--- name: audit-and-optimize-an-existing-landing-page description: >- Diagnoses why a live landing page converts below expectations and produces an evidence-backed, prioritized fix list, then verifies the fixes actually moved conversion rate. Reach for this when a page's CVR is disappointing, bounce is high on paid traffic, or CPA is over target even though click metrics look healthy - and always before scaling spend into a page you haven't inspected. VigilDog pulls the Google Ads and GA4 evidence (CVR, CPA, device split, bounce, engagement); the page edits themselves are human work. Do not use it to build a page from nothing (build-a-high-converting-landing-page), to repair a Below Average Landing Page Experience label specifically (improve-landing-page-experience), to run a controlled split test (run-a-landing-page-a-b-test), or as the recurring account-wide sweep (monitor-landing-page-performance). --- # Landing Page Audit and Conversion Fix Cycle ## Purpose A structured pass over one underperforming landing page: capture a metrics baseline, walk the page against a structural standard, locate where visitors drop off, apply the highest-leverage fixes, and prove the improvement with before/after data. The output is a documented fix log, not a redesign - the goal is attributable conversion lift. ## When to run - A page's conversion rate sits clearly below the account average or your vertical's norm. - Bounce rate on paid traffic to the page exceeds roughly 60%. - CPA is above target while CTR and Quality Score look fine - the click is working, the page is not. - You are about to raise budget on a campaign and its destination page has never been audited. - monitor-landing-page-performance flagged this page and routed it here. ## When NOT to run - No page exists yet, or the page is broken beyond patching → build-a-high-converting-landing-page. - The specific problem is a Below Average Landing Page Experience label in Quality Score → improve-landing-page-experience. - The page already meets baseline quality and you want incremental, proven gains → run-a-landing-page-a-b-test. - You want a recurring sweep across every landing page, not a deep dive on one → monitor-landing-page-performance. - Add-to-cart is healthy but purchases lag (ecommerce) → optimize-cart-and-checkout-flow. ## Prerequisites - Conversion tracking verified as accurate (a broken pixel makes every diagnosis wrong). - GA4 property collecting data for the page (bounce/engagement, ideally scroll events). - Access to edit the page, or a channel to whoever can. - Knowledge of which campaigns/ads send traffic to this URL. - At least ~30 days of traffic history for a stable baseline. ## Procedure 1. **Pull the Google Ads baseline.** Use `gads_run_gaql_query` on the `landing_page_view` resource for the last 30 days: clicks, cost, conversions, conversion value per final URL. Compute CVR and CPA for the audited page and for the account, so you know the gap you are closing. 2. **Split by device.** Use `gads_get_device_performance` for the campaigns feeding this page. Record mobile vs desktop CVR separately - a blended number hides most mobile problems. If regional behavior is suspected, add `gads_get_geo_performance`. 3. **Pull the behavioral baseline.** Use `ga4_run_report` (or `ga4_batch_run_reports` to fetch several at once) with the landing-page dimension plus device category: sessions, bounce rate, engagement rate, average session duration, and scroll/key events if configured. Save all numbers as the "before" snapshot. 4. **Capture the ad-side promise.** Use `gads_run_gaql_query` on `ad_group_ad` to list the headlines, descriptions, and final URLs of the ads driving this page. Note the offer and language the visitor was promised before landing. 5. **HUMAN STEP (outside VigilDog):** take full-page screenshots of the current page on mobile and desktop. These are the before-reference for every later comparison. 6. **Walk the page structure.** Compare the page against a standard persuasion sequence: a hero that states the value in one glance, outcome-framed benefits, credibility signals, customer proof, objection answers, a legitimate reason to act now, and one repeated call to action. Also check focus: navigation stripped, no links that lead away from the goal, one primary CTA. Score each element present/missing/weak. The page-reading itself can be done by the model from fetched page content; visual and layout judgments are a HUMAN STEP. 7. **Verify message match.** Compare step 4's ad language against the page headline and offer. For Search traffic the primary keyword theme should appear in the page headline; the offer in the ad must be visible above the fold. 8. **Diagnose drop-off** using the decision rules below - each symptom in the data points at a page section. Attach the metric evidence to every suspected issue. 9. **Prioritize.** Sort issues into a 2x2 of impact vs effort. High-impact/low-effort items are the first batch; high-impact/high-effort get scheduled; low-impact items are batched or dropped. 10. **HUMAN STEP (outside VigilDog):** apply the first batch of fixes on the page (headline rewrite, CTA placement, proof replacement, form trimming, image compression, nav removal). Deploy in batches, not all at once, so lift is attributable. 11. **Validate.** After the waiting gate in the decision rules, re-run steps 1-3 with the same date ranges and compare CVR, bounce, CPA against the snapshot. Log each change with expected vs actual effect and a keep/revert verdict. 12. **Route the outcome.** Improved → schedule the next fix batch or graduate to run-a-landing-page-a-b-test. Flat → the diagnosis was wrong, return to step 8 with fresh eyes. Worse → revert immediately and re-diagnose. ## Decision rules - **Bounce > 60% on paid traffic** → the top of the page fails: check message match, headline clarity, and load speed before anything else. - **Under ~30% of visitors reach mid-page** (scroll data) → the benefits section is not hooking; reframe features as outcomes. - **Visitors scroll fully but do not convert** → unanswered objections or a weak CTA; add guarantees, FAQ, and sharpen the action copy. - **Mobile CVR below ~50% of desktop CVR** → treat as a mobile-specific defect (tap targets, form usability, responsive layout), not a general page problem. - **High form-start but low form-complete** → too many fields or unclear next step; cut every non-essential field. - **Validation gate:** do not judge a fix batch before 50+ post-change conversions AND at least 2 weeks of data - earlier reads are noise. - **Fix ordering:** offer, hero, and objection handling before any cosmetic work; design polish rarely moves CVR on its own. ## Common failure modes - **Shipping every fix at once.** You get a number that moved and no idea why. Batch changes and log each batch separately. - **No baseline saved.** Without the step 1-3 snapshot there is no before/after and the audit proves nothing. Snapshot first, always. - **Calling the result at day 3.** Small samples produce fake wins and fake losses. Respect the 50-conversion / 2-week gate. - **Auditing desktop only.** Most paid traffic is mobile in many verticals; a page can pass every desktop check and still be the CPA problem. - **Polishing a page that sells a weak offer.** If the value proposition itself is unconvincing, no structural fix will save it - escalate the offer question to the business owner. ## Related skills - **Before:** monitor-landing-page-performance (flags which page deserves this audit). - **After:** run-a-landing-page-a-b-test (structured testing once the fundamentals pass). - **Instead, when applicable:** improve-landing-page-experience (Quality Score label repair), build-a-high-converting-landing-page (full rebuild), optimize-cart-and-checkout-flow (post-add-to-cart leaks).