Skip to content
SiteRevenue
Log inAdd the free bot
LIGHTHOUSE VS A CONVERSION AUDIT

Lighthouse 100, conversion 0

Published:

A Lighthouse 100 does not mean the page converts. Core Web Vitals measure load, interactivity, and layout shift — not whether a paid click can see the offer and one primary action. SiteRevenue does not collect Core Web Vitals and will not invent them. Speed still matters. It is not the conversion audit. Evergreen comparison: Lighthouse vs a conversion audit.

What Lighthouse actually scores

Lighthouse and Core Web Vitals are good at a specific job: is this page a reasonable citizen of the web? Largest Contentful Paint, Interaction to Next Paint, Cumulative Layout Shift — those names describe load and stability, not purchase intent.

A page can ship a 100 because:

  • The hero is a compressed image and almost no copy.
  • The primary action is a ghost link in the footer, so nothing “long” blocks the main thread.
  • Third-party tags were stripped on the lab URL you tested, not on the URL ads hit.
  • The fold is a beautiful video poster that never states who the product is for.

None of those failures show up as a red CWV. They show up as a paid click that cannot answer three questions in about five seconds: what is this, who is it for, what do I do next? That is hero clarity, not a performance budget.

Performance work is still real work. Slow, jumpy pages waste money. A conversion audit that invented a Lighthouse number to look complete would be lying. We do not do that. The mobile conversion guide is explicit: we will not invent Core Web Vitals.

Conversion is a different job

Conversion on a public marketing page is a path:

  • The arriving promise (search, ad, or the job they thought they clicked) still exists in the H1.
  • One filled primary CTA finishes that job. A second equal button splits the click — CTA hierarchy.
  • Price or an honest sales path is findable — offer and pricing clarity.
  • Proof sits next to the claim, not three scrolls down, and is not invented — social proof that converts.
  • The form asks for the minimum and does not silently fail — form friction.
  • Trust is a real policy or cancel path next to the ask, not a lock icon — trust and risk reversal.
  • On a phone, offer + one CTA fit the first screen. Dual sticky bars clipped by iOS Safari are a conversion bug even when Lighthouse is green.

That list is the conversion leak checklist. A 100 on performance does not tick those boxes. A 60 on performance with a clear fold can still take the sale. Both can be true at once. They are not the same score.

Lab tools also lie about which URL you measured. Operators test the cached marketing homepage on a fat pipe, then send spend to a builder landing URL stuffed with chat widgets, cookie stacks, and a hero image taller than the viewport. The Lighthouse run was honest. It was honest about the wrong page.

SiteRevenue scores the public URL you paste. Optional outcome and competitor URLs frame the job (full reports fetch those public pages; fail-open). We do not connect Google Ads. We do not crawl a private dashboard. If the live HTML is a third-party widget with no offer in the document, that is a finding — the page may also “pass” a speed lab because the widget failed to load.

The 100 / 0 pattern (illustrative, not a case study)

You will recognize the pattern even without a named brand. This is a composite of public-page shapes, not a verified customer story and not a promised lift.

  • Lighthouse looks done. Images are sized. Fonts are subset. CLS is quiet.
  • The fold is a slogan. Product name, atmospheric photo, two equal pills: “Learn more” and “Talk to sales.”
  • The form eyebrow switches the job. The ad said “start free.” The form says “book a product tour.”
  • Mobile hides the ask. A dual sticky bar sits under the iOS home indicator. The actual submit is below a full-viewport video.
  • Proof is a logo strip in the footer that does not support the claim in the hero.

A speed tool will congratulate you. A buyer who paid for the click will leave. The leak was never LCP.

The sample report scores an illustrative Mobile 7/10 because a fixed dual-button bar clips under Safari chrome — effort M, labeled as a wireframe, not a live screenshot of a customer. That is the kind of miss a 100 does not catch.

If your only dashboard is PageSpeed Insights, you will keep polishing the wrong surface. If your only dashboard is a roast thread, you will keep collecting vibes. The missing artifact is a file of record that ranks conversion leaks on the URL you actually bought.

What SiteRevenue measures (and will not invent)

A full report is a conversion file, not a CWV file:

  • Letter grade (A–F) — overall conversion health. A is healthier than F. Same model every run so re-audits stay comparable.
  • Leak score — on a live report, higher = more leaks found. Do not treat it as a “health” ring.
  • Dimension scorecard — modules 0–10, higher = healthier on that slice. New full jobs emit Hero, CTA, Message match, Offer, Proof, Form, Trust, Mobile (Message match omitted only when there is no ad copy). Older jobs stay omit-safe.
  • Page captures — fail-open. A missing shot is not a failed audit and not a fake Lighthouse clip.
  • Action box — fix these 3 this week, with effort S/M/L.
  • Findings — typically 5–8 with evidence and a fix. That is a quality target, not a refund SLA.
  • Dollar ranges — estimates for prioritization, not guarantees.

We do not emit Core Web Vitals. We do not scrape CrUX. We do not put a invented “Speed: 94” in the scorecard to look thorough. If a finding mentions slowness, it is because the public HTML or the path is visibly hostile (blocked offer, clipped sticky, hover-only nav) — not because we ran your lab profile.

Read the stack in order: How to read a revenue audit. Teaser omits the full modules. Unlocking the $19 page file starts a new full job that captures the public page again. It does not reuse the teaser scrape.

Ship the fold, then measure speed

The honest sequence for teams who pay for traffic:

  1. Make the job obvious on the phone fold. Offer + one primary. HTML, not only a widget. Walk mobile conversion.
  2. Align the arriving promise with H1, CTA, and form eyebrow — message match. Message match is usually cheaper than another week of bid rules.
  3. Ship the action box, then re-audit the live URL for a letter-grade strip. Empty re-runs waste a credit.
  4. Then spend the performance sprint on the URL that already has a clear path. A faster mismatched page is a faster bounce.

Speed and conversion are complementary. Collapsing them into one vanity 100 is how operators get a green dashboard and a red CPA.

Public pages only. Not a chatbot. Not an accessibility certificate. Agent job webhooks are later — poll if you automate. None of those caveats change the thesis: CWV ≠ conversion.

More notes: A roast is not a system · Message match is the real ad waste · Re-audit or you are guessing · All posts