Preflight the page before discussing CRO
A landing page must first be reachable, accurate and measurable. Test the final URL and tracking parameters in Google Ads, then open it on a real mobile device. Google’s landing-page test guidance covers URL mismatches, parameters, crawl access and HTTP errors.
| Check | Evidence | Fail condition |
|---|---|---|
| Offer continuity | Query, ad and first screen captured together | Price, product or promise changes after click |
| Functional path | Recorded mobile form/checkout test | Error, hidden CTA, impossible input or broken payment |
| Measurement | Unique test ID reconciled | Missing, duplicate or wrong-value conversion |
| Performance | Field and lab data by template | Material delay or instability on the decision path |
| Policy/trust | Visible business, price and privacy information | Misleading claim or missing material condition |
Audit the complete promise
Message match is more than repeating a keyword. Compare audience, problem, product, price, incentive, location and next step. “Free audit” leading to a generic agency homepage fails because the action and terms disappeared. A Shopping click should land on the advertised item or a defensible equivalent, with price and availability consistent.
Put decision information where doubt occurs
For lead generation, readers usually need scope, fit, evidence, process and what happens after submission. For ecommerce, they need product identity, variants, price, delivery, returns, proof and stock. Place proof beside the claim it supports. A logo strip does not prove a specific outcome.
Use funnel evidence to locate friction
Measure page view, meaningful interaction, form start, validation error and completion, or view item, add to basket, begin checkout and purchase. A drop is a location, not a cause. High form abandonment could mean excessive fields, but also a technical error or poor traffic. Google Analytics’ funnel exploration supports open or closed funnels and segment comparisons; understand those definitions before interpreting rates.
Turn findings into tests
Write: “Because paid mobile visitors from query group X start the form but fail at the postcode field at twice the desktop rate, we believe mobile validation is blocking completion. Fixing validation should increase valid submissions without reducing qualified rate.” That is a hypothesis. It becomes evidence only after QA and a credible comparison.
- Choose one primary business metric and guardrails such as qualified rate or refund rate.
- Change one decision mechanism, not an entire redesign.
- Estimate required sample from baseline rate and minimum worthwhile effect.
- Run across representative weekdays and conversion lag.
- Predefine win, loss and inconclusive outcomes.
Google Ads custom experiments can test Search landing pages by splitting traffic. Google recommends a clear hypothesis and warns against changing the base during the test. Where platform experiments are unavailable, use a suitable testing tool with stable allocation and server-side outcome checks.
Do not optimise a page for Quality Score alone
Google calls Quality Score a diagnostic. Landing-page experience is one component alongside expected CTR and ad relevance. Improve usefulness and relevance for the visitor; do not claim that a design change guarantees a particular CPC reduction.
Lead-generation page checks
- State who the service is for and where it is available.
- Define the deliverable, process and any material exclusions.
- Show evidence relevant to the claim, not generic logos.
- Explain what happens after submission and expected response.
- Ask only fields needed for routing or qualification.
- Test validation, error recovery, spam controls and mobile keyboards.
- Send a lead ID into the CRM and reconcile its final status.
Fewer fields are not always better. Removing budget or service selection may increase raw conversion rate while burdening sales with poor-fit enquiries. Use qualified rate as a guardrail.
Ecommerce page checks
- Match advertised product, image, price, promotion and stock.
- Make variants, dimensions, compatibility and included quantity clear.
- Show delivery, returns, warranty and payment conditions before commitment.
- Test galleries, sticky add-to-cart, basket, voucher and payment paths.
- Pass a unique transaction ID, correct value and currency.
- Monitor refunds and contribution, not only purchase count.
For product-specific Shopping traffic, review the Shopping product-page CRO guide. A separate campaign page should not contradict the Merchant Center offer.
A ten-minute mobile QA script
Start from the actual ad preview or final URL, not the homepage. Load on mobile data. Accept and reject consent in separate sessions. Rotate the device, zoom text, open navigation, trigger an invalid form field, select a variant, add/remove basket items and complete a test payment or lead. Record the screen and note timestamps.
Check sticky elements do not cover fields or consent controls, keyboard type matches input, error messages remain visible, back navigation preserves state, and the thank-you page cannot accidentally refire conversion on refresh.
Prioritise findings
| Evidence | Reach | Action |
|---|---|---|
| Confirmed defect | High | Fix and regression-test immediately |
| Strong behavioural pattern | High | Research cause, then test |
| One anecdote | High | Collect corroborating evidence |
| Stylistic preference | Low | Deprioritise |
Page audit output
For every finding include affected URL and campaign, device, evidence, expected mechanism, commercial exposure, proposed change, owner and validation. “Move reviews higher” is incomplete. “Mobile visitors from non-brand Search reach the CTA but open the returns accordion before leaving; testing delivery and returns beside price may reduce uncertainty” is a testable hypothesis, not yet a fact.
Worked audit example
A hypothetical non-brand Search page receives 3,000 mobile clicks, 240 form starts and 72 completions. Desktop completes 45% of starts; mobile completes 30%. The initial hypothesis is mobile form friction, but QA shows the postcode field loses input when the keyboard closes and error logs confirm the same pattern. That is a defect, not an A/B-test idea. Fix it, regression-test it and compare the post-release cohort.
After repair, completion rises but qualified rate falls because the form no longer captures service area. The next decision is not to restore the bug. Add deliberate qualification or routing, and evaluate qualified CPA. Page conversion and business conversion are separate.
Evidence to gather before rewriting copy
- Actual query or audience themes and ad promises.
- Sales-call and support objections in customers’ words.
- Form errors, checkout failures and field abandonment.
- Differences by device, location and customer type.
- Competitor price and offer conditions.
- CRM quality and loss reasons.
Use this evidence to decide which uncertainty the page must resolve. Avoid manufacturing fake specificity or rewriting every paragraph when one missing eligibility condition causes the loss.
Handoff to development
Specify URL/template, affected campaign, screenshots, target devices, acceptance criteria, event changes, accessibility requirements, rollout, rollback and owner. QA staging and production. Preserve URL parameters and consent behaviour. After release, compare real outcomes with the baseline and document whether the original hypothesis was supported.
Final sign-off questions
- Can a new visitor explain the offer, audience, price or next step?
- Can the business evidence every material claim?
- Can mobile users complete and recover from errors?
- Does one unique outcome reconcile with the source system?
- Are downstream quality, revenue or refunds available as guardrails?
- Is there an owner and rollback plan?
If any answer is unknown, record it as an evidence task. A checklist should expose uncertainty, not convert it into a green tick.