Experiments — how it works

What this page is for

Experiments is the portfolio of every CRO change Levri has suggested for you, plus any you have added manually. It tracks each one as a card moving through five stages: Backlog, Prioritised, Design, Running, and either Won or Lost at the end.

If Analyses tells you what is broken and Funnels tells you where you are losing people, Experiments is where you decide what you are actually going to ship and when.

Where experiments come from

Most experiments are generated automatically. When Levri analyses a page or scores a funnel, the issues it finds become experiment suggestions. The source chip on each row tells you where it came from:

• Page (Analysis). From a page analysis. Targets a single page issue.
• Diagnose. From a funnel run. Targets a step-level health problem.
• Explain. From a funnel run. Targets a drop-off between two steps.
• Compare. From a Compare run. Targets a structural gap vs another page.

You can also add experiments manually if you have a hypothesis Levri did not surface.

Reading the priority columns

Each experiment row shows four numbers that together answer "should I ship this?":

1. 01
   Opportunity score (0–100)

   Levri's estimate of how big a win this could be. Higher is better. Combines severity of the underlying issue with the size of the audience that sees it.
2. 02
   Est. lift (%)

   Expected conversion lift if the experiment wins. A 4% lift on a high-traffic page is usually more valuable than an 8% lift on a low-traffic one.
3. 03
   Revenue ($)

   Rough dollar estimate of the lift, using your revenue inputs (or a sensible default if none are set). Treat it as a relative comparison tool, not a forecast.
4. 04
   Effort (Low / Medium / High)

   How much work to design, build, and run the experiment. Copy changes are Low. Layout reworks are Medium. Architectural changes are High.

Moving experiments through the pipeline

Every experiment has a Status dropdown you can change inline. The expected flow is:

1. Backlog. Suggested but not committed to.
2. Prioritised. You are committing to running this soon.
3. Design. Being designed or built right now.
4. Running. Live on your site, gathering data.
5. Won or Lost. The test concluded. Won means it beat the control. Lost means it did not.

You do not have to use every stage. Many teams skip Design and Prioritised and only use Backlog, Running, Won, Lost. Use the stages you find useful and ignore the rest.

Portfolio view vs Board view

Portfolio is the default table view. It is best for sorting, filtering, and looking across the whole pipeline at once.

Board view is a Kanban layout grouped by status. It is better for tracking what is in flight right now. Drag a card between columns to change its status.

Top experiments highlights

At the top of the page you will see four highlighted experiments. These are the highest priority items based on opportunity score and confidence, normalised by effort. Treat them as your shortlist for what to ship next.

Common new-user questions

• Does Levri run the experiment for me? No. Levri suggests what to test and ranks the suggestions. You (or your team, or your A/B testing tool) ships the change and runs the test.
• What does "Won" actually mean? You decide. Most teams use it for any experiment where the variant beat the control with statistical significance. Some teams use it for any experiment they kept live. Both are fine, just be consistent.
• Do I need an A/B testing tool? Not necessarily. If you ship the change live with no control group, you can still mark the experiment Won or Lost based on whether the page metric improved. An A/B testing tool gives you a cleaner signal but is not required.
• What is the difference between opportunity score and confidence? Opportunity is how big the win could be. Confidence is how sure Levri is the win will happen. High opportunity plus low confidence is a bigger bet. High confidence plus low opportunity is a safer but smaller bet.
• Can I un-Win an experiment? Yes. Status is mutable. Levri keeps an audit trail so changing from Won back to Running (or Lost) is logged.
• Why is the revenue number sometimes blank? You have not set revenue inputs for that site yet. Add monthly traffic, conversion rate, and average order value in the site settings and the dollar estimates start filling in across all experiments for that site.