Skip to content
SiteRevenue
Log inAdd the free bot
WHEN TO RE-AUDIT YOUR SITE

Re-audit or you are guessing

Published:

If you ship the action box and never run the same URL again, you are guessing. SiteRevenue stores version history on a site so letter grades stay comparable — v1 Grade D → v2 Grade B. A second full report is a new capture of the live public page, not an edit of last week’s file. Empty re-runs burn a credit. Ship, then re-audit.

Shipping is not evidence

The most expensive sentence in a growth meeting is “we shipped it, so it should be better.” A headline went live. A second button became a text link. The form dropped the company field. Someone pasted a screenshot in Slack. The ads dashboard is noisy this week — maybe that is the lift.

That is a story. Stories are how teams protect the week of work they already did.

A first full report is also a story until you close the loop. It is a snapshot of the public HTML on the day you paid. That file does not update when Webflow publishes. Park Solo, Park Agency, or Agency ($99/mo, 25 page credits) watch can email a mini grade when the public HTML hash changes. Watch is opt-in and off by default. It never spends a page credit. That is a signal, not a v2 letter. Improved / Dropped / Holding still needs a second full snapshot of the same URL, same model, new job. That is the only reason a strip like v1 Grade D → v2 Grade B can exist.

The when to re-audit your site guide is the rule sheet — when a second run is worth a credit, when it is not, how to read the strip without mixing letter grade and leak score. This post is the argument for why skipping that sheet is how operators stay confident and wrong.

Why the ads dashboard cannot close the loop

After a landing change, people open the auction. CPA moved. Or it did not. Either way they narrate.

The dashboard is a blender. Creative, bid, budget, daypart, audience, competitor, season, and the page all dump into one number. If you changed the H1 and raised the bid and launched a new RSA, you cannot tell which lever did what. A quieter week looks like a win. A sale-week competitor looks like a failed rewrite.

A letter on the same site is a narrower question: did the public conversion path look healthier to the same model after the HTML changed? That is not revenue. Dollar bands in findings stay prioritization estimates, not a forecast. It is still a better question than “how did the account feel.”

Confounding is why empty re-runs are also a waste. Re-running the identical live HTML “to see if the model changed its mind” does not remove the blender. It spends a credit on noise. The guide’s yes/no list exists so you do not have to invent a ritual. Use it.

The fear under the skip

Teams skip v2 for reasons that sound operational and are usually emotional.

  • The letter might not move. Then last week’s tickets were the wrong tickets — or they never made it to the URL ads still hit.
  • The letter might get worse. Then the rewrite was a mismatch, or the capture saw a widget fail. That is information. Pretending the first file is still current is not.
  • The client already saw v1. A second letter feels like reopening a closed sale. A file of record is supposed to reopen. See a roast is not a system.
  • Lighthouse is 100 now. Speed is not the conversion path. Lighthouse 100, conversion 0.
  • We’ll know from pipeline in a quarter. Pipeline is slower and more confounded than a public-page snapshot. You can do both. You should not do only the slow one.

Skipping v2 keeps the first report looking like a deliverable. A deliverable that cannot be checked is a roast with better typesetting.

What people call “we’ll know”

“We’ll know” usually means one of three substitutions:

  1. Anecdote. A sales call went well. Someone mentioned the new headline. N=1 dressed as a read.
  2. Account weather. Spend was efficient on Tuesday. Nobody can name the landing commit that caused it.
  3. Another roast. A new thread says the fold looks cleaner. No site identity, no obligation to come back.

None of those are v2. None of them tell you whether the old finding titles dropped. None of them survive a personnel change — the next contractor cannot open last month’s belief.

My Audits is the private library for that belief: versions, a strip when two full jobs exist, a re-audit when credits remain. Session URLs under /audits/:id stay noindex. A share link is read-only, not a public gallery. Recover with the checkout email at Open with email. The sample shows the chrome as illustrative, not a verified customer before/after.

The free teaser is a gut-check (titles only). It does not replace a full version when someone needs evidence-level diffs. Unlocking the $19 page file starts a new capture. It does not edit the last file and it does not reuse the teaser scrape.

Guessing is cheaper until it is not

A credit feels like the expensive move. Guessing is cheaper in the hour and expensive in the month: you keep buying traffic onto a page you have already congratulated. You brief the next sprint from memory. You tell the founder the fold is “fixed” because the ticket is closed in Linear.

The cheap sequence is still: ship the action box on the live public URL (cache actually cleared), then take the second full run. If nothing on the conversion path changed, do not re-run — that is the guide, not a vibe. If you launched a new campaign URL, do not bury it under the corporate homepage’s history.

Public pages only. Not a chatbot. Not a forecast. Agent jobs poll; completion webhooks are later. None of that replaces v2.

  • Matching guide: When to re-audit your site — the rule sheet for v1 letter → v2 letter, and when not to burn a credit.
  • See the open sample for the illustrative grade strip and action box.
  • Get a free grade on the live URL. If you already paid, recover via Open with email, ship, then re-audit for v2.

More notes: A roast is not a system · Lighthouse 100, conversion 0 · Message match is the real ad waste · All posts