/* CRESTIAL — brand-features. GENERATED by _source/_docs/build_page_css.py
 * from _source/brand-features.src.html. Do not edit by hand. */

/* --- page index ----------------------------------------------------------
   Twelve anchors, four across. A features page is a reference page — the
   client's own brochure puts these across four spreads — so it gets a
   contents list rather than making people scroll to find one of them.
   Typographic, like the rest of the page. */
.index{display:grid;grid-template-columns:repeat(4,minmax(0,1fr));
  gap:0 clamp(24px,3.2vw,52px);margin-top:clamp(38px,4.5vw,60px)}
/* No rule over each entry (6 Oct 2026). The twelve were separated by hairlines,
   which made the index read as a table; the numbers already do that job. */
.index a{display:flex;align-items:baseline;gap:15px;padding:11px 2px;
  font-size:14px;line-height:1.5;
  transition:color .35s var(--ease)}
.index a:hover{color:var(--orange)}
.index .n{font-size:10.5px;font-weight:500;letter-spacing:.18em;color:var(--taupe)}
/* The index closes the intro and the next section opens with its own centred
   header. Two full section paddings back to back put a 250px hole between
   them, so the intro gives up most of its bottom one. */
#intro{padding-bottom:clamp(26px,3vw,46px)}

/* --- feature row ---------------------------------------------------------
   The six NAMED features. Two columns, 46/54, alternating sides, with the
   media column capped below its track width so the gutter is uneven — that
   is the brief's "abstract design, not following rigid alignment", done with
   column widths and padding rather than anything Elementor cannot express. */
.frows{display:grid;gap:clamp(56px,7vw,112px)}
.frow{display:grid;grid-template-columns:minmax(0,46fr) minmax(0,54fr);
  gap:clamp(30px,5vw,84px);align-items:center}
.frow__media{max-width:520px;width:100%}
.frow__media img{width:100%;border-radius:var(--rad);background:var(--stone)}
.frow--r .frow__media{order:2;margin-left:auto}
/* the stagger, set per row rather than by :nth-child so the wide Sundance
   band can sit in the middle of the sequence without re-keying everything */
.frow--drop .frow__media{margin-top:clamp(18px,3.4vw,60px)}
.frow__text{max-width:50ch}
/* ⚑ THE FEATURE'S NAME IS THE HEADING NOW, 16 NOV 2026, AND THE OLD HEADING IS THE
   STANDFIRST UNDER IT. The client reported the feature titles as "too small and hard
   to see" and they were right about the page, though not about which element: the
   <h2> has always been 23-32px. What was small was the feature's NAME — the only
   place on the page each feature is actually called what the index above calls it —
   and it was set at 10.5px in --stone-2.

   MEASURED BEFORE IT WAS CHANGED, because "looks faint" is not a finding:
   10.5px, rgb(196,188,183) on the rgb(238,235,228) paper, a contrast ratio of
   1.57:1. WCAG AA wants 4.5:1 for text that size and 3:1 for large text; it failed
   both by more than half. It is now 12px in --taupe, which is 4.62:1 — and it no
   longer carries the name, only the ordinal.

   ⚑ THE INDEX AND THE SECTION NOW AGREE. A link reading "01 Wasser Musik" jumped to
   a block headed "The pure sound of water music", so the heading never answered the
   link that reached it. That is the part a screen reader feels hardest: the only
   accessible name of the destination was a sentence that did not contain the words
   the user clicked.

   ⚑ NOTHING IS LOST AND NOTHING BECOMES UNEDITABLE. Both strings were already
   separate fields — the name is `name`, the sentence is the post title — so this is
   which element each one lands in, not new copy and not new boxes. Twelve <h2>s
   before, twelve after; the page keeps exactly one <h1>.

   ⚑ 36px AT 1440 AND NOT MORE. Measured on the built page: hero 50px, section
   heading 42px. A feature title above 42 would out-shout the section it sits in. The
   names are short — two or three words against a nine-word sentence — so 36px reads
   considerably larger than the 32px sentence it replaces, at a smaller number.

