Choose the one your order volume, your team and your legal counsel can support, in that order.
Under a few hundred orders a month, don't buy any license: Clarity is enough to observe, and Shopify's native tests are enough to experiment. Above that, the choice comes down to hours of work and to the contract.
The benchmark
First, find the row in the table that looks like your store.
To detect a 20% lift, a test needs a little over 400 orders per variant (the calculation is here), or nearly 850 to collect in four weeks on the tested page. Tests run by the authors point the same way.
On two stores with a few hundred orders a month, tests reached no conclusion in two weeks on gaps of a few percent. On a third, with several thousand orders, they resolve in three to five weeks.
| Your situation | Starting point | Public entry price, early October 2026 |
|---|---|---|
| A few hundred orders a month, or fewer | Microsoft Clarity, then Shopify Rollouts | Clarity is free, Rollouts tests start at the Grow plan |
| Testing Shopify templates, by device | Shoplift | $74 a month billed annually, up to 50,000 people a month |
| Also testing a price or shipping fees | Intelligems | Pay as you go from $69 a month, $1,279 a month for unlimited annual |
| No data team, French vendor | Kameleoon | From $495 a month, 50,000 people tested |
| Frankfurt hosting listed | Convert | $299 a month billed annually, 100,000 people tested |
| Someone on the team who codes | GrowthBook self-hosted, PostHog EU Cloud | GrowthBook is free self-hosted, PostHog is free under a monthly cap |
| Multiple markets, high volumes | Kameleoon or Wingify (AB Tasty and VWO), custom quote | No public pricing at Wingify |
Since June 5, 2026, Rollouts compares two themes or two checkout configurations. If you're already on the Grow plan, the tool is paid for.
On the testing side, the authors use only Intelligems day to day. The other tools in the table are listed based on their price lists, with no field feedback from us.
The rule
Before buying a license, wait until your traffic lets you conclude a test in two to four weeks and you have hypotheses in reserve. The authors did the opposite.
The tool was already running at a brand the authors work with when they counted its orders: fewer than a hundred in nearly a month. At that pace, judging a single test would have taken five months.
The developer then asked the question that should have come before the installation: "Aren't there other things we should work on first?" For lack of a better option, they ended up judging on add-to-carts.
Factor in consent too. In its July 2025 self-assessment tool, the CNIL excludes from the audience measurement exemption "any creation of a cohort of users to present them with differentiated content," whether or not the cohort is drawn at random.
We infer that an A/B test requires consent whatever the tool, and your legal counsel should confirm that reading.
The steps
- Volume. Calculate how long a test would run on your most-viewed page before asking for a quote. Beyond four weeks, stay with Clarity and fixes that need no test.
- Subject. A price or shipping fees call for a specialized tool like Intelligems, whose pricing module is excluded from the free trial because the vendor builds the integration itself.
- Production. Ask who will build the variants. GrowthBook and PostHog cost nothing up front, as long as you have someone to code them.
- Hosting. Have the vendor name the region and the infrastructure provider, because an "EU" mention is too vague. Get in writing any transfers outside the EU and the vendor's role under Article 28 of the GDPR.
- Refusal. If only visits that accept cookies are tested, the test runs correspondingly longer. Check what the tool does with the others, and that your banner passes it the signal.
- The contract. With Wingify, ask for the renewal grid at 24 and 36 months. With Kameleoon, have the currency written in, since the same tier shows as $495 on its site and as €495 on Capterra.
The trap
The trap is choosing from a comparison, which goes stale fast. The merger of French vendor AB Tasty with Wingify, VWO's parent company, was announced in January 2026, and the two products have been combined under the Wingify brand since September.
In our view, renewal terms and the subcontracting chain will take a year or two to settle, and a multiyear commitment signed before then deserves written guarantees.
Above all, a comparison lines up licenses, and in practice the license is the smallest line item. Against a $69 entry subscription, our estimates put development at two to eight hours per test depending on the component, then one hour to launch. We put the removal of a losing variant left in the code at six hours.
The case to make
Before any quote, write down three numbers: the monthly orders on the page to test, the hours available to build variants and the cost of the license.
A fictional store with 300 monthly orders would wait nearly a quarter to see a 20% lift. Copy this, then rewrite it with your numbers for your leadership:
"With 300 orders a month, no tool will tell us within three months whether a variant wins 20%. We're keeping Clarity to fix what we observe, and we'll reopen the question when the page to test approaches 850 orders a month."
