A/B testing

A/B Testing Built Into Your Website

Show part of your visitors a new version of a page, and keep the one that brings more leads. Pick a page from your most visited, make its new version in the editor and press Start. Within about a minute, Hiveku splits your visitors on your live site.

Split tests
Routes for ad visitors
Leads and calls counted
Runs on your live site

Start Free Trial View Pricing

The A/B tests tab in the Hiveku editor for Locus Digital's website: the three steps (pick a page, make a new version, split your visitors), New route and New test, and 'Choose a page to test' listing the site's 530 pages most viewed first, with section chips and a Test this page button on every row. — A/B tests in the editor: every page of your site, most viewed first, ready to test.Point at a numbered area to see it up close.

Set up a test of any page straight from its row.
Pick a page, make a new version, split your visitors.
Every page of your site, most viewed first, with sections and search.

Your site's pages in one list, most viewed first, with a search box and, on large sites, sections.
Hiveku copies the page for you. Change the copy yourself, or ask Hiveku's AI.
Count leads, forms, calls, button clicks, pages reached or scrolling.
No winner is called until the numbers are clear.

Start from your data

Find the Page Worth Testing

Start with a page people already visit. In Analytics, open one of your top pages and press Test this page. Heatmaps and session recordings show where visitors click, how far they scroll and where they get stuck, so you know what to change.

Top Pages in Hiveku Analytics for Locus Digital's website, most viewed first: the first row (Digital Marketing Agency Plano TX, 6,910 views) is open with its full URL and the Open page, Test this page and Copy URL buttons; Contact Us, Our Work, Testimonials, Free Quote and Case Study follow. — Analytics' top pages: open one and press Test this page to set up a test of it.Point at a numbered area to see it up close.

Open a top page and test it straight from Analytics.
Your most visited pages, ranked by views.

A click heatmap over the Locus Digital home page, from 357 clicks: warm spots on the Get Proposal button and the navigation, with modes for clicks, rage clicks, dead clicks, scroll and movement and filters for device, country, browser, visitor, UTM source and referrer. — Heatmaps: where visitors click, scroll and move on each page.Point at a numbered area to see it up close.

See where people click and how far they scroll.
Switch between clicks, rage clicks, dead clicks, scroll and movement.

A session replay open over the recordings list: a visit to the Locus Digital home page played back in Hiveku, with the timeline, speed buttons up to 8x, skip inactive and full screen. — Session recordings: watch a visit back to see what got in the way.Point at a numbered area to see it up close.

Watch a visit back, click by click.
Play at up to 8x and skip the idle parts.

How it works

Set Up a Test on One Form

Everything a test needs is on one form in the editor's A/B tab. Nothing changes on your live site until you start the test.

1
Pick the page
Choose the page you want to do better. Hiveku shows its views for the last 28 days.
2
Make the new version
Let Hiveku copy the page and change the copy, or pick a page you already built.
3
Choose who is in it
Everyone who opens the page, or only some visitors, like those from one ad or on a phone.
4
Say what counts as a win
Leads anywhere on the site are the usual choice: a form sent or a call made.
5
Set the split
Half your visitors see each version. Or pick 60/40, 80/20 or 90/10, or drag the slider.

New A/B test form in the editor for Locus Digital's Home page: 1 Page to test (Home, 675 views in the last 28 days), 2 The new version (Copy this page, then change the copy, recommended; or use another page I already have), 3 Who is in the test (Everyone, or only some visitors) and 4 What counts as a win (Leads anywhere on the site, most used). — A new A/B test on one form: the page, its new version, who is in it, what counts as a win and the split.Point at a numbered area to see it up close.

Split visitors 50/50 or pick another share; nothing changes until you start.
Pick the page; Hiveku shows its visits for the last 28 days.
Copy the page and change the copy, or use a page you already built.
Test everyone, or only some visitors, like those from one ad or on a phone.

A copy gets its own address, like /pricing-b, and goes live when you deploy. Change it in the editor's preview, or ask Hiveku's AI to change only the copy. Blog posts and other pages made from a template can't be copied: build the new version as its own page, then pick it.

A new test waits as a draft in the A/B tab, listed with your other tests and routes by status. A draft can show the original and the new version side by side, each with its share of visitors.

Start a new test or route, and see your tests by status.
Each row shows its pages, who it is for and how visitors are shared.

Both pages load as visitors see them; looking at them here never counts in the test.
One line says which page is tested and how visitors are split.
Choose what counts as a win, track extra goals and set the split.

What counts

Count the Leads That Matter

Pick one main goal. Hiveku counts it for each version, on the tested page and anywhere on the site afterwards, from the same forms and calls its analytics records. Each visitor counts once.

