A/B testing software.
Ship the winner.

Visitors are split between versions on the traffic you set — 50/50 for a classic A/B, even slices for an A/B/n test, or 45/45 with 10% held out on the untouched site to measure the lift you actually caused. By default a visitor keeps the same version on every visit.

Then the test decides one of two ways. Pick a winner at significance: traffic stays as you set it, and when a version beats the rest on your goals at the confidence, session floor, runtime and minimum improvement you chose, it is promoted and the test concludes. AI: optimize continuously: there is no conclusion — every night the split shifts toward the version performing now, weighting recent traffic, moving only with statistical confidence, and always keeping 10% exploring. Either way the result is also read per discovered audience, so you see who each version won for.

Pick a winner when you need an answer you will act on once: a redesign, a pricing layout, a checkout change. It runs a fair fixed split until one version proves itself, tells you the lift and how sure it is, and stops. You can defend that call.

Optimize continuously when the content is always-on and the right answer drifts: hero banners, offers, seasonal blocks, anything you would otherwise re-test every quarter. It never asks you to come back and declare anything; it keeps the split where the recent data says it should be, and keeps 10% exploring so it notices when that changes. Switch between them on the campaign at any time; the choice applies when you save. The long version, with the nightly pass step by step.

Each night the optimizer re-runs the test thousands of times in simulation from the data it has, and counts how often each version comes out on top. That share is its chance of being the true best; near-even numbers mean the data cannot tell the versions apart yet, and then the split stays put. The recommended split follows that chance, with 10% of traffic always divided evenly among the versions — so a losing version keeps earning the data that could redeem it, and a version that was best two months ago can still lose today. Views and clicks are decayed with a 28-day half-life, so recent traffic counts most.

You choose per test. Rotate users, the default: a visitor keeps the same version on every visit, right for funnels and anything measured over days. Visits: the same visitor can get a different version on each visit, for banner and headline tests where the impression is what is measured. Random: every page view picks afresh. No rotation turns the split off, so the groups are layers of content that all apply rather than competing versions.

Yes, once the results are in. While the test runs, one split serves everyone, and every test is scored inside each discovered audience separately — the rate per version, the leader there and how sure it is. Then each audience can get the version that won it: apply the winners to their audiences with one click, or let Personyze apply them automatically. Each winning version is shown only to the audiences it won; where the control group won, those visitors keep the original, and a version that won no audience is switched off. A holdout can keep running, so the result keeps being checked. The campaign is saved as a new version, and you can undo it as long as the campaign is not edited again. The audiences can be the ones Audience Discovery finds in your data, or any audience you build.

No. Personyze is built to be no-code: the visual editor lets you change copy, swap images, redesign CTAs, or replace whole sections by clicking on them on the live page. Popups, banners, and forms are built from a 20+ widget template library — no engineering team required.

For the rare cases where you want bespoke logic (custom JavaScript, server-rendered variants, integrations with proprietary data) there\'s a JS action you can drop in and a REST API for server-side variant assignment. Most teams never need either.

Personyze A/B testing is fully platform-agnostic. It works on Shopify, Magento, BigCommerce, WooCommerce, Salesforce Commerce, WordPress, Webflow, custom React or Vue apps, Next.js, headless setups, and plain HTML sites. Integration is a single JavaScript snippet.

For server-side variants there’s a REST API. For email A/B tests, Personyze injects content at email-open time. And for data, Personyze plugs into your existing martech: Google Analytics 4, Mixpanel, Segment, HubSpot, Salesforce, Marketo, ActiveCampaign, and any tool with a REST API or webhooks.

A/B testing, AB testing and split testing are used interchangeably — all three describe splitting traffic between two or more variants and measuring which performs better. “A/B testing” is the more common phrasing for landing-page and CRO experiments; “split testing” is sometimes used to specifically describe URL-based tests where each variant lives on a different page.

Personyze handles both: same-page A/B variants via the visual editor, and full-page split tests where you redirect a percentage of traffic to a different URL. Multivariate (A/B/n) and holdout tests are also supported in the same UI — no separate tool required.

From a fresh account, most teams launch their first campaign in under 15 minutes: pick a template, choose a trigger, define the audience, hit publish. No code, no developer queue.

For more complex setups (custom CRM data, ABM IP lookups, multi-step flows) plan on a few hours or schedule a free implementation session with our success team.

No. The Personyze script is under 30KB gzipped, loaded asynchronously, and decisions are made in under 50ms. We don’t block render. Customers consistently report no measurable impact on Core Web Vitals after deploying.

Yes — every popup, bar and banner is built from standard HTML/CSS, so you can use the visual editor, or drop in your own custom HTML and inherit your site’s CSS. Though you rarely need to: Personyze offers to apply your brand to any template it shows you — your colours, your fonts and your language — so a design starts on-brand instead of being restyled into it.

Yes — native integrations for HubSpot, Salesforce, Mailchimp, Klaviyo, ActiveCampaign, and Marketo. Captures from popup forms can be auto-pushed to your CRM as new leads, with custom field mapping. Webhooks let you connect anything else.

Yes — testing is built into every capability rather than bolted on. Test recommendation algorithms against each other, pit popup offers and triggers head-to-head, and prove which landing page variants win for each audience.

Yes, and the report is built to say so. Every test is also read per discovered audience: the rate per arm inside the audience, the leader there, and the confidence of a disagreement with the overall leader, measured against the arm it disagrees with. A one-line verdict at the top of the results names the biggest audience that disagrees; the A/B tests by audience report lists every audience with enough traffic, with a conclusive-only filter. Membership is as of the pass that tagged the sessions, which the panel says in its tooltip.