests need no class at all.

Until a test is started, visitors see the original version only; the new version stays hidden, so you can add the classes whenever you like. Only deleting a test removes its hiding rule, so remove the extra markup from the page first.

Is there a traffic or visitor limit?

No. Because tests run on your own site, there are no view caps, no visitor limits, and no overage bills.

Will visitors see a flash of the wrong version?

No. The plugin hides untested variants until the assigned one is ready, so the page reveals flicker-free.

Does it work with a strict Content Security Policy?

Yes. The runtime is a normal external script, its configuration is an inert JSON block, and the small inline scripts (anti-flicker reveal and URL-split redirect) are printed through WordPress’ inline-script functions, so a CSP tool that gives WordPress’ inline scripts a nonce gives these one too. The anti-flicker rules are inline styles that WordPress itself prints, like any theme’s or plugin’s: a CSP tool that adds nonces to the inline style tags WordPress prints covers this plugin’s too, and the plugin’s script gives the styles it adds the same nonce. Without such a tool, a policy that refuses inline styles refuses the anti-flicker rules as well: the page can show more than one version for a moment, until the plugin’s script runs, and from then on exactly one version.

Does it set cookies or send my data to a third party?

No cookies, and nothing is ever sent to us. Visitor and test data never leaves your site: a random, pseudonymous identifier is stored in the visitor’s browser so they keep seeing the same version, and your own database records which version that identifier saw and whether it converted, with the device type and the time. No names, emails or IP addresses are stored with your results.

Does it work with caching and CDNs?

Yes. Variant assignment happens in the visitor’s browser, so the same cached HTML can be served to everyone. It works behind page caching plugins and CDNs, with no cache exclusions needed.

One detail is handled for you: a small snippet in every page’s head changes when a test starts or stops, so cached copies made before that moment are stale. The plugin purges the page cache at those moments when it recognises the cache: WP Rocket, W3 Total Cache, WP Super Cache, LiteSpeed Cache, WP Fastest Cache, Cache Enabler, Hummingbird, Breeze, Autoptimize, SiteGround Optimizer, Nginx Helper, WP Engine, Kinsta, Pantheon and the Cloudflare plugin. With any other cache or CDN, clear it yourself right after starting or stopping a test (developers can hook the improveab_purge_caches action).

If your site sits behind a proxy or CDN that hides visitors’ addresses from WordPress, the plugin’s per-address rate limit for tracking events (600 a minute) is shared by everyone behind that address. Cloudflare is recognised automatically, and so is a reverse proxy on a private address that appends the visitor’s address to X-Forwarded-For; for any other proxy, restore the real address at the web server or return it through the improveab_client_ip filter.

Does it work with script optimizers (WP Rocket, Autoptimize, LiteSpeed Cache, Perfmatters, Cloudflare Rocket Loader)?

Yes. The plugin marks its own tags so optimizers leave them alone and adds itself to the exclusion lists of WP Rocket, Autoptimize, LiteSpeed Cache, Perfmatters and SiteGround Optimizer automatically. If you use a tool that only has a settings screen (Flying Scripts, WP Fastest Cache’s delay, Asset CleanUp), exclude runtime.js, improveab-pending and improveab-redirect: the two small inline scripts must run before the page paints and the runtime must not wait for the first click. If they are delayed anyway, tested elements are still revealed after two seconds and no visitor is ever stuck, but the flicker-free hiding and the URL split redirect happen later than intended and results from that period deserve caution.

Do ad blockers stop it?

Not with the lists they ship with. uBlock Origin, AdGuard, Brave and similar blockers work from public lists of ad and tracker addresses, and cloud testing tools, which load from their own servers, are on those lists. This plugin’s script and its tracking address are on your own site, and before each release they are checked against EasyList, EasyPrivacy and the uBlock Origin and AdGuard lists. A visitor whose browser runs no scripts or stores nothing sees the original version and is not counted, and anyone can still block a site’s files by hand.

Emails from the plugin don’t arrive. Is it broken?

Almost certainly not the plugin. Every email it sends (the “Email summary” button on a test’s results page) goes out through your site’s own WordPress mail, the same route as password-reset emails. The plugin has no mail server of its own. If the summary does not arrive, check spam first, then press “Email summary” again: when the mail server refuses, the page shows the server’s own error message. The fix is on the hosting side: install an SMTP plugin (WP Mail SMTP, FluentSMTP, Post SMTP or similar), connect a mail service, and send its test email. Once that works, the plugin’s emails work too. Nothing is ever emailed to your visitors.

