How do you decide what to test first on a CRO backlog?
Score ideas by impact, confidence, and effort, and fix the broken things before testing.
Score every idea on three things, impact, confidence, and effort, then start with the highest scoring idea on the highest traffic page closest to the money. Before any of that, fix what is broken, because a broken thing is a repair, not an experiment.
Fix these before you test anything
- Conversion tracking that is missing, double counting, or firing on the wrong event.
- A form that does not create a CRM record, or creates one with no owner.
- A page that fails Core Web Vitals on mobile, where Google's threshold is a Largest Contentful Paint of 2.5 seconds or less at the 75th percentile.
- Anything actually broken: a dead link in the nav, a button that does nothing on iOS, a form that errors on a valid phone number.
None of these need a hypothesis. They need a ticket. Testing on top of broken measurement produces confident nonsense.
The scoring
Rate each idea 1 to 10 on three axes and multiply or average them:
- Impact. How much could this move the number, considering the page's traffic and its distance from revenue? A change on a page with 10,000 monthly visits that sits one step from a quote beats a clever idea on a page with 200 visits.
- Confidence. How strong is the evidence? Scroll data showing nobody reaches the form is strong. A blog post saying green buttons work is not.
- Effort. Developer days, design time, and approvals. A copy change you can ship today outranks a redesign you can ship in six weeks, even at lower impact, because you get the answer sooner.
Write the score down. The point of scoring is not precision, it is forcing a conversation about evidence and making the order defensible to whoever asks why their idea is fourth.
Sequence and cadence
Work top down, one meaningful change at a time per page so you can attribute the result. Expect most changes to land flat. That is normal, and the losers are still information as long as you record what you learned.
Review the backlog monthly against real data. New evidence should reorder it, and an idea that has sat at the bottom for six months should be deleted rather than carried forever.
Ongoing work like this is bought as a retainer because the value is in repetition. Ours run Level 1 at 10 hours a month for $1,900, Level 2 at 15 hours for $2,775, Level 3 at 20 hours for $3,500, and Level 4 at 40 hours for $6,750, Net 15 with 30 days notice to change or cancel.
Related: What is conversion rate optimization?, How do you A/B test a landing page?, Does LevelUp Digital offer retainer services?, and our retainers page.