   ⚠ THREE OF THE SIX KICKERS NOW RESTATE THE TITLE. `CRESTIAL Wasser Musik` sits
   directly above `Wasser Musik`, and the same for Environ Pro and Easy Installation.
   The kicker is the client's own field with the client's own words in it, so it is
   still rendered and still theirs to clear; it is NOT suppressed by a rule here,
   because a line that vanishes when you type a particular thing into it is a bug
   from the editor's side. Raised with them instead. See DESIGN-FEEDBACK.md. */
.fnum{font-size:12px;font-weight:500;letter-spacing:.22em;color:var(--taupe);
  display:block;margin-bottom:12px}
/* The wide row is not a .frow, so every shared rule has to name it too —
   grouped here rather than duplicated, so a change to one changes both. */
.frow h2,.fwide h2{font-size:clamp(26px,2.6vw,36px);font-weight:200;
  letter-spacing:-.015em;line-height:1.12;margin:0 0 14px}
/* ⚠ SCOPED TO THE ROW, AND `.fsub` ALONE WOULD NOT HAVE WORKED. `.frow p` below is
   (0,1,1) and beats a bare class at (0,1,0) wherever they disagree, whatever the
   order — so a <p class="fsub"> would have come out at 15px while this rule sat
   right above it looking applied. Naming the row makes it (0,2,0), which wins.

   ⚠ AND THAT IS WHY THE KICKER HAS BEEN RENDERING AT 15px. `.fkick` declares 11.5px
   and has been losing the same fight since the page was written: measured on the
   built page before this change, the kicker computed to 15px with its tracking and
   its uppercase intact — so it looked designed rather than broken. A 15px tracked
   capital line sitting above the heading is most of why this block read as muddled:
   the NAME was 10.5px, the kicker 15px and the heading 32px, which is the hierarchy
   upside down. Scoped below with the same fix. */
.frow .fsub,.fwide .fsub{font-size:clamp(16px,1.6vw,20px);font-weight:300;
  line-height:1.45;color:var(--taupe);margin:0 0 24px}
.frow p,.fwide p{font-size:15px;line-height:1.85;color:var(--taupe);margin:0 0 1em}
.frow p.lead,.fwide p.lead{color:var(--ink);font-size:16px}
/* No icons on this page. The twelve 2018 marks were removed 24 Sep 2026 — they
   were never a coherent set (three hand-drawn doodles, one a photograph of a
   feather) and holding them at 34px only made that quieter, not untrue. The
   blocks are typographic: number, kicker, headline. If the pictogram set that
   is on request arrives, it comes back as one rule here and one img per block.
   See DESIGN-FEEDBACK.md. */
.frow .fkick,.fwide .fkick{font-size:11.5px;font-weight:500;letter-spacing:.2em;
  text-transform:uppercase;color:var(--taupe);margin:0 0 18px;line-height:1.4}
/* ⚑ THE KICKER SITS UNDER THE TITLE NOW, 16 Nov 2026 — Keith: "first line should be
   the feature title, followed by the kicker below it. we should standardise it for
   every section row." Sequence on both paths: .fnum / <h2> / .fkick / .fsub.

   THE TWO MARGINS ARE NOT THE SAME AND THAT IS THE POINT. A kicker ABOVE a title
   needs equal air either side — it is a label introducing what follows. Below it, it
   belongs TO the title and has to sit closer to it than to the sentence underneath,
   or it reads as a third separate line floating between two things. So the heading
   gives up 6 of its 14px when a kicker follows it (8px above the kicker) and the
   kicker takes 18px below — roughly 1:2, which is what makes the pair read as one
   unit rather than as three.

   ⚠ AND THE HEADING RULE CANNOT SIMPLY CHANGE, because seven of the twelve features
   have no kicker and go straight from <h2> to .fsub, where 14px is the spacing this
   page has always used. :has() asks the question per row instead of splitting the
   rule — the same device, and the same reason, as the empty-contact-block rule on
   the homepage, Contact Us and Be Our Partners. */
.frow h2:has(+ .fkick),.fwide h2:has(+ .fkick){margin-bottom:8px}
/* The rule over this line came off 6 Oct 2026. Its 24px margin plus 18px of
   padding existed to clear a hairline; with the hairline gone that is 42px of
   nothing, so it collapses to one margin and the small caps do the separating. */
.fmeta{margin-top:26px;padding-top:0;border-top:0;
  font-size:11px;font-weight:500;letter-spacing:.16em;text-transform:uppercase;
  color:var(--taupe)}

/* --- the two diagrams inside the named rows (6 Oct 2026) ------------------
   Velvet to the Touch and Environ Pro are the two claims on this page that are
   about BEHAVIOUR rather than about an object — a comfort zone across a handle's
   travel, and a flow rate holding steady as pressure rises. The brochure draws
   both, and a photograph of a mixer cannot.

