What Experiment Should I Run?
“What should I even test first?” is the right question — and the answer is a decision tree, not a list of best practices. Tell the tool your goal, where your funnel leaks, your traffic, and your research, and it recommends the specific experiment to run next, and why.
The right next experiment depends on where your funnel leaks, how much traffic you have, and whether you have research. This recommender maps those to a specific test — the lever to change, the funnel step to target, and the test type that fits your traffic — and, crucially, tells you when to do research instead of testing. It encodes the triage RGM runs on a new account; runs entirely in your browser.
What Experiment Should I Run? inputs and result
How to use this tool
- Answer the four questions.Your goal, where the funnel leaks most, your traffic level, and the research you already have.
- Read the recommended experiment.The tool maps your situation to a specific test — what to test, where, and which test type fits your traffic.
- If it says “research first,” do that.Testing without research wastes scarce traffic on low-win-rate guesses. The biggest unlock is often not a test at all — it’s a week of analytics, recordings, and a survey.
- Right-size and design the test.Use the linked sample-size and duration calculators, write a structured hypothesis, and pre-commit one primary metric with guardrails.
- Export the plan.Copy a share link, download the CSV, or print a one-page PDF for your test brief.
RGM Expert Says
The question we hear most from teams new to CRO isn’t ‘how do I test’ — it’s ‘what should I even test first?’ The honest answer is a decision tree, not a list of best practices, because the right next experiment depends entirely on where your funnel leaks, how much traffic you have, and whether you’ve done any research. This tool encodes the same triage we run on a new account.
Two answers surprise people. First, if you have no research, the recommendation is not a test — it’s a week of analytics, session recordings, and a survey, because testing blind produces the ~1-in-3 win rate that makes teams quit. Second, if your traffic is low, we steer you toward bold, high-leverage swings (value proposition, full-page redesigns) rather than tweaks, because small traffic can only ever detect large effects — a 2% button test is mathematically doomed on low volume.
The throughline is leverage: test the step with the most traffic and the biggest drop-off, change the argument (value prop, clarity, friction, proof) rather than the decoration, and match the test type to your traffic. Get those three right and your first experiments actually move something — which is what earns the program the credibility to keep going.
How it works
The recommender applies the prioritization logic from the RGM CRO program:
- Research gate — no research → research first; win rate is decided before the test.
- Leverage — test the high-traffic, high-drop-off funnel step you flagged.
- The lever — map the leak to what actually moves behavior: entry → value proposition; consideration → clarity & proof; conversion → friction; retention → lifecycle (not a page test).
- Test type by traffic — low → bold A/B swings; medium → standard A/B (A/B/n for distinct concepts); high → A/B/n, multivariate, or bandits for short-lived calls.
It’s a heuristic starting point, not a substitute for your own research. Runs entirely in your browser. Pair it with the sample-size and duration calculators.
Most programs test the wrong thing first
The single most common reason a young CRO program stalls is starting in the wrong place: tweaking a low-traffic checkout step while 80% of visitors bounce off the landing page, or testing button colors when the value proposition is unclear. Every test you run is several you didn’t — so the first question isn’t ‘what’s a good test idea,’ it’s ‘where is the leverage, and am I ready to test at all?’
This tool exists to force that triage. By tying the recommendation to your funnel leak, traffic, and research, it pushes you toward the high-leverage step and the behavior-changing lever, and it refuses to recommend a test when you have no research to base one on. That single discipline — research before testing, leverage before tweaks — is most of the gap between a program that compounds and one that spins.
Use it as a starting map, then sharpen with your own data. The recommendation tells you the category of experiment that fits your situation; your analytics and qualitative research turn that into the specific hypothesis worth its slot in the backlog.
The lever for each funnel leak
Match the change to where the leak is — and change the argument, not the decoration.
| Leak location | Highest-leverage lever | Avoid |
|---|---|---|
| Entry (landing/ads) | Value proposition & message match | Button colors, hero swaps |
| Consideration (product/content) | Clarity, info hierarchy, social proof | Cosmetic restyling |
| Conversion (cart/checkout/form) | Friction: fields, steps, trust | Adding more upsells |
| Retention (post-purchase) | Onboarding & lifecycle experiments | One-off page A/B tests |
What operators say
The hard part is testing the right things — having the right treatment — not setting up the tests.
If you have no research, your best next move usually isn’t a test — it’s a week of analytics, recordings, and a survey.