[Sigