One conversion, with the part that went wrong
Ritla has no client logos and no testimonials, because it has no clients yet. What it has is two conversions run against public open-source sites, and this is the one that produced a before and an after.
Before
After
+3 points, and the physical-CSS findings went from about two dozen declarations to 0. The headline moved three points because the other categories were already clean: what a conversion fixes is one category, and on a page that was otherwise fine that is what one clean category is worth.
The target, and why that one
varadbhogayata/varadbhogayata.github.io
An MIT-licensed personal portfolio people actually fork. Chosen for one property and verified before committing to it: its stylesheet is hand-written plain CSS carrying about two dozen physical direction declarations and no right-to-left handling at all. Static, with no build step — which matters more than it sounds, because with nothing between the codemod and the scan, nothing can explain away a difference in the numbers.
The one decline
A `padding` shorthand that could express a left-right asymmetry. It happened to be symmetric, and it was still right to decline: the difference is intent, and intent is not in the stylesheet. Every place the codemod stops is listed with a reason, and that list is the deliverable's most useful page rather than an apology — the kinds of refusal are on the service page.
It converted nothing the second time
The codemod was run a second time and converted nothing, writing no bytes — idempotency demonstrated on a third codebase we did not write.
What the score got wrong
The score went up and the page got visibly worse in one respect. The client's own CSS is logical now, so the content padding correctly moved to the right edge — and the sidebar stayed pinned left, because it is positioned by Materialize loaded from a CDN, a stylesheet in nobody's repository. A conversion of a site whose layout comes from a CDN framework is incomplete by construction, and at the time nothing in the tooling said so. It does now, and it is scoped out of an engagement by default.
This is on the page because it is the most useful thing on it. A number that rose while a layout got worse is exactly the failure a tool like this has to be honest about, and it is the reason a conversion is now qualified before it is quoted.
And the other run
The first target was a Bootstrap 5 admin template, and the codemod converted nothing at all across 84 files — correctly, because Bootstrap 5 is already written in logical properties. That is not a failure of the tool; it is the answer to a question a prospect should ask before buying, which is why the conversion plan is free and runs in a minute.