Leads anywhere on the site: a form sent or a call made, on any page. The usual choice.
A form sent anywhere on the site, or only on the tested page.
Reached a page, like a thank-you or checkout page.
Clicked a button your site marks as a call to action.
Phone calls to a tracked number from a visitor in the test.
Scrolled at least half, or 75%, of the tested page.
An event your site sends to Hiveku, or a conversion goal from Analytics.
Up to 5 extra goals to report on. Only the main goal decides the winner.

Step 4 of a new A/B test, What counts as a win: Leads anywhere on the site (most used, selected), a form sent anywhere on the site, a form sent on the tested page, reached a page, clicked a marked button, phone calls, scrolled at least half or 75% of the page, a tracked event, or a goal from Analytics. — What counts as a win: leads, forms, a page reached, button clicks, calls, scrolling or your own goals.Point at a numbered area to see it up close.

Count leads from any form or call, a form on the page, or a page reached.
Or button clicks, phone calls, scrolling, your own events and Analytics goals.

The Forms view: 56 submissions across 7 forms with 66 started and 100% completion, a submissions inbox (names, emails and phones blurred) showing which form and page each came from, and each form's completion with submitted and started counts. — Forms: every form on the site captured, with submissions and completion rate.Point at a numbered area to see it up close.

Every submission in one inbox, and how each form converts.
Submissions, forms, starts and completion at a glance.

Recent calls in Hiveku website analytics: each call with its time, disposition (AI handled), duration, the source that brought the caller (direct, Google Ads, Bing, a directory), conversion uploads to Google Ads and GA4, and a recording to play; caller and tracking numbers blurred. — Calls from the website, each with the source that brought the caller.Point at a numbered area to see it up close.

Every website call, answered by people or the AI receptionist.
Each call tied to its source, sent to Google Ads and GA4, with its recording.

Routes

Send Each Ad's Visitors to a Page Made for Them

A route sends the visitors you choose from the page they land on to another page, like a landing page for one ad. Everyone else sees the page as usual, and your ads stay as they are.

Came from an ad on Google Ads, Microsoft Ads, Meta, TikTok, LinkedIn or X, read from the click id the ad adds to its link.
Came with a UTM tag: a source, medium or campaign that is, or contains, the words you set.
Came from another website you name.
Use a phone, a tablet or a computer.
Are in a country, or in a state or region.
Keep 5% to 50% of these visitors on the original to compare, or send them all.

Add up to 5 conditions, and a visitor must match every one. Hiveku checks each visitor as they arrive, before the page opens. The same conditions can limit a split test to some visitors.

New route form for Locus Digital's website: 1 Page they land on (Home), 2 Who to send (visitors who came from an ad, Google Ads, and use a device, Phone: 'Visitors who came from Google Ads and use a phone.'), 3 Where to send them (Free Quote, /free-quote) and 4 Compare with the original (keep some of these visitors on the original). — A route: send the visitors you choose, like Google Ads clicks on phones, to a page made for them.Point at a numbered area to see it up close.

Choose who to send, like visitors from Google Ads on a phone.
Start from the page your ad points to.
Send them to a page made for them, and keep a share on the original to compare.

A draft route lists what is left before it can start, and shows both pages with the share each will get. Once you start it, matching visitors are sent on within about a minute, with no deploy.

A draft route, "Google Ads phone visitors to Locations", in the Hiveku editor's A/B tab for the Northwind Coffee Roasters demo site, scrolled to its checklist: 1 The page it sends them to: /locations (Edit it in the preview, Ask the AI to change it), 2 Put both pages live ("Both pages are live. The route is ready to start."), 3 Start the route, then Counts as a win: phone calls, Also tracked: leads anywhere on the site, and Compare: 10% of these visitors stay on the original to compare. — A draft route with its checklist: Hiveku checks that both pages are live before it can start.Point at a numbered area to see it up close.

Send visitors from one ad to a page made for them, and keep a share on the original to compare.
A checklist shows what is ready before anything goes live.
Who it is for and where they go, in one line.
The goal, an extra goal and the share kept on the original.

Hiveku already knows where your visitors come from: Analytics breaks traffic down by source, campaign and ad platform, and the ads dashboard shows each connected platform side by side.

The Sources view of Hiveku website analytics: sessions by channel (direct, organic search, referral, organic social, AI assistants, paid social, paid search) and a source and medium table with sessions, visitors, bounce rate and share. — Traffic sources: search, ads, social, email and referrals.Point at a numbered area to see it up close.

Break traffic down by source, campaign, keyword or ad platform.
Where your visitors come from, including AI assistants.