   They are redrawn here as inline SVG rather than lifted as images, for three
   reasons and one of them is not laziness: they stay sharp at any width, they
   cost about 2KB against 40 on a page that inlines everything, and the
   brochure's primary red / green / blue reads as a spreadsheet on a warm cream
   page whose brief is "must look expensive". Every curve, label and value is
   the client's — only the palette is ours.

   The two rows that carry one are WIDER ON THE TEXT SIDE, and that is the
   whole trick. At the standard 46/54 the diagram had 290px, which is where a
   470-unit viewBox puts five labels on top of each other — the first build of
   this page had "environmental protection" printed through "without any
   restriction". Spanning the full row instead opened a ~240px dead band under
   the plate, because the row centres a short text column against a tall one.
   Re-proportioning the row fixes both: the diagram gets 640px and the row
   keeps its rhythm. A diagram that has to be read does not go in a 50ch column. */
/* Three rows carry evidence in the text column rather than prose alone — two
   diagrams and the five spray discs — and all three need the same extra width.
   WATCH THE REVERSED ROW: .frow--r does not swap the TRACKS, it swaps the
   order of the children inside them, so on a reversed row the first track is
   the text. Setting 36/64 flat gave Velvet a 310px text column and a 790px
   photograph, which is the opposite of the intent. */
.frow--evidence{grid-template-columns:minmax(0,36fr) minmax(0,64fr)}
.frow--evidence.frow--r{grid-template-columns:minmax(0,64fr) minmax(0,36fr)}
.frow--evidence .frow__text{max-width:720px}
/* ⚑ THE 720px TRACK IS FOR THE EVIDENCE, NOT FOR THE PROSE — 16 Nov 2026.
   `max-width` above is on the whole COLUMN, so it widened the heading, the
   standfirst and the body copy as a side effect of widening the diagram. The
   client reported the feature titles as inconsistent, "some small, some big",
   and this is where that came from.

   MEASURED AT 1440 BEFORE THE CAP, as the proportion of its own column each
   title actually fills — every one of the twelve is the same 36px:

       04  Environ Pro                 185px in 720px   26%
       05  Vita Sprays                 184px in 720px   26%
       09  Anti-Vandal                 202px in 443px   46%
       11  Adjustable Aerator          319px in 443px   72%
       02  CRESTIAL Surface Finishes   422px in 510px   83%

   A 3.2x spread in how much of its line a title occupies, at one type size.
   Nothing here was ever a font-size: a short name alone in a 720px measure
   reads small, and the eye is comparing it against a neighbour that fills
   its own. ⚑ THE SIZE WAS THE FIRST THING I MEASURED AND IT WAS IDENTICAL
   ON ALL TWELVE — had I trusted that and stopped, the report would have been
   "cannot reproduce" on a fault that is plainly there on the screen.

   ⚠ AND ROW 10 IS AN EVIDENCE ROW CARRYING NO EVIDENCE. Adjustable Shower
   Rail takes .frow--evidence for its proportions but has no diagram and no
   disc row, so its body copy had been running the full 720px — about 95
   characters a line against the 50ch (443px) every other row uses. That is a
   readability fault in its own right and it has been live on this page since
   the layout was written; the cap below fixes it in the same line.

   So: the TRACK stays 720px, because the diagram needs it. Everything in the
   column that is text is capped to the measure the other ten rows use. 27.5rem
   is 440px, which is the 50ch of `.frow__text` resolved at its own 15px. */
.frow--evidence .frow__text > *{max-width:27.5rem}
.frow--evidence .fdiag,.frow--evidence .discs{max-width:none}
.fdiag{margin:clamp(30px,3.4vw,44px) 0 0}
.fdiag svg{width:100%;height:auto;display:block;overflow:visible}
/* the client's own chart file, inlined — it carries its own padding */
.cfr-src{width:100%;height:auto;display:block}
.fdiag .dlab{font-size:9.6px;font-weight:500;letter-spacing:.14em;
  text-transform:uppercase;fill:var(--taupe)}
.fdiag .dax{font-size:9px;letter-spacing:.12em;text-transform:uppercase;
  fill:var(--taupe)}
.fdiag figcaption{margin-top:16px;font-size:11px;line-height:1.6;
  letter-spacing:.04em;color:var(--taupe)}

