54.6% fewer 404s, at zero cost per run
- BRP
- Global Digital Transformation Specialist
- 1 month, proposal to first use
- 2026
Staff role, not a client engagement
PageGuardian is a Python and Playwright validator anyone in the company can run, over thousands of pages, the moment content goes up. It replaced a paid crawler that billed per crawl and only ran on Mondays.
At a glance
54.6%
fewer 404 errors year over year
20 of 52
weeks with no 404 at all, up from 13
Zero
cost per run, against a crawler billed per crawl
The situation
Content went up and packages were installed on production and staging every day, but the only broad check on any of it was a crawl that ran on Monday. It came from a paid vendor tool that billed per crawl, so nobody ran an extra one just to check their own work. Between Mondays, finding a broken link or a page sitting in the wrong language meant opening pages by hand and hoping you opened the right one.
What I did
A browser, not a parser
PageGuardian is a Python service that drives Playwright across a site the way a browser would, so it reads the page after it renders rather than the HTML that came off the server. It runs against production and staging, and each pass checks the things that had been going unnoticed between crawls: broken links and 404s, SEO fields, missing or broken assets, text sitting in the wrong language, and staging URLs and other environment leaks that had made it into production.
Internal, and unlimited
Anyone in the company can point it at a set of pages, and a run costs nothing, so it happens right after a content upload or a package install instead of the following Monday.
What changed
The QA gain was the direct one. The weekly average of 404 errors fell from 5 to 2.73, and the number of weeks ending with no 404 at all went from 13 to 20, leaving 38% of the year clean. The two failure modes the vendor crawler never caught, wrong-language text and environment leakage, started getting caught at all.
The bigger gain was quieter. The weekly page-by-page hunt for problems stopped being part of anyone's week, and what users saw stopped depending on which day of the week they saw it.
Results
- Validation runs on demand instead of waiting for the Monday crawl
- Thousands of pages checked minutes after a content upload or a package install
- Unlimited internal runs replaced a vendor tool that billed per crawl
- Wrong-language text and environment leaks started getting caught at all
- Weekly 404 average down from 5 to 2.73: 54.6% fewer
- 20 weeks in the year with no 404 at all, up from 13
- Built in 3 weeks, validated by the product team in 1
Skills used
- Data Pipelines & Automation
- Web Development
- SEO
Systems: Python, Playwright
