How to Use VWO to Test a Pricing Page With Claude Watching

The short answer: VWO, now sold as Wingify, publishes no prices, so run your first pricing-page test inside the 30-day free trial before you take a quote (homepage FAQ, checked on 26 October 2026); install SmartCode and check it, settle consent, create an A/B test with one primary metric, keep the traffic split equal, choose Bayesian or frequentist before launch, review and start it, connect Claude through the MCP server at the lowest permission that works, and let it summarise the report while a person makes the call.

Disclosure: I may earn a commission through some links on this site. It does not change what this page says.

A pricing page is where a visitor decides whether the product is worth the money, and small changes there move revenue, not just clicks. That makes it the page where a sloppy test does the most damage. Wingify's help centre is the source for each step below; the product still answers to the VWO name in places, including the user agent it scans your pages with (ideas guide). For the merger and the tiers, see the VWO glossary entry.

Before you start:

• The traffic sum. For example, 6,000 pricing-page visitors a week split two ways is 6,000 / 2 = 3,000 per variation a week. Reports show Collecting Data until 500 visitors and one conversion on the baseline (statistics guide), so probabilities appear within days and a decision takes longer.
• A trial account. When it ends, tests pause and their reports stay readable; an account inactive across all products is deleted after 45 days (trial guide).
• Access to the page head, or to the tag manager that loads scripts.
• A consent decision, made with your legal counsel, which Wingify's own guide recommends.

1. Install SmartCode and check it

Every page in the test needs the SmartCode snippet. A tag manager works but loads asynchronously, so visitors can glimpse the original before the variation appears (setup guide). Then open Configurations, Websites and Apps, select your domain, and on the Code tab enter the pricing page URL in the checker; it waits up to 30 seconds and tells you whether this account's code ran. If uBlock is on in your browser, whitelist Wingify's domain first, or the checker, previews and editor will not load (checker guide).

2. Settle consent

Wingify documents three ways to run SmartCode: without consent, conditional on the visitor's cookie choice, or with consent required only for analytics and push while tests run regardless (consent guide). Article 5(3) of the ePrivacy Directive on EUR-Lex requires consent for storing information on a device unless it is strictly necessary, and some customers class Wingify's cookies as necessary, per the guide. That classification is a legal judgement for your counsel, not a setting to copy.

3. Create the test

Go to Experimentation, then Web Experimentation, click Create and choose AB, Single Page. The Configuration tab walks through Pages, Variations and Traffic, Metrics, Targeting and Integrations, and the Review tab opens when they are complete (campaign guide). Enter the pricing page URL under Pages.

4. Build one variation and leave the split alone

Click Add variation, open it in the Visual or Code editor and make the one change your hypothesis is about: the order of plans, the annual-billing default, a line under the price. Keep the traffic split on Equal Distribution, the default; the alternatives are a custom split or a multi-armed bandit that moves traffic toward the leader (variations guide). A bandit finds a winner to exploit. It does not give you a clean read.

5. Pick one primary metric

The primary metric decides the winner, and secondary metrics add context (campaign guide). On a pricing page, make it the click into checkout or trial signup, and keep revenue per visitor as a secondary if checkout carries the code.

6. Choose the statistics before launch

Pick Bayesian or frequentist now, because the model cannot be switched on a running campaign. Winners are typically declared when the probability to be better, or the significance level, passes 95%, and a variation below 5% gets a recommendation to disable it (statistics guide). The region of practical equivalence sets how small a lift counts as no lift.

7. Review and start

On the Review tab, read the configuration summary, run the Debugger, use the QA configurations and live preview, and check the variation screenshots (campaign guide). Then start it.

8. Connect Claude to read the report

Add the Wingify connector from Claude's directory, or under Connectors, then Add connector, then Browser connectors. Choose the server for your account's region, for example https://mcp-eu.wingify.ai/mcp for an EU account, select the account and the permission level, and approve (Claude connector guide). The server's read tools fetch campaign reports, daily data, heatmaps and recordings; its write tools can create tests and change a campaign's status (MCP guide).

Each Monday, ask: "@Wingify show the report for campaign 123, the visitors and conversions per variation, and whether any variation has crossed its winner threshold." The agent reads and drafts the weekly note. The human check: grant the lowest permission level that can read reports, and leave the write and delete tools off Always allow in Claude's connector settings; the guide's troubleshooting table shows that without it, Claude cannot start or pause campaigns. A person decides. Claude cannot modify a running campaign in any case; it must be paused first.

9. Do not touch it mid-test

Wingify flags three mistakes in the report. Changing the traffic split mid-test produces Simpson's paradox, where combined totals point the wrong way; stopping short of full weeks biases the sample toward weekday or weekend buyers; and changing targeting or metrics after launch mixes two tests into one report (warnings guide).

10. Decide

When a variation crosses the threshold after whole weeks, roll it out, and write down the hypothesis, the samples and the result before the next test starts.

Every Wingz action, an MCP question included, spends from one shared credit pool, and the larger allowance sits in Wingz Advanced, the paid upgrade (credits guide), so an agent that asks a lot of questions is also a sales route for Wingify. One question a week is plenty.

FAQ

Do I need Claude for this?

No. The same report sits in the campaign's Reports tab. Claude saves the weekly reading and the note, nothing more.

Does the MCP connection cost anything?

Connecting is included: Wingz Core, free with paid products, covers the MCP server without behaviour analytics data, and each request uses credits from your allowance (pricing page).

Why does my report say Collecting Data?

Statistics appear only once there are at least 500 visitors and one conversion on the baseline (statistics guide).