Shopify A/B testing where you only pay for wins.

Ecommerce founders shouldn’t be running A/B tests themselves.

Trusted by Shopify brands

Northbound GearPixie WingNourished LifeFlora & FaunaBacked byStartmate
Scroll

A year of tests, and what you pay for it

A year is a lot of tests.

Most stores can run 2 to 4 tests a month.

You approve. Cascayd does the rest.

AI does the research and drafts each test fast. A person checks every one before it reaches your customers.

  1. 1

    Cascayd studies your whole store.

    It reads your site, past tests and customer personas, then ranks test ideas by impact.

    Reading every page

    Free shipping on all orders

    ≡DAWN

    Leather goods, built to last.

    Shop bestsellers
    ✓
    Home
    ✓
    Collection
    ✓
    Product
    ✓
    Cart

    Ranked test ideas

    • 1. Hero headlineHigh
    • 2. Product page reviewsMedium
    • 3. Cart drawer upsellMedium
  2. 2

    Every change runs as an A/B test you approve.

    Shoppers see either your page or the new version, so you learn what sells more. Live within 12 hours of your yes.

    ACurrentNew versionB

    Free shipping on all orders

    ≡DAWN

    Leather goods, built to last.

    Shop bestsellers

    Free shipping on all orders

    ≡DAWN

    Leather goods, built to last.

    Shop bestsellers
    A vs B

    ✓Reviewed by the Cascayd team

  3. 3

    Every test makes the next one better.

    Wins and losses both teach Cascayd what works for your store, so the gains compound.

    Each result shapes the next test

    Learn

    Shoppers care about durability

    Wins make you money

    Ideate

    Durability badges on product pages

    Launch a test

    Live within 12 hours of your yes

Works with what you use

No flicker, and no blank screen

Most testing tools stop flicker by hiding your page until they load. Shoppers stare at a blank screen, and the wait hurts your speed score on Google.

Testing tools, in their docs: we hide your page until our script loads, so shoppers never see flicker.

Meme from The Big Short. “I don’t get it. Why are they confessing?” “They’re not confessing. They’re bragging.”
A blank screen isn’t a fix for flicker. It’s a slower page.

Flicker method from each tool’s own docs. Size is compressed, measured on a live store running each tool, November 2026.

Your shoppers stay anonymous

Each visitor gets a random ID, so Cascayd never knows who they are. These limits are written into our services agreement.

JR

Jamie Rivera

What your store knows

  • jamie@example.com
  • +1 555 0134
  • 14 Harbour St, Portland
  • IP 73.42.118.9

Stays with you. Cascayd never receives it.

?

v_8f3a21c9

What Cascayd sees

Page
/products/canvas-tote
Saw
Homepage hero, version B
Did
Added to cart
Ordered
$84.00, 2 items

Enough to tell which version sold more. Nothing about who they are.

Never collected: names, emails, phone numbers, addresses, IP addresses, full URLs or device fingerprints. IP address and browser details are checked only to filter out bots, then discarded.

What founders ask us first.

Do you store all my customers’ data?

No. Cascayd tracks visitors against a random ID and never collects names, emails, addresses, form inputs, IP addresses or location. From Shopify it reads order totals and line prices only, to count revenue for each test. These limits are written into our services agreement.

Will the pixel slow down my site?

No. The script is 10.5kb compressed, loads in one request from a cache close to your shoppers, and never hides your page while it loads. In our benchmarks, pages with Cascayd loaded as fast as the same page without it, with 0ms flicker.

What counts as a win?

A test wins when we’re statistically confident the new version earns more revenue than your current one, meaning the low end of the estimate is above zero. If the result is unclear, it isn’t a win and you don’t pay for it.

What does the 80% lower bound mean?

Every test gives a range for how much extra revenue the winning version brings in each month. The 80% lower bound is the low end of that range: we’re 80% confident the true uplift is at least that much. We bill 35% of that low number, so if the lower bound is $10k a month, you pay $3.5k and keep $6.5k plus everything above it. That’s pretty confusing but you get the good end of the deal!

What happens when a test loses?

You pay nothing, your store goes back to the original, and the result still shapes the next test. A losing version can cost some sales while the test runs, which is why a person reviews every test before it launches.

How much traffic do I need?

Traffic decides how many tests your store can run, usually 2 to 4 a month. We’ll look at your numbers first and tell you plainly what they support.

Do you implement the winner after the test?

Yes, if you want us to. If you’d rather your own team ships it, we’ll hand over everything they need.

How long until my first test is live?

Within 12 hours of you approving it.

Is it all done by AI?

AI analyses your store and drafts tests quickly. A person reviews every test before it goes live, and you approve it too.