Choosing A/B testing tools for UK ecommerce stores

What to take away

Start with VWO or Convert on a self-serve plan for page-level tests. Move to Optimizely or AB Tasty when variants need assigning inside checkout.
UK-only hosting is rare. Look for a named EU region, a published sub-processor list and a transfer mechanism you could show a regulator.
Client-side tools cannot edit a Stripe, PayPal or Shopify checkout page. Server-side assignment plus a returned order event ties revenue to a variant.
These vendors price on monthly tested visitors, so session volume sets the bill more than the feature list does.

Four vendors and the Shopify app route

For UK stores choosing A/B testing tools, assignment and data location decide the shortlist, not the size of the feature grid.

Where assignment happens

VWO: browser script, plus a server-side SDK
Convert: browser script
AB Tasty: browser script, with server-side testing
Optimizely: server-side flags
Shopify App Store testing apps: browser script in the theme

Reach into checkout

VWO: product and cart pages; checkout needs the SDK
Convert: theme pages; hosted payment steps sit outside it
AB Tasty: can assign before a redirect with the SDK installed
Optimizely: payment steps run by a third party
Shopify App Store testing apps: cannot modify the checkout page

Licence model

VWO: self-serve tiers by monthly tested visitors
Convert: published price list by monthly visitors
AB Tasty: entry plan, then quoted traffic tiers
Optimizely: annual contract, quoted, no self-serve
Shopify App Store testing apps: app subscription by traffic band

Confirm each line on the vendor's current plan page.

Platform choice shapes the work as much as the tool. Our Shopify vs Magento CRO comparison sets out what each platform allows without a developer.

Data residency questions for UK stores

Experiment data carries a visitor identifier, so it is personal data with a lawful basis and a retention period attached. The ICO's UK GDPR guidance hub explains how to record both.

Ask every vendor three questions: the named hosting region, the sub-processor list, and what happens to identifiers after cancellation.

EEA hosting is straightforward, because the UK treats the EEA as adequate. US hosting needs a transfer mechanism such as the IDTA or the UK Addendum.

Run those three checks before you open a pricing page.

Three filters before pricing

Where does experiment data sit?

Yes

check transfer mechanism and lawful basis

still record lawful basis for processing

Consent, refused cookies and PECR

Testing and analytics cookies need consent unless an exception applies, and the ICO guide to PECR sets out the cookie rule.

A refusal must be honoured inside the test. If a tool keeps assigning variants after a visitor declines, the consent banner is doing nothing for that visitor.

Ask whether suppression happens before the script loads or only inside the tag manager. Only the first answer is reliable.

Example: testing a hosted checkout step

Testing a hosted checkout step

Write the hypothesis in one sentence, such as a shorter delivery promise lifting completed orders.
Install the server-side SDK on your own domain so the variant is assigned before the browser reaches the provider.
Carry the variant identifier into the order metadata your gateway returns, which is the event chain explained in our comparison of conversion attribution methods.
Place a sandbox order with live credentials and check the variant appears against that order.
Run a full week and a full trading cycle before reading the result.

Assessing the monthly licence

Self-serve plans are quoted in US dollars by monthly tested visitors. Convert publishes a price list, VWO lists rates per tier, and the other two move to a quote as traffic grows.

A tool cannot fix thin traffic, so check session and order volume before paying for a premium tier.

Check traffic before premium tier

Count monthly sessions and orders
Estimate detectable change size
Check ONS monthly online share
Is the test month typical?
Yes -> proceed with test plan
No -> adjust timing or category benchmark

Seasonality distorts that check. The ONS retail sales reference tables publish the monthly online share of retail sales, so you can see whether your test month looks typical.

Record monthly sessions, orders and how many tests run at once.
Ask the price at that level and at the tier above, so the step is visible.
Ask what happens to running tests if you cross a band mid-month.
Request export format, retention after cancellation and any exit fee.

Convert the dollar figure at a rate you note that day, and check whether VAT is included, since a registered business reclaims it. Put subscription, implementation hours and migration on one sheet with a source beside every figure, as in our store-change budget template with a source field on every row.

Common questions

Which tool should a small UK store start with?

VWO or Convert, on a self-serve tier, while the tests are page-level. Move up when checkout tests become the priority.

Do I need UK data residency?

No. A named EU region with a valid transfer mechanism and a short retention period is defensible, and usually cheaper than a UK-only demand.

Can I test a Stripe or PayPal checkout?

Not with a browser-only tool. You need server-side assignment before the redirect, and the gateway must return the variant with the order event.

How long should a test run before I trust it?

Cover a full week and a full trading cycle for your category, and treat peak weeks as their own case. A quiet fortnight does not predict December.