The Hiveku PPC dashboard for Locus Digital, all time: $35,198.99 spent, 19.4K clicks and 1.8M impressions, conversions counted per platform (Microsoft Ads 79, Google Ads 66), and cards for Google Ads, Microsoft Ads, LinkedIn Ads, TikTok Ads and Meta Ads. — The ads dashboard: spend, clicks, impressions and conversions across every connected ad platform.Point at a numbered area to see it up close.

Spend and results across your ad accounts.
Google, Microsoft, LinkedIn, TikTok and Meta in one place.

Start, watch, decide

Checked Before It Starts, Called Only When It Is Clear

Hiveku checks a test before it changes anything, watches it while it runs, and calls a winner only when the numbers say so.

Checked before it starts

Both pages must answer on your live site, no redirect may move either one, and no other test may use them. A draft lists each check, and nothing changes until they all pass.

Yours to change while it runs

Change the split, pause or resume at any time. Each version keeps between 5% and 95% of visitors. If you save a change to a page under test, the editor warns you that it mixes two versions in one result.

Health checks every 10 minutes

A split check warns when visitors are not shared as set. If your live site can no longer split visitors as set, Hiveku pauses the test and says why.

A winner only when it is clear

No call before 7 days, 100 visitors on each version and 20 conversions in all. A version wins at a 95% chance of beating the original. After 14 days, versions within 10% of each other are called no meaningful difference.

Told when it matters

The people who manage the site get an inbox item, a bell notice and an email when a result is called, a health check warns, or Hiveku pauses a test.

End it your way

Keep the original, or send everyone to the winner with an ordinary temporary (302) redirect that you can see and edit on the Redirects page. A route, or a test for some visitors, ends by keeping the original.

The full report shows each version's visitors, conversions and rate, a chart over time, what the leader would add across the whole site, results by device, health checks and a timeline of every change.

Put the new version live from the editor's Deploy menu before you start. A test that ends by sending everyone to the winner leaves its redirect on the Redirects page.

Hiveku builder with the Deploy menu open over the live Northwind site: Development and Production, each with '1 file changed', its address and a Deploy button, and Deploy All. — The Deploy menu: send the site to development or production, one tier at a time or all at once.Point at a numbered area to see it up close.

Each tier shows what changed since it last went live, with its own Deploy button.
Deploy from the editor's top bar, right beside the live preview.

Hiveku URL Redirects page for the Northwind site: four 301 redirects such as /old-menu to /shop and /cafes to /locations, the 404 page setting, search and filters. — Redirects for the site in one list: old addresses sent to new ones, with import, export and a deploy button.Point at a numbered area to see it up close.

Old addresses sent to new ones, with the redirect type and status.
Add a redirect in one line.
Import, export or deploy redirects, and choose how missing pages are handled.

Questions

A/B Testing Questions

Which sites can run A/B tests?

Next.js sites that Hiveku builds and hosts, once they are live on production. Tests use the pages on your live site, never a branch. Sites published as plain HTML files and sites hosted somewhere else cannot run tests, and the A/B tab says why when a site can't.

What happens to SEO during a test?

A test copy is left out of your site's sitemap, and while a test runs, the copy points search engines to the original with a canonical link. If you end a test by sending everyone to the winner, Hiveku adds an ordinary temporary (302) redirect that you can see and edit on the Redirects page.

What about visitors who decline cookies?

Visitors who have not allowed analytics in your cookie banner see the original page and are not counted. A route still sends matching visitors on, but does not count them. Tests read that choice from Hiveku's cookie banner, so a page that uses another consent tool, such as OneTrust or Cookiebot, needs Hiveku's banner before a test can start.

Does a visitor see the same version each time?

Yes. Hiveku remembers which version each visitor got, so they see the same one when they come back.

How long does a test take?

It depends on how many people visit the page. Hiveku calls nothing before 7 days, 100 visitors on each version and 20 conversions in all. When you set up a test on the A/B Tests tab, a rough guide says about how many weeks the split needs at the page's daily views.

How many versions can a test have?

Two for now: the original and one new version.

Can I test a blog post?

Pages made from a template, such as blog posts, cannot be copied. Build the new version as its own page in the editor, then pick it as the new version.

Can Hiveku's AI help?

Yes. A draft test can hand the new version to Hiveku's AI, with a request to change only the copy and leave the original as it is. Hiveku's AI agents can also set up and run tests, and each change to your live site waits for your yes.

I'm an agency. How should I set up my clients?

Give each client its own Hiveku account. Memory, brand voice, AI agents, contacts and reports belong to an account, so each client's tests and results stay with that client. You switch between client accounts from one sign-in.