ChiefOSby ConversintStart free
Menu
← Back to the library
Marketing and creative direction

Landing page review

Audit, create, and improve Conversint /start/... paid-ad landing pages. Use this skill whenever the user mentions start pages, ad landing pages, Meta/Facebook/Instagram campaign pages, lead-magnet pages, claim pages, page conversion, landing-page copy for ads, or asks whether an ad page is good enough to run, even if they do not explicitly say "review."

Works with the context you provideVersion 2.0.1Historical revision

# Start Page Review

Use this as the quality gate for Conversint paid-ad landing pages, especially `https://conversint.ai/start/<slug>` pages. The purpose is not to make a prettier page; it is to make sure the page can turn paid traffic into a useful signal or a real lead.

Class and Boundaries

  • Skill class: marketing execution plus website quality gate.
  • This skill may inspect and edit the Conversint landing-site repo when the user asks for implementation and the repo is available.
  • This skill does not publish ads, change budgets, pause campaigns, or create ad spend. Route those decisions through `/fb-ads` or `/paid-ads`.
  • This skill does not rewrite global tracking architecture from scratch. Route missing events, consent behavior, GA4, Meta Pixel, CAPI, UTMs, or source capture to `/analytics-tracking`.
  • This skill does not silently write CRM, Gmail, Drive, Calendar, or project ledgers. CRM writes belong to `/crm` or `lib/crm.py`; Gmail/email work belongs to `/gmail`; Drive work belongs to `/gdrive`; durable project tasks belong to `/project-manager`.
  • If this skill saves a durable Markdown audit or implementation note, follow `./references/dependencies/conversint-current/references/output-paths.md`. Most quick reviews can stay in the chat; website code changes are committed in the website repo, not as ChiefOS deliverables.

Required Context

Load these before any material audit, new page, or page rewrite:

  1. `./references/dependencies/conversint-current/references/meta-ads-client-acquisition.md`
  2. `./references/dependencies/conversint-current/references/marketing-message-governance.md`
  3. `./references/dependencies/conversint-current/research-thought-leadership/conversint-messaging-optimization/active-message-catalog.md` when the page serves a testable message
  4. `./references/dependencies/conversint-current/references/paid-ads-audit-engine.md`
  5. `./references/dependencies/conversint-current/references/facebook-ads-playbook.md`
  6. `./references/dependencies/conversint-current/references/digital-marketing-funnel-best-practices.md`
  7. `./references/dependencies/conversint-current/references/org/positioning.md`
  8. `./references/dependencies/conversint-current/references/org/anti-patterns.md`
  9. `./references/dependencies/conversint-current/references/output-paths.md` when saving any durable review or note
  10. `./references/dependencies/conversint-current/references/conversint-brand.md` when copy or visual design will be changed
  11. `./references/dependencies/conversint-source/.claude/skills/copywriting/SKILL.md` when writing buyer-facing copy
  12. `./references/dependencies/conversint-source/.claude/skills/analytics-tracking/SKILL.md` when checking events or source capture
  13. `/website-feedback` and `/vercel-feedback` when the page has Vercel Toolbar comments or other stakeholder feedback

When the page is tied to an active campaign, inspect fresh evidence when available:

  • Meta campaign/ad/ad-set metrics from `/fb-ads` tooling or current exported report.
  • GA4 journey data, including `/start/*` page views, CTA clicks, form starts, submit attempts, successful submits, and downstream page paths.
  • Postmark contact-form messages to `hello@conversint.ai`; this is the source-of-truth lead count for Conversint landing pages.
  • Current landing repo implementation, usually `~/Dev/Conversint/conversint-landing/src/lib/adLandingPages.ts`, `app/start/[slug]/page.tsx`, claim/form components, and tracking helpers.

If the review uses public research, platform documentation, or competitor examples, cite sources inline and include a short Sources section. Internal Meta, GA4, Postmark, and repo evidence should be named with dates, campaign/page identifiers, and file paths rather than treated as public citations.

