~15 Options before NN/G says switch to a combobox
7 USWDS: use radio buttons under this count
3 IBM Carbon's small-list cutoff
200+ Options in a typical country field

What happened

NN/G's July 17 guidance is blunt: dropdowns earn their place only in a moderate range of roughly 5–10 options. Below that, show the choices outright. Past roughly 15, switch to a combobox — a text field paired with a filterable list, so users type to narrow instead of scrolling to hunt.

The major design systems already draw their small-list lines low. The U.S. Web Design System says use radio buttons under 7 items. Google's Material Design puts the cutoff at 6. IBM's Carbon goes as low as 3. For genuinely long lists — the 200-option country field is the canonical case — NN/G points to type-to-filter comboboxes as the only sane control.

The sharpest material comes from GOV.UK's accessibility research. Users with disabilities struggle to close dropdowns, try to type into them, confuse the focused item with the selected one, miss options hiding below the fold, and fight pinch-zoom on small screens. GOV.UK's own position: dropdowns are a last resort in public-facing services.

Where the design systems draw the line (option counts)
NN/G combobox switch point
USWDS radio-button cutoff
Material Design cutoff
IBM Carbon cutoff

Why this matters

Checkout is where this bites. The country and state fields are the exact long-list case the research describes, they sit at the highest-intent moment of your funnel, and on mobile every extra scroll is a chance to bail. A shopper thumbing through 200 countries to find theirs is friction you chose.

Variant pickers fail the same test. A 20-option size dropdown hides your catalog behind a click; swatches or buttons show it. Filter menus deserve the same scrutiny — a brand filter with 40 entries crammed into a dropdown is a search problem pretending to be a design choice. And the GOV.UK failure list doubles as legal exposure: accessibility demand letters aim at checkout flows first, and a control a keyboard user can't operate is exactly the kind of thing they cite.

Don't over-rotate, though. NN/G isn't banning dropdowns — a cohesive 5–10 option field that's secondary to the main task is still fine. The useful part is that this is a counting rule. You can audit an entire store against it in an hour.

What to do about it

Count the options in every selector

Walk your product pages and checkout and note the option count for each dropdown. Anything over ~15 goes on the combobox list; anything under 7 is a candidate for radios, buttons, or swatches. It's a one-hour spreadsheet job.

Check what your checkout actually renders

Load your own checkout on a phone and open the country field. If it's a raw 200-option dropdown with no type-to-filter, that's the highest-value fix on the list — it sits in front of every international order you'll ever take.

Turn small lists into visible choices

Anything under the design-system cutoffs — 7, 6, or 3 depending on whose line you take — should be radios, buttons, or swatches the shopper can see without a click. On product pages this doubles as merchandising: visible options get considered, hidden ones get skipped.

Keyboard-test whatever remains

Tab into each surviving dropdown, try typing a value, then try Escape to close it. If any of those fail, you're reproducing the exact failure cases GOV.UK documented — fix the control or replace it before someone else documents it for you.