Stop guessing. Know what sells.

Orders and revenue, not clicks

Sessions, paid orders and revenue attributed to the test, and a Leader that stays honest: it says “not enough data yet” until there is.

One primary metric, chosen up front

CVR, RPV and AOV side by side against the control. The starred metric is the one the experiment is read and decided on.

Two statistical reads, side by side

A Bayesian read tells you how likely the variant is to beat the control. A classical p-value sits next to it. They answer different questions, so you get both.

The whole checkout, per variant

From session to cart, checkout, payment and order, with each variant measured against its own traffic. Descriptive on purpose: no verdicts where there is no test.

And a check on the data itself

Was the split what you asked for? Did visitors keep their variant? Is checkout coverage healthy? You see it before you trust the number.

The page answers the ad

A visitor clicks a promise and should land on that same promise. Proa reads the URL parameter your ad already carries and changes the page to match it.

No duplicate landing pages

Headline, image, offer and button change per ad, on one URL. One page to maintain, one theme, and the campaign says the same thing from the ad to the cart.

Show it to everyone, or test it

Deliver a version to every visitor from that ad, or run a campaign test: one experiment per ad, so you know which message sells and not just which one gets the click.

Automatic winner, with rules you can read

Proa checks every hour. A variant rolls out only when every condition on the list is met, and the list is right there on the screen.

Roll out & validate

The winner goes to everyone while Proa keeps watching it against the numbers that made it win. If it starts underperforming, you get an alert, not a surprise.

One click back

Revert rollout returns the store to the original. Pause, resume or end a test whenever you like; making a winner permanent is a separate, deliberate step.

Your storefront, inside the editor

The real store renders next to your edits, at desktop or mobile width. No staging copy, no screenshots to guess from.

Pick from page

Click an element and Proa writes the edit for you: type, target and content filled in, with a live count of what it matches.

A variant is a list of edits

Text, HTML, image, attributes, hide, remove or CSS only. Each edit is targeted on its own and applies as you type, before anything is saved.

Check it where it matters

Flip to mobile width, keep an edit alive when the theme redraws the page, and export a version as JSON to reuse it on another site.

Only what changes waits

The rest of the page paints as usual. Just the elements a variant touches hold back for a moment, so nobody watches the old headline turn into the new one.

Then it appears, already changed

Each element is released the moment its edit lands. A visitor sees one version of your store, the one they were given.

And the page always shows up

If anything goes wrong on our side, the page reveals itself anyway. Your store is never left waiting on Proa.

Pick the kind of test

Change the page in place, swap the theme template, send visitors to another page, or give each ad its own experiment. Every later step adapts to the answer.

Nothing goes live by accident

The wizard ends in a draft. Six short steps, a review at the end, and the store stays untouched until you press Start.

More on the way

Price, post-purchase, theme and checkout tests are on the roadmap, already visible in the wizard so you know where this is going.

Split the traffic your way

Set a share per variant and watch the bar fill. Visitors keep their variant from page to page, and any share you hold back simply sees the store as it is.

Say where it runs

Include or exclude pages by path, limit by screen size, or only enter visitors whose URL carries a given parameter, like the ad that brought them.

No two tests fight over a page

If another running test already claims the same pages, Proa says so before you start, not after the numbers went strange.

Open preview, per variant

A link puts you inside one variant on the real store, for your session only. Nothing is assigned, nothing is counted, and it works on drafts.

On your phone in one scan

Every preview has a QR code. Switch variants from the badge on the storefront without leaving the page you are checking.

A checklist that reads the test

Before Start, Proa checks the things that quietly ruin experiments and tells you, in plain sentences, what still stands in the way.

Ask the results a question

Analyse this experiment writes up what happened over the window you are reading. Chat about your results picks up from there.

Yours to export

Copy the results as one self-contained Markdown file for any AI agent, or export CSV from any experiment view. Aggregates only.

Connect, activate, done

Install the Proa app, switch it on in your theme, and the checkout pixel is set up for you. No code in the theme, and Proa tells you what it actually sees on the storefront.

Not on the app? One line works too

Copy the Proa snippet into your theme and you are running. Same tests, same reports, on Shopify or any other platform.