Analyze an A/B Test and Call the Winner (Claude Cowork Self-Install Kit)

Call the winner with a reason you can say out loud, and stop re-litigating it next month.

You run the tests, but you're never sure the winner is real or just noise, so you second-guess every call. You called a subject line early once, watched the winner lose the next month, and now you sit on results for weeks instead of deciding.

The leak isn't your testing. It's that your tool shows you percentages but never says plainly whether there's enough data to decide, so the decision gets deferred instead of made.

What this kit does

The kit compares control against variant, computes the lift and the statistical significance, and recommends the winner with a plain-English rationale and a confidence caveat. It works on email subject lines, landing pages, or ad creative. Paste in the raw numbers, visitors and conversions per arm, and you get the call back without needing an analytics stack.

What you can expect

Every test gets a statistically grounded winner with the lift percentage and a rationale you can actually explain to a client. You stop sitting on results for weeks, because the kit tells you plainly when the data supports a call and when it doesn't yet.

Who it's for

Built for agencies and marketing agency owners, coaches and consultants, and e-commerce operators who test regularly and want the call made on evidence. This is a self-install kit you run yourself with your own AI agent, built for Claude Cowork first.

What's included

Skip building from scratch. No coding. You get the complete kit your own AI agent reads, sets up, and runs for you, step by step:

• A Start Here guide, so you know exactly what to open first
• The agent install instructions, the part that lets your own AI agent stand the workflow up for you
• A guided setup that interviews you about your business, so the workflow fits how you actually work
• Workflow Briefing: the plan, the inputs, and what done looks like
• Task List your agent follows to build it
• MCP Connection Guide for the tools involved
• Tool Setup notes for accounts, keys, and permissions
• Prompt Library, including the ready-to-run prompt for this workflow
• Troubleshooting and maintenance guide for the common snags
• A Next Steps and Roadmap, so you know what to automate next once this one is running
• A Run Guide you keep, so you can run it any time with no setup again

It's a small kit that earns its keep on repetition. One test you call correctly instead of guessing at is worth more than the kit, and every test after that is free.

What you'll need (not included)

The kit is the workflow, the prompts, and the setup guide. It does not include the tool subscriptions it runs on. You'll need your own Claude with Cowork account, and the results from whatever testing tool you already use. Claude is billed by the tool itself, not by us.

You'll also need your Claude account on Cowork's Pro plan or higher, so your agent can connect your tools and use the most capable model while it installs the workflow.

Not on Claude? The kit is plain files that describe what to achieve rather than which buttons to press, so a capable AI agent on another platform, like ChatGPT Work, can follow the same steps. If anything doesn't map to your setup, email support and we'll get you working.

How to install this kit

Open the kit in Claude Cowork and run the included prompt with your visitor and conversion counts per arm; the analysis runs inside Cowork, so there's nothing extra to connect. Step-by-step instructions ship inside every kit, so you're never guessing. See How Self-Install Kits Work for the full walkthrough.

Install once, not every month

If you've been pricing A/B testing and analytics software, this is the install-it-once alternative. You set it up yourself with your own AI agent and keep it, with no monthly seat. Add it to your cart, open it in Claude Cowork, and settle the test you've been sitting on today.