docs: small multiples can split by two dimensions - #11523

• Read PR diff against origin/master
• Read the full page for internal consistency
• Check sibling pages for stale cross-references
• Post findings

Medium

1. small-multiples.mdx:18 — "single row of panels" contradicts the page (inline)
Line 16 says one split dimension gives one panel per value, and the Grid row says a four-value dimension opens as a 2 × 2 grid — so a single dimension wraps across both axes. Clearing the second picker therefore returns you to that behaviour, not to a one-row layout.

2. small-multiples.mdx:28 — Grid cell fights itself, and "the split dimension" is now ambiguous (inline)
The lead-in ("the most tiles each axis may draw") states the two-dimension reading, then the following sentence carves out a different single-dimension reading. A neutral gloss ("Columns × rows, up to 5 × 5") lets the distinction land. Separately, with two pickers filled it's unclear which dimension drives the 2 × 2 preselection.

Low

3. small-multiples.mdx:24 and :30 — singular "dimension" left un-updated.
"These options appear once a Split by dimension is chosen" and "Orders the panels by the dimension's own values" both predate the second axis. The Sort panels by row in particular raises a real question the page doesn't answer: with two split dimensions, does one sort setting apply to both axes? Line 49 ("Each axis drops its own overflow in that order") implies yes — a clause in the table row would save the reader the inference.

4. Legacy duplicate not touched — correct, but worth confirming intent.
docs/explore-analyze/charts/configuration/small-multiples.mdx (the deprecated Nextra site) still carries the old text. Per CLAUDE.md that site is deprecated and shouldn't be edited, so leaving it is right; noting it only so the divergence isn't a surprise later.

Verified fine

• 5 × 5 / twenty-five is consistent at lines 28, 43, and 45 — no leftover 5 × 4 or "twenty panels" anywhere in docs-mintlify.
• The 5 × 40 → 25-panels example is arithmetically right under per-axis caps.
• The Limitations bullet rewrite is consistent with the new body text, and the remaining bullets are untouched and still accurate.
• No docs.json change needed (no new page), and no sibling page's description of small multiples is invalidated — configuration/index.mdx:13 and series-mapping.mdx:18 both describe the split generically.
• N/A for this PR: security, performance, tests.

One process note from the PR body: since this also carries the 5 × 5 correction from #11498, whichever lands second will conflict on lines 28 and 43.

• Re-read PR diff against origin/master
• Re-read the full page for internal consistency
• Check prior review threads (both resolved/outdated — suggestions applied)
• Post findings

Low

1. small-multiples.mdx:31 — Sort order row doesn't inherit the two-dimension caveat. (inline)
Line 30 now says the row order is "always … ascending", so with two split dimensions Sort order governs the columns only. The adjacent row still reads as a global control.

2. small-multiples.mdx:47 — "more values than the grid has tiles" is still the one-dimension framing.
With two split dimensions the overflow test is per axis (column dimension vs. the columns number), not value-count vs. total tiles — which is exactly the point lines 45 and 49 make. A short "each axis against its own number" clause would keep the three paragraphs telling one story.

3. Frontmatter description (line 3) and the intro (line 6) still say "one per value of a dimension".
Accurate for the single-dimension case and fine as a summary; flagging only because the description is the SEO/nav preview and the grid-of-two is now the headline capability. Optional.

Verified fine

• Lines 18, 28, 30, 45, 49, and 61 all agree: first dimension → columns, second → rows, each number caps its own axis, columns follow Sort panels by, rows ascending by value.
• 5 × 5 / twenty-five consistent at 28, 43, 45. No stale 5 × 4 or "twenty panels" left in docs-mintlify.
• The 5 × 40 → 25-panel example is right under per-axis caps.
• Limitations bullet 1 rewrite matches the body; the other bullets are untouched and still accurate.
• No docs.json change needed (no new page). Sibling pages (configuration/index.mdx:13, series-mapping.mdx:18) describe the split generically and aren't invalidated.
• Legacy docs/explore-analyze/charts/configuration/small-multiples.mdx correctly left alone per CLAUDE.md.
• N/A for this PR: security, performance, tests.

Unverifiable from the repo

• The and by picker label and the "a dimension already used on one axis is not offered on the other" behaviour are UI claims with no source in this repo — worth a spot-check against the shipped UI, since the PR body notes it depends on that change landing first.

Still true from the earlier review: this carries the 5 × 5 correction from #11498, so whichever lands second will conflict on lines 28 and 43.

• Re-read PR diff against origin/master
• Re-read the full page for internal consistency
• Check prior review threads (all four resolved — suggestions applied)
• Check sibling and legacy pages for stale cross-references
• Post findings

Low

1. small-multiples.mdx:16 — the page never says what clearing the first picker does when both are filled. (inline)
Line 16 ("Clearing the picker returns the chart to a single plot") predates the second axis, and line 18 only covers clearing the second one. With both filled, does clearing Split by drop to a single plot, or promote the row dimension to the column axis? Your own note on the earlier thread suggests the code has a promotion path (setFacetDefFor(spec, axis, …) promoting the row axis) — worth a clause either way, since it's the one round trip the page leaves open.

2. Frontmatter description (line 3) and the intro (line 6) still say "one per value of a dimension".
Carried over from the last pass, still optional. Accurate as a summary of the single-dimension case; flagged only because the description is the SEO/nav preview and the two-dimension grid is now the headline capability.

