Not a GOV.UK service, but I quite like how google deal with international phone number entry. The select-list combined with country icon and name is nice. Though it introduces other problems for people who struggle with select lists. Having a default (eg UK) appears to work well and makes it obvious which it is, and how you change it.
It sort of feels like we should let browsers deal with this. If it's marked up as a telephone number it's for the browser to do something sensible with it. NB I've used `<meta name="format-detection" content="telephone=no">` to try to disable auto-detection in some cases - though mostly only safari uses it AFAIK.
Use this section to draft guidance for the pattern. This will become the first draft of the pattern page when it's added to the Service manual, at which point it will be removed from this page. It's quite common for a single pattern page to describe a number of different approaches. We're using the following structure to document each approach:
Example: 1 line intro plus screenshot or HTML
Pros: What this pattern is good at and when to use it
Cons: What this pattern is less good at and when not to use it
Guidance: General guidance on implementing this pattern
Was any A/B testing performed? That would be particularly valuable.
I'd also love to know whether you've tested a registration or login process with only one question per screen. Once, when I was a user of such a form, I found it frustrating, and it also slowed me down, because it was just so unexpected. De facto conventions may go against the pattern in a case like login/registration.
I appreciate that this is a good pattern for small screens. On Twitter Caroline indicated that more than 30% of your forms are filled via mobile. I was wondering if you had tested the pattern on desktop/large screens (which are still the vast majority of your usage)? Also, did you think about allowing for a different approach on large screens versus small screens (e.g. allowing multiple questions per screen when screen is above a certain size)? The large screen offers the opportunity to provide context that may help the form filler (e.g. grouping questions in to steps and showing what steps there are).
One thing that I'm not clear on is how the pattern reconciles with the pattern on progress indicators. If you have 20 questions, one per screen, should you have a progress indicator with 20 steps? Or are you advocating dropping the progress indicator altogether (given some of your other research has shown it doesn't really help users)? If it's the latter, I wonder if people might not feel the need for signposting more when there are more screens/questions not grouped into steps (Am I half way through? 10% done? Nearly finished?).