/* --- the wide row (Sundance) ---------------------------------------------
   One row in six breaks the ladder, and it is the row whose content demands
   it: the five finishes are laid out diagonally across a single frame, so a
   portrait window would crop two of them off. Content-driven asymmetry.
   Below the plate: heading block against body copy, then the swatch strip
   runs the full width — five chips on one line is the point of them, and
   squeezed into half the row they wrapped two/two/one. */
.fwide__img{width:100%;border-radius:var(--rad);background:var(--stone)}
.fwide__grid{display:grid;grid-template-columns:minmax(0,42fr) minmax(0,58fr);
  gap:clamp(28px,4.5vw,76px);margin-top:clamp(30px,3.6vw,52px);align-items:start}
.fwide__body{max-width:56ch}
/* A WIDE ROW WHOSE BODY IS A SINGLE PARAGRAPH IS NOT A TWO-COLUMN ROW.
   The shape was drawn around 02, whose body runs eight lines and so outweighs
   the heading block beside it; the heading nestles into its top-left and the
   swatch strip closes the article underneath. Feature 08 has one three-line
   lead and nothing after it, and in the same grid that reads as an orphan:
   measured at 1440, the text sat 34px ABOVE the heading it belongs to (both
   columns are align-items:start, so the body lines up with the little 08
   marker, not with the title), finished 36px SHORT of the heading column it
   was supposed to balance, and stopped 208px short of its own grid track --
   because the track is 705px wide and .fwide__body can never exceed the 56ch
   measure of 497px. Nothing follows the grid on 08, so the article simply
   ended in that hole.
   ⚑ MOST OF THIS PREDATES 16 Nov and the 208px dead track predates the page
   being built: the shortfall was 23px before the title rework made the
   heading column taller, and 36px after. Measured both ways by simulating the
   old arrangement in the DOM rather than reasoning about it.
   Collapsing to one column puts the paragraph under the standfirst at the
   same left edge, at the same 56ch measure, with NO invented spacing: .fsub
   already carries 24px below it, which is exactly the standfirst-to-paragraph
   gap all ten .frow rows use. gap:0 lands on 24px by itself.
   Keyed on the CONTENT, so it is self-correcting -- the day anyone adds a
   second paragraph to 08 it becomes a two-column row again with no edit here,
   and 02 is never touched, which matters because 02 is signed off. */
.fwide__grid:has(.fwide__body > p:only-child){grid-template-columns:1fr;gap:0}
.fwide .fmeta{margin-top:22px}
/* The five finishes, as the brochure's own swatch discs rather than the flat
   CSS gradients that stood in for them until 6 Oct 2026. They are 74px, not
   26px, because at 26px a photograph of brushed nickel and a photograph of
   chrome are the same grey dot — the grain and the PVD sheen ARE the content,
   and the client's brochure gives them a quarter of a page for that reason.
   Round in CSS, square on disk: see prep_images_d.py. */
/* Five columns inside a CAP, which is the only version of this that holds in
   both places it is used. Plain columns stretched to the container and gave
   86px discs under the Sundance band and 52px under Vita. Flex with a sized
   cell fixed that and broke the other way: five 138px cells plus gaps overran
   Vita's column and wrapped the last pattern onto a line of its own, so Air
   Dynamics sat alone under the other four. A cap sizes them once, for both. */
.discs{display:grid;grid-template-columns:repeat(5,minmax(0,1fr));
  gap:clamp(16px,2vw,26px);max-width:680px;margin-top:clamp(30px,3.4vw,44px)}
.disc{text-align:center}
.disc img{width:clamp(58px,6vw,86px);aspect-ratio:1;border-radius:50%;
  display:block;margin:0 auto 14px;background:var(--stone);
  box-shadow:inset 0 0 0 1px rgba(20,18,15,.16)}
.disc b{display:block;font-size:11.5px;font-weight:400;line-height:1.5;
  letter-spacing:.04em;color:var(--ink)}
.disc .cd{display:block;margin-top:4px;font-size:9.5px;font-weight:500;
  letter-spacing:.2em;color:var(--taupe)}
/* the spray patterns are photographed on black, so they need no inner hairline
   and carry their name in the brochure's own uppercase */
.discs--spray .disc img{box-shadow:none}
.discs--spray .disc b{font-size:10.5px;font-weight:500;letter-spacing:.17em;
  text-transform:uppercase;color:var(--taupe)}

/* --- the picture panel ----------------------------------------------------
   The artwork features 07-12 carry, which the first six do not have.