Stakeholder Feedback Gate

Before changing or approving a page that has Vercel Toolbar comments, run the `/website-feedback capture` and `triage` flow for `conversint-site`. Use `/vercel-feedback` only for the Vercel source adapter. Treat each distinct comment as a linked acceptance criterion with its original wording, page path, thread ID, and implementation pointer. Do not close or mark a comment done from code changes alone; prepare the `dev` preview and keep it in review until the stakeholder approves it.

If Vercel comments cannot be fetched because the connector or Toolbar is unavailable, say so explicitly and ask for the preview URL or copied comment text before claiming the feedback has been addressed. Never infer the comment content from a screenshot, page metric, or prior task.

Operating Principle

Every start page needs one clear job:

```text qualified paid visitor -> understand the pain -> complete a useful first step -> see a result -> optionally request contact -> qualify the conversation ```

Do not approve a page that asks for a meeting or email before giving value. Replace abstract artifacts such as "map," "assessment," or "diagnosis" with a bounded exercise, worksheet, or reflection that gives the visitor a practical next step first.

Do not approve or generate these abstract CTA patterns: "Send Me The First-Pass Map," "Get the Map," "Claim your ... map," or any equivalent promise that hides the useful result behind contact. Name the actual exercise, worksheet, reflection, or result instead.

For paid pages, use one primary mobile action. Remove main-site navigation, newsletter signup, generic contact links, and competing CTAs unless they are required privacy access or a deliberate secondary trust/retargeting path. Cookie consent must not cover the first action.

Review Workflow

1. Identify the Cell

Name the page's campaign cell:

  • Persona or buyer: usually SMB owner/operator, CEO, COO, or CTO.
  • Value family: operational overload, software/vendor relationship, AI visibility, or another documented family.
  • Manifestation: the concrete version of the pain.
  • Give-first: the artifact the visitor receives.
  • CTA: the action phrase used in the ad and on the page.

If the cell is vague, fix the cell before fixing layout.

2. Score the Page

Use this 100-point scorecard:

| Area | Weight | Pass Standard | |---|---:|---| | Message match | 20 | The ad promise, page hero, form panel, and fulfillment note say the same thing. | | Offer clarity | 20 | The visitor can tell exactly what they will receive and why it is useful. | | Conversion path | 20 | CTA is visible, form is easy, no meeting-first ask, mobile path is obvious. | | Trust and friction | 15 | Low-pressure language, privacy path, no overclaiming, no jargon, clear next step. | | Tracking readiness | 15 | Page, CTA, form start, field behavior, submit attempt, success, and lead delivery are tracked. | | Implementation quality | 10 | Page is static/cacheable where possible, responsive, accessible, fast, and visually coherent. |

Decision bands:

  • 90-100: launch-ready or continue running.
  • 75-89: usable with small fixes.
  • 60-74: repair before scaling or changing budgets.
  • 40-59: do not send new spend until the main blocker is fixed.
  • Under 40: pause traffic or rebuild the page.

3. Audit the Copy

Check these in order:

  • The H1 names an owner-recognizable pain, not a service category.
  • The first paragraph makes the pain specific enough to self-qualify the buyer.
  • The CTA promises the give-first artifact, not a call.
  • The first action starts the useful experience without requiring email.
  • The form prompt says what information to provide and what will happen next.
  • The "what you get" bullets are specific to the page's pain.
  • The language avoids banned terms from `./references/dependencies/conversint-current/references/org/anti-patterns.md`, especially "Readiness Assessment," "Strategic Workshop," "Visual collaboration," "leverage," and "synergy."
  • The page says enough to create trust but not so much that the visitor has to study before submitting.
  • Any exercise or worksheet uses deterministic routing and shows practical steps before optional contact actions.
  • "Other" free text is clearly marked private, stays in the browser, and is never sent to GA4 or Meta automatically.

4. Audit the Funnel

Walk the visitor journey:

  1. Page loads from ad URL with UTMs preserved.
  2. The first CTA starts the useful exercise or worksheet without email.
  3. The visitor can complete the steps and see a bounded result.
  4. Practical result actions work before contact is requested.
  5. Optional contact submission carries only required contact data plus non-PII exercise/result IDs.
  6. Successful submit triggers browser tracking and server/contact delivery.
  7. Postmark sends the contact email with page, source, UTM, and non-PII result context.
  8. A human can fulfill the requested worksheet quickly from the submitted context.

Do not restart paid traffic after a page change until the deployed production URL has passed the full mobile journey: start, complete, result, copy/email action, optional contact submit, GA4/Meta payload review, and Postmark delivery.

If the evidence shows landing-page views but no CTA or form starts, prioritize offer/page clarity. If there are form starts but no submits, prioritize form friction. If there are submits but no Postmark leads, prioritize tracking/contact delivery.

5. Visual and Mobile QA

Before approving or committing page implementation:

  • Check desktop and mobile layouts.
  • Confirm text does not overlap, truncate, or overflow buttons/cards.
  • Confirm dark-mode and hover states preserve contrast.
  • Confirm the form panel and modal are not too large on mobile.
  • Confirm images load and are meaningfully tied to the page promise.
  • Confirm the page does not become dynamically rendered unless the page truly requires request-time data.
  • Test the first action at 390px before testing desktop polish.
  • Confirm cookie consent never obscures the first action or result controls.

6. Learning Loop

For every material page change, prepare a learning-loop entry:

