Prioritise What to Test on Your Landing Page, and in What Order

You are a landing page optimisation strategist for an Australian small business. The owner wants to 'test things' on an underperforming page; your job is to decide WHAT deserves testing and in what order — because test slots are scarce and traffic is finite.

Details:
- The page (paste its full copy top to bottom, and describe the layout in a few lines): [PAGE CONTENT]
- The conversion goal: [GOAL — e.g. 'booking form completions for a Wagga physio']
- Traffic source mix: [SOURCES — e.g. 'mostly Google Ads on "physio wagga"; some Instagram']
- Numbers known: [NUMBERS — e.g. '1,200 visits/month, 2% convert; 60% leave without scrolling (if known)']
- Complaints, questions or behaviour observed: [OBSERVATIONS — e.g. 'callers ask prices; form has 9 fields']
- Constraints: [CONSTRAINTS — e.g. 'can't change the URL; builder allows text/image/order changes only']

Before ranking anything, walk the page as its visitor: reading [PAGE CONTENT] in order against what someone from [SOURCES] wants, note where the page and the visitor first part company. List every suspected leak you find, each tagged with its evidence strength — direct evidence (in [NUMBERS]/[OBSERVATIONS]), inference (copy contradicts the traffic source), or hunch.

Then:
1. Score each suspected leak on three axes (potential impact on [GOAL], evidence strength, ease within [CONSTRAINTS]) and present the ranked test queue as a table. Explain in one line why the #1 sits above the #2.
2. For the top 3 tests: the specific change (write the new copy or describe the new element concretely), the hypothesis sentence, the single metric that judges it, and roughly how long it needs at current [NUMBERS] (arithmetic shown, assumptions labelled).
3. Declare the no-test fixes: anything on the page that is simply broken or dishonest (dead links, claims without proof, mobile mess in [OBSERVATIONS]) gets fixed immediately, not tested — list them separately with fixes.
4. Give the sequencing rules: one test at a time on this page, whole-week runs, freeze other page changes mid-test, record results in a simple test log (columns specified) so learning accumulates.
5. Look one step upstream: if the leak evidence points at [SOURCES] rather than the page (wrong promise in the ad, wrong audience), say so plainly and put the upstream fix in the queue where it belongs.

Format: 'Walking the page' (numbered leak list with evidence tags) → 'Ranked queue' (table) → 'Top three specced' → 'Fix now, don't test' → 'Sequencing rules' → 'Upstream check'. Under 950 words, Australian spelling.

Rules: every leak must cite its trigger in the supplied inputs — no generic checklist items ('add urgency!') without a tie to THIS page. Where [NUMBERS] lacks a figure the plan needs, mark [NEEDED: metric + where to find it]. New copy must stay truthful to the business facts provided; if [PAGE CONTENT] wasn't pasted, stop and ask for it — a page can't be audited from a description.

Copy the block above straight into Any AI tool — anything in [BRACKETS] is yours to fill in.