   WHERE IT CAME FROM, 6 Oct 2026: these six used to be type-only cards, because
   the page had no imagery for them and would not substitute a mood shot of a
   bathroom for a mechanism. The brochure pages the client supplied carry a
   cutaway, a four-step sequence, an aerator key, a plug section, an annotated
   shower and three flow angles — drawings that say exactly what the sentence
   says. So the six became plates on a twelve-column grid, each taking the span
   its artwork needed: 7/5, then a full 12, then 3/5/4.

   ⚡ THE PLATE GRID IS GONE, 5 Nov 2026, and so is the band it stood in. The
   client asked for one list of twelve they can add to and reorder, so these six
   became rows like the other six and their artwork moved into a row's own
   picture column. What was `.plates`, `.plate`, `.plate--7` and `.detail` is now
   `.part` inside `.frow__media`, and the span each plate took is a proportion it
   no longer gets to choose — see the min-height note below, which is the cost.

   Everything from here down describes the panel as it is. Where a comment
   mentions a plate it is history, marked as such.
--------------------------------------------------------------------------- */
/* --pp IS GONE, 5 Nov 2026. It existed so .pfull could cancel the PLATE's own
   padding exactly — written as a second clamp the browser resolves independently,
   the image stood a few pixels proud of the rounded corner. There is no plate any
   more: the panel sits directly in the row's picture column, which has no padding,
   so there is nothing to cancel and .pfull simply fills the column. Left in place
   for one commit it pulled the EasyClean photograph 36px past the column on each
   side and 20px past the page's own right gutter — measured, not spotted. */
/* The artwork sits on its own ground, so a cut-out render and a line drawing
   read as plates from one sheet rather than as loose images. That is the whole
   job of this rule and it is why the background colour below matters more than
   it looks.