```markdown ## Learning Loop Entry - Date: - Page: - Campaign/ad cell: - Hypothesis: - Change made: - Evidence before: - Metric to watch: - Review window: - Ledger path: research-thought-leadership/conversint-messaging-optimization/marketing-learning-ledger.md ```

Route the write through `/marketing-learning-loop` when the user wants the ledger updated, or update the ledger directly only if that has already been the active workflow in the current session.

Output Format

When reviewing, answer in this structure:

```markdown ## Start Page Review - Page: - Score: - Decision:

## Main Finding - First blocker: - Why: - What to change:

## Scorecard | Area | Score | Evidence | Fix |

## One Change Next - Change: - Why this first: - Metric to watch: - Review date:

## Handoffs - Ads: - Tracking: - Learning loop: ```

When implementing, keep edits scoped to the page data, page component, form/tracking helpers, and necessary styles. Build or run the nearest available verification command before reporting done.

Required Exercise Measurement

When a page contains an exercise or worksheet, require these events or their established equivalents:

| Stage | Event | Required non-PII fields | |---|---|---| | Start | `exercise_start` | `exercise_session_id`, `offer_id`, `landing_slug`, `cta_id`, `cta_location` | | Step | `exercise_step_view`, `exercise_step_complete` | `exercise_session_id`, `exercise_step`, `offer_id` | | Result | `exercise_complete`, `result_view` | `exercise_session_id`, `result_id`, `result_category`, `offer_id` | | Worksheet | `worksheet_copy`, `worksheet_email_request` | `exercise_session_id`, `result_id`, `result_category` | | Contact | `form_start`, `form_submit_attempt`, `generate_lead` | `exercise_session_id`, `result_id`, `result_category`, source/UTM context |

Never send email addresses, names, free-text answers, or private exercise responses in analytics or ad-platform parameters. Register event-scoped GA4 custom dimensions for the stable categorical fields before relying on them in reports.

Eval Prompts

Save and maintain runnable eval prompts in `./evals/evals.json`. These prompts should test whether the skill:

  • reviews a live `/start` page using ad, GA4, and Postmark evidence;
  • creates a new `/start/<slug>` page from an ad hypothesis without a meeting-first CTA;
  • catches tracking or form-delivery gaps before recommending ad changes.