Can the same visitor see both versions?

Rarely. Assignment is stored in the browser’s local storage, so it is per browser and per device: one person on a phone and a laptop counts as two visitors, and clearing site data or using a private window starts over. Safari and iOS also delete script-written storage after seven days without a visit to your site, so a visitor who returns later than that on Safari may be assigned again. This is inherent to cookieless testing and rare enough not to move a result; run tests for a week or two and read the final numbers.

Why are the numbers different from Google Analytics?

They are meant to be. The plugin counts people, not pageviews or sessions: a visitor is counted once per test however many pages they view, and each conversion once per person. A visitor is only counted when the tested element was actually on the page they loaded and the test was running. Logged-in staff are always excluded, events from crawlers and headless browsers are dropped, visitors with JavaScript or site storage blocked see the original and are not counted, and a conversion only counts for someone who was entered into the test. Expect the plugin’s visitor numbers to be below your analytics users and well below pageviews; what matters is that both versions are counted the same way.

My variant is not showing, or both versions show at once

While a test is Running, each visitor sees their assigned version; while it is a draft, paused, or stopped without a kept winner, everyone sees the original and the other versions stay hidden (only deleting a test stops the hiding, so remove the extra markup first). If both versions show at once, the classes are on the page but the plugin’s small head snippet is not: the page is a stale cached copy, or the plugin is not active on that site. While a test runs: logged-in staff always get the original, so check in a private window; make sure the variant’s class is on the new version of the element on that page; and clear your page cache and CDN if the cached copy predates the start. To see a specific version regardless, open the page with ?improveab=your-test-slug&improveab_variant=v1 added to the address (nothing is recorded). If a running test shows both versions to everyone, the plugin’s script is not executing on that page: a speed plugin is delaying JavaScript (exclude runtime.js), another script on the page throws an error, or the page is a stale cached copy. No visitors recorded at all usually means a security or firewall plugin is blocking the WordPress REST API for visitors; the improveab/v1/event route must answer for logged-out visitors.

Can I test a redesign on a different domain or subdomain?

Yes. A full-page URL test can send part of your traffic to a variant hosted on another WordPress site. Install the free plugin on both sites, create the URL test on the original site and use the redesign’s address (for example new.yoursite.com/pricing/) as the variant. Visitors are split between the two sites and every result is collected on the original. The original site must use https. A subdomain of your site works as it is (on a shared name such as co.uk or a host’s site address, your own name is the domain, so another customer’s site never counts as yours); for a different domain altogether, the variant’s site must name the original’s domain with the improveab_handoff_hosts filter, or no conversion is reported back. A domain with international characters works however it is stored (as typed, or in its xn-- form). Keep both sites on the same release and update them together: a site still on an earlier release can lose the conversions it reports back in some browsers. Your own address with or without www. is the same site, so the variant’s site needs a name of its own. For a redesign on the same domain, a standard URL test is all you need. See “Testing across two sites or subdomains” in the built-in guide for the SEO notes.

My site also answers under another domain. Are visits there counted?

Visits are counted from pages on your site’s own address (Settings, General, Site Address) and on any subdomain of its domain, so www. and the like are covered. If the same site also answers under another domain without sending visitors on to that address (a second brand’s domain, a language on a domain of its own), the plugin cannot tell its pages there from another site’s, so visits there record nothing. Redirect that domain to your site’s address, or name it with the improveab_handoff_hosts filter: add_filter( 'improveab_handoff_hosts', function ( $hosts ) { $hosts[] = 'example.net'; return $hosts; } );

Can I move my site to a new address with tests on it?

Yes. Migration and search-replace tools (WP-CLI, Better Search Replace, your host’s migration tool) update the plugin’s tables like any other, as long as the tables whose names contain improveab_ are included. Visitors’ versions are remembered by their browser for each address, so after a move to a new domain a running test sees every visitor as new. Where you can, start tests after the move.

Will A/B testing hurt my SEO?

No. Element tests change content client-side and don’t create duplicate URLs. For URL/redirect tests, each variant page is given a canonical tag pointing to the original (works alongside Yoast and Rank Math), so search engines understand they’re variants of one page.

Can I use it in place of Google Optimize?

Yes. Google Optimize was shut down in 2023. ImproveAB runs A/B tests inside WordPress, with no traffic limits and no external dashboard.