   flex-wrap, and the SVG set not to shrink: a flex item with an intrinsic width
   still gives it up under pressure, and at 800 the Cu/Zn ring was squeezed to
   about 40px beside the cutaway — a chart too small to read its own labels.
   Below the page's own breakpoint the pieces stack instead. */
/* ⚑ min-height IS THE PRICE OF THE MERGE, and it is worth naming. These six were
   drawn as plates on a twelve-column grid, each taking the span its artwork needed —
   7/5, a full 12, then 3/5/4. As rows they take a row's proportions instead, and
   measured at 1440 the six of them came out 282, 531, 312, 392, 197 and 443 tall
   against 693 to 1300 for the six above. The aerator row was a sixth of the Sundance
   one. A floor under the panel is not the same as redrawing them, and it stops the
   lower half of the page reading as an afterthought. */
/* ⚡ --card, NOT --paper, SINCE 5 Nov 2026 — AND THE PANEL WAS INVISIBLE UNTIL IT WAS.
   The well was #EEEBE4 because it sat inside .detail, a band of #F8F6F1: lighter
   band, darker inset. Those six rows moved into .frows, which sits on the page's
   own #EEEBE4, so the well and the page behind it became the SAME COLOUR. Measured,
   not noticed: computed background rgb(238,235,228) on both. What you saw was each
   picture's own baked-in ground and nothing holding them together — the exact
   "loose images" this rule exists to prevent. One step lighter restores the
   relationship on the ground the rows actually stand on. */
.part{--pgap:clamp(14px,2vw,28px);
  border-radius:14px;background:var(--card);
  padding:clamp(16px,2vw,26px);display:flex;flex-wrap:wrap;align-items:center;
  justify-content:center;gap:var(--pgap);
  min-height:clamp(200px,20vw,280px)}
.part img{display:block;max-width:100%;height:auto}
/* ⚑ ONE RULE WHERE THERE WERE THREE INLINE STYLES. The cutaway, the aerator key and
   the plug section each carried their own max-height typed onto the element — 230,
   190 and 150 — which is fine in a mockup and impossible in the build: the pictures
   come out of the media library now and nobody is going to type a style attribute
   onto an upload. The direct-child selector keeps it off the stepped sequence and
   the square row, which size themselves. */
.part>img{max-height:clamp(200px,20vw,280px);width:auto}
/* ⚑ AND A WIDTH CAP WHEN THERE IS MORE THAN ONE, or a pair stops being a pair. The
   panel wraps before it shrinks — that is what flex-wrap does — so with only a height
   cap the alloy ring and the cutaway came out 280 and 242 wide in a 468 column and
   stacked into a tower 640 tall. Measured, not guessed: their rects had different
   tops. Half the panel each, less half the gap, and both designed pairs sit side by
   side again at 220 wide.
   `:only-of-type` is doing real work here: a panel holding ONE picture is not capped
   to half a column, which is the common case for anything the client adds. */
.part>img:not(:only-of-type){max-width:calc(50% - var(--pgap) / 2)}
.part svg{flex:none}
/* The narrow media column - the shower rail standing on end. It was
   .plate--tall, which was a column span; on a row it is the evidence variant's
   36% track doing the same job. */
.frow--evidence .part img{max-height:340px;width:auto}
/* 10 is the client's own 4-second render (19 Oct), not a drawing like its
   neighbours. The source is 2000x2000 and the shower occupies a strip
   between 31% and 70% across it — measured over the whole loop, not off one
   frame, because the handshower swings out and the headshower rises. So
   `cover` in a tall box crops to the subject with room either side and
   nothing is re-encoded; the file ships exactly as supplied, renamed from
   .m4v (it is already H.264 in an MP4 container, and silent — there is no
   audio track in it at all).
   The background is the render's own #2b2b2b, so there is nothing to see
   in the box before the first frame decodes and no poster file to keep in
   step with it. */
.vloop{display:block;height:340px;width:auto;max-width:100%;
  aspect-ratio:92/190;object-fit:cover;border-radius:10px;background:#2b2b2b}
/* a caption the client wrote on the artwork itself, kept with it */
.pcap{margin-top:12px;font-size:11px;line-height:1.55;letter-spacing:.04em;
  color:var(--taupe);text-align:center}
/* the four diverter frames, captioned in our type rather than the brochure's */
.steps4{display:grid;grid-template-columns:repeat(4,minmax(0,1fr));
  gap:clamp(14px,2vw,30px);width:100%}
.step4{text-align:center}
.step4 img{display:block;margin:0 auto;height:clamp(150px,15vw,208px);width:auto}
.step4 b{display:block;margin-top:16px;font-size:10px;font-weight:500;
  letter-spacing:.22em;color:var(--taupe)}
.step4 span{display:block;margin-top:7px;font-size:12.5px;line-height:1.6;
  color:var(--taupe)}
/* The three flow angles are black-backed photographs — they keep their ground.
   The three source frames are 693x401, 582x559 and 606x581, so laid out raw
   they stepped down in height and read as three unrelated pictures. One ratio
   and object-fit makes them one sequence of the same mixer. */
.angles{display:grid;grid-template-columns:repeat(3,minmax(0,1fr));gap:8px;width:100%}
/* Square, not 3:4: two of the three frames are near-square already and the
   third is 693x401, so a portrait crop threw away the basin in all three and
   left black with a sliver of chrome in it. */
.angles img{width:100%;aspect-ratio:1;object-fit:cover;border-radius:8px}
/* A photograph that IS the demonstration fills the picture column: no tint, no
   padding, no rounded corner and nothing inset. It keeps the panel's min-height so
   the row holds its place in the rhythm. */
.pfull{background:none;padding:0;border-radius:0;overflow:hidden;margin:0}
/* ⚑ max-height:none IS NOT TIDYING UP. `width:100%` below sets the width outright, so
   the height cap above had nothing left to scale — it simply clipped the box, and the
   EasyClean photograph rendered 520x280 where its own 912x689 wants 520x393. A 29%
   vertical squash on a close-up of a thumb on a nozzle. It reads as a slightly odd
   crop rather than as a bug, which is exactly why it needs a rule and not an eye. */
.pfull img{width:100%;max-width:100%;max-height:none;border-radius:0;display:block}

