We graded our own homework — and found seven failures
We sell evaluation. "Measured, not vibes" is on the homepage. So at some point the obvious, uncomfortable question lands: have we run that rigour on our own site? We hadn't. So we did.
What the audit caught
An automated pass over every page found two things we'd shipped without noticing. Seven listing pages
had no <h1> at all — their title was rendering as a sub-heading, so the page had no top-level
topic for a search or answer engine to grab. And a batch of titles ran past the length a search result
will show, getting truncated mid-thought.
The deepening underneath it
The audit sat on top of a wider GEO pass: titles rewritten to lead with the message and fit the result; a breadcrumb trail emitted on every sub-page; and a per-service FAQ whose accordion and structured data read from one source, so they can never drift apart.
Before the audit
- Seven pages with no h1
- Titles past the ~60-char limit a search result shows
- FAQ schema maintained by hand
After
- Exactly one h1 per page
- Every title ≤ 60 characters, every description ≤ 160
- FAQ accordion and schema from one source
The corroboration we'd promised ourselves
We also added Crunchbase to the entity's sameAs — a third party confirming who "Pangaea Labs" is. An
earlier entry said that corroboration was still missing; this is us closing the loop.
Do
- Run your own evaluation tooling on yourself
- Generate structured data from visible content, one source
- Let a third party corroborate the entity
Don't
- Assume "looks fine" means "passes"
- Maintain parallel copies of the same facts
- Claim rigour you won't apply to your own work
You don't get to say "evaluated, not vibes" if you won't point the evaluation at yourself first.
Sources