Can I test WooCommerce products?

You can test any element or page, including WooCommerce product and shop pages. A purchase can be the conversion goal, with revenue and revenue per visitor shown per variant. It counts paid WooCommerce orders. An order waiting for payment counts when it is paid; refunds and cancellations are taken off its revenue. So a stopped test’s purchase figures can still move a little: an order placed while it ran counts once it is paid, a deleted refund goes back on, and a cancelled order (cash on delivery that never arrived, say) gives its revenue back until it is paid after all.

Does it slow down my site?

It’s built to be lightweight. Visitors load one small script (about 18 KB compressed), deferred on WordPress 6.3 and later, plus a few lines in the page head that keep the wrong version from flashing; nothing is fetched from third-party servers, no cookies are set, and pages stay fully cacheable. Results are read from your own database only in wp-admin.

Is it GDPR-friendly? Do I need a cookie banner for it?

The plugin was designed for privacy: it sets no cookies, loads nothing from third parties, and stores no IP addresses with your results. A random identifier saved in the visitor’s own browser keeps their assigned version consistent, and your database records which version that identifier saw and whether it converted, compacted into daily totals once a test has ended and the retention period has passed. Whether you need visitors’ consent for this depends on the laws that apply to your site, so review it with your own privacy policy or counsel. If you ask for consent first, a consent tool can hold the plugin’s script back (developers can use the improveab_enabled filter); visitors without consent see the original version and are not counted.

Can I edit a test after creating it?

Yes. The name, description, notes and variant labels can be changed at any time. The goal and pages can also be changed on a live test, after a clear warning, since data already collected followed the old rules. Variant URLs are only editable while a test is still a draft; to try different pages once it has started, duplicate the test.

How long should I run a test?

Until it reaches the sample size you planned, and always in full weeks so weekday and weekend behaviour are both covered. The built-in calculator (in the Guide) estimates the visitors and days you need from your traffic and conversion rate before you start. Until then the results page names the variant with the highest rate so far (“Variant B leads so far.”): that is a reading, not a test of significance, so decide at the sample size you planned and then declare the winner.

What happens to my data if I uninstall the plugin?

By default, nothing is lost: deleting the plugin keeps your tests, results and settings, so you can reinstall later and pick up where you left off. If you prefer a complete clean-up, turn on “Delete all data when the plugin is removed” in Settings before uninstalling: every table, option and user setting the plugin made is then removed, whichever version of it made them. If ImproveAB Pro ran on the site before, deactivate its license on its License screen before you install this plugin in its place: this plugin cannot release a license, so a license still active stays counted against your plan until you release it in your account.

What happens to my pages if I deactivate the plugin?

Your tests and results stay. URL split tests stop redirecting and point-and-click changes stop showing, so visitors see your original pages. Element tests are different: every version you built is in the page itself, and only the plugin hides the ones a visitor should not see, so while it is deactivated visitors see every version at once. Remove the versions you no longer need from your pages before deactivating, or keep the plugin active. The Plugins screen reminds you while element tests exist.

Can editors or other roles manage tests?

Yes. Administrators manage tests, and the plugin checks its own capability, improveab_manage_tests, the standard WordPress way: any role you give it can create, run and read tests in this free plugin. Give it with WP-CLI (wp cap add editor improveab_manage_tests) or with any role editor plugin. Settings, and a URL test’s version on another site, stay with administrators. ImproveAB Pro only adds a settings screen for choosing the roles; this plugin’s Settings names every role that holds the capability, each with a button to take the access back.

Can I share results with my team?

Yes. Every test has a one-click plain-text summary you can download, or email to yourself and forward, including the variants, numbers, verdict and your notes.

Does it work on WordPress multisite?

Yes, per site. Each site in the network gets its own tests, results and settings. Network activation installs the plugin on every existing site, and a site created later sets itself up on its first page load. You can also activate it on individual sites only.

Can I create a test without touching my page builder?

Yes. While logged in as an administrator, open any page of your site, click “A/B Tests” in the admin toolbar and choose “Start a test on this page”. Click the element you want to improve, write the alternative wording, choose a goal, and start the test. The plugin swaps the text for visitors assigned to the variant; your page, theme, and builder are never modified. If the element contains inline formatting, such as “This is a heading in hero.” with a differently colored word, it opens as editable HTML so you can keep the tags and change only the words.