/* --- standards -----------------------------------------------------------
   White band, the same quarantine the homepage's Assurance section uses, so
   the two certification blocks read as one voice across the site. */
.stds{background:#fff}
.stds__list{max-width:1080px;margin:0 auto;display:grid;
  grid-template-columns:repeat(2,1fr);gap:0 clamp(30px,4.4vw,72px)}
/* Ruled rows until 6 Oct 2026. Without them the code column carries the list:
   it is the only thing set in small caps, and it starts every row. */
.std{display:grid;grid-template-columns:minmax(0,150px) minmax(0,1fr);gap:20px;
  padding:14px 0}
.std .c{font-size:12px;font-weight:500;letter-spacing:.13em;text-transform:uppercase;
  line-height:1.6}
.std .d{font-size:13.5px;line-height:1.7;color:var(--taupe)}
.stds__foot{text-align:center;margin-top:clamp(38px,4.4vw,58px)}

/* --- closing call to action ---------------------------------------------- */
.cta{background:var(--ink-2);color:var(--cream);text-align:center}
.cta .sub{color:rgba(252,251,248,.72)}
.cta .head{margin-bottom:clamp(30px,3.6vw,44px)}

/* ===== responsive (this page only) ===== */
@media(max-width:1080px){
  .index{grid-template-columns:repeat(2,1fr)}
}
@media(max-width:900px){
  .steps4{grid-template-columns:repeat(2,minmax(0,1fr));
    gap:clamp(20px,3vw,34px) clamp(14px,2vw,30px)}
}
@media(max-width:860px){
  /* One column, and the media goes FIRST on every row regardless of side —
  an alternating stack reads as a mistake once the columns are gone. */
  /* .frow--evidence.frow--r is two classes and this is one, so the reversed
  evidence rows went on holding their 64/36 split well below the breakpoint —
  Vita kept a 230px text column at 560 and stacked its five spray discs three
  and two inside it, with CHAMPAGNE running into AIR DYNAMICS. Re-stated at
  equal specificity. */
  .frow,.frow--evidence,.frow--evidence.frow--r{
    grid-template-columns:1fr;gap:clamp(24px,4vw,34px)}
  .frow--evidence .frow__text{max-width:none}
  /* ⚠ AND THE PER-CHILD CAP COMES OFF WITH IT. The desktop rule caps every
  text child of an evidence row to 27.5rem so a short title is not marooned
  in the 720px diagram track. Once the row is one column that track is gone,
  and leaving the cap on made the four evidence rows the ONLY ones with a
  440px measure while the other eight ran the full 723 — which is the exact
  inconsistency the cap was added to remove, moved to tablet width. Measured
  at 820 before this line existed. */
  .frow--evidence .frow__text > *{max-width:none}
  .frow--r .frow__media{order:0;margin-left:0}
  .frow--drop .frow__media{margin-top:0}
  /* Wider than the desktop cap, not narrower: once the text runs the full
  column, a 440px plate left a 260px hole beside it that read as an
  orphan rather than as asymmetry. min() lets it go near-full-bleed on
  a phone without a third breakpoint. */
  .frow__media{max-width:min(100%,540px)}
  .frow__text{max-width:none}
  .fwide__grid{grid-template-columns:1fr;gap:clamp(24px,4vw,32px)}
  /* ⚑ The one-column rule above is :has(), so it OUTRANKS the line before this
  one on specificity whatever the source order -- its gap:0 would otherwise
  follow 08 down to the phone and give it 24px where the other wide row has
  32px. Same trap as the evidence-row measure cap on 16 Nov: a desktop rule
  added above this block is invisible to it until it is named here. */
  .fwide__grid:has(.fwide__body > p:only-child){gap:clamp(24px,4vw,32px)}
  .stds__list{grid-template-columns:1fr}
}
@media(max-width:620px){
  .index{grid-template-columns:1fr}
  /* Five discs across a phone gives each one 48px and a two-line name under it.
  Three and two reads as a set; five reads as a crush. */
  .discs{grid-template-columns:repeat(3,minmax(0,1fr));row-gap:clamp(20px,5vw,30px)}
  .std{grid-template-columns:1fr;gap:6px}
}
