data center prep
- navy, #071426
- ink, #0B1117
- cobalt, #1B44D9
- cobalt hover, #173AB8
- cobalt active, #133098
- cobalt on dark, #3E63F0
- wash, #E8ECFB
- mist blue, #D8E6FF
- slate, #5F6B7A
- steel, #7A8594
- soft white, #F7F9FC
this is the studio’s own product business, not a client engagement. it is in the register because it is the only project where the same hands did the identity, the site, the products and the fulfillment, and because a studio that will not show its own work under the same rules it applies to a client’s is not applying the rules.
datacenterprep sells interview preparation to engineers going for data center jobs: a question bank, a fundamentals course and career material. the buyers are technical, they are spending their own money, and they check things. that audience decides the whole brief, because a claim that cannot be checked is worse than no claim in front of a reader who will check it.
the identity
- primary lockup
the icon
- navy#071426
- ink#0B1117
- cobalt#1B44D9
- cobalt hover#173AB8
- cobalt active#133098
- cobalt on dark#3E63F0
- wash#E8ECFB
- mist blue#D8E6FF
- slate#5F6B7A
- steel#7A8594
- soft white#F7F9FC
11 values · eleven values from the written brand book, including the three cobalt interaction states, published together because the book forbids one pairing of them by name and the ratio is printed beside it
what was wrong
- four webfonts shipped as the foundry sent them, with every glyph for every language, on a site that sets latin and a handful of symbols. inter alone was 259,524 bytes of which about a quarter was reachable.
- the brand had a documented palette and no published contrast figures, so the one pairing that fails was written down nowhere and could be reached by anyone following the guidelines.
what was decided
subset every face to the characters the site can set, and keep the untouched originals in the repository beside them.
why
a subset you cannot reverse is a subset nobody will re-run. _fonts-original/ is committed so the operation can be redone from source when the character set changes, and the saving can be re-derived by anyone with the repository.
rejected: a font cdn, which would have moved the identity onto a third party domain and made the one thing that carries the brand in every line of copy the one thing not under our control.
publish the failing pair in the brand book as a hard rule, with its measured ratio.
why
cobalt on navy is the pairing a designer reaches for first, and it measures 2.53:1. a rule that names its own failure and prints the number is the only kind anyone follows.
the numbers
2.53:1
DCP Cobalt #1B44D9 set on DCP Navy #071426, the pairing the brand book forbids by name
m LOGO DCP/brand-colors.md, recomputed here from the hex values with the WCAG 2.1 relative luminance formula
does not prove: one pairing. it says nothing about the rest of the palette, and it is a contrast figure, not a judgment about whether the colors look right together.
68 percent
the reduction across all four faces, 400,956 bytes to 127,776, when the two conditional faces are included
m commit 6b0a50c
does not prove: not what a visitor pays on a first view. the headline figure above is the one that reaches a browser.
one more figure, with what it measures, the file it was read from, and what it does not prove
m
does not prove:
the useful thing about running the studio’s own product business is that every shortcut costs the studio directly. a claim that cannot be checked does not get quietly shipped and forgotten, it gets checked by a stranger who has paid for it.
what they own now
- the repository, the deploy and the fulfillment, in one place
- the mark suite in every state, as flat vector
- the palette as a written book with its failing pairing named
what this does not prove
this section is here on every file. it is the reason the rest of the page is worth reading.
blocked out: no before and after page weight for the whole site.
why we are not claiming it
the saving was measured on the font files, which is where it happened. no full page transfer was recorded on either side of the change, so the site-wide figure does not exist and is not estimated.
this is the studio’s own product, so nothing here was negotiated with a client, and a case file where the designer and the client are the same is easier than one where they are not.