Verified fine

• Lines 18, 28, 30, 31, 45, 47, 49, and 61 all agree: first dimension → columns, second → rows, each number caps its own axis, Sort panels by / Sort order govern the columns, rows ascending by value.
• 5 × 5 / twenty-five consistent at 28, 43, 45. No stale 5 × 4 or "twenty panels" anywhere in docs-mintlify.
• The 5 × 40 → 25-panel example is arithmetically right under per-axis caps.
• Line 47's overflow rewrite reads correctly against both 43 (product bound, one dimension) and 45 (per-axis cap, two).
• Limitations bullet 1 ("At most two split dimensions") matches the body; the other three bullets are untouched and still accurate.
• No docs.json change needed (no new page). Siblings configuration/index.mdx:13 and series-mapping.mdx:18 describe the split generically and aren't invalidated.
• Legacy docs/explore-analyze/charts/configuration/small-multiples.mdx correctly left alone per CLAUDE.md.
• N/A for this PR: security, performance, tests.

Unverifiable from the repo

• The and by picker label and "a dimension already used on one axis is not offered on the other" are UI claims with no source in this repo — worth a spot-check against the shipped UI, since the PR body notes this depends on that change landing first.

Still true from earlier passes: this carries the 5 × 5 correction from #11498, so whichever lands second will conflict on lines 28 and 43.

• Re-read PR diff against origin/master
• Re-read the full page for internal consistency
• Check prior review threads (all five resolved — suggestions applied)
• Check sibling and legacy pages for stale cross-references
• Post findings

Low

1. small-multiples.mdx:16 — clearing the first picker is still unstated. (inline)
The new sentence answers "how do I get back to a single plot" via the toggle, but it also replaced the old "Clearing the picker…" clause — so the page now describes clearing only for the second picker (line 18). A reader who clears Split by with both filled still has to guess between promotion and collapse. If Split by simply isn't clearable in the UI, a half-clause saying so would keep lines 16 and 18 from reading as an inconsistency.

Not re-raising the frontmatter description point from the last pass — the author's reasoning (single-panel-per-value is what a search arrival is looking for; the grid is the page's second sentence) is sound and I agree it should stay.

Verified fine

• Lines 16, 18, 28, 30, 31, 45, 47, 49, and 61 tell one story: first dimension → columns, second → rows, each number caps its own axis, Sort panels by / Sort order govern the columns, rows ascending by value, toggle clears both.
• "taking both split dimensions with it" on line 16 is consistent with the Limitations bullet and with line 18's single-dimension fallback.
• 5 × 5 / twenty-five consistent at 28, 43, 45. No stale 5 × 4, "twenty panels", or "One split dimension" anywhere in docs-mintlify.
• The 5 × 40 → 25-panel example is arithmetically right under per-axis caps.
• Line 47's per-axis overflow framing reads correctly against both 43 (product bound, one dimension) and 45 (per-axis cap, two).
• Siblings configuration/index.mdx:13 and series-mapping.mdx:18 describe the split generically and aren't invalidated. No docs.json change needed (no new page).
• Legacy docs/explore-analyze/charts/configuration/small-multiples.mdx correctly left alone per CLAUDE.md.
• N/A for this PR: security, performance, tests.

Unverifiable from the repo

• The and by picker label, the toggle beside Split by, and "a dimension already used on one axis is not offered on the other" are UI claims with no source in this repo — worth a spot-check against the shipped UI, since the PR body notes this depends on that change landing first.

Still true from earlier passes: this carries the 5 × 5 correction from #11498, so whichever lands second will conflict on lines 28 and 43.

• Re-read PR diff against origin/master
• Re-read the full page for internal consistency
• Check prior review threads (all six resolved — suggestions applied)
• Check sibling and legacy pages for stale cross-references
• Post findings

Verified fine

• The clear/toggle asymmetry now reads as design, not oversight. Lines 16 and 18 together say: the toggle switches the whole split off and takes both dimensions with it; only the second picker is clearable. That matches the behaviour you described from the code (Split by isn't isClearable, its handler ignores a null selection) and leaves no round trip unstated.
• Lines 16, 18, 28, 30, 31, 45, 47, 49, and 61 tell one story: first dimension → columns, second → rows, each number caps its own axis, Sort panels by / Sort order govern the columns, rows ascending by their own values.
• 5 × 5 / twenty-five consistent at 28, 43, 45. No stale 5 × 4, "twenty panels", or "One split dimension" anywhere in docs-mintlify.
• The 5 × 40 → 25-panel example is arithmetically right under per-axis caps.
• Line 47's per-axis overflow framing reads correctly against both 43 (product bound, one dimension) and 45 (per-axis cap, two).
• Limitations bullet 1 ("At most two split dimensions") matches the body; the other three bullets are untouched and still accurate.
• Siblings configuration/index.mdx:13 and series-mapping.mdx:18 describe the split generically and aren't invalidated. No docs.json change needed (no new page).
• Legacy docs/explore-analyze/charts/configuration/small-multiples.mdx correctly left alone per CLAUDE.md.
• Frontmatter description intentionally left as-is — agreed with your reasoning; not re-raising.
• N/A for this PR: security, performance, tests.

Unverifiable from the repo

• The and by picker label, the toggle beside Split by, the cross on the second picker only, and "a dimension already used on one axis is not offered on the other" are UI claims with no source in this repo. You've cited the code for each on the threads, so this is just a note that the page can't be checked from here — and the PR body already flags that it depends on the UI change landing first.

Still true from earlier passes: this carries the 5 × 5 correction from #11498, so whichever lands second will conflict on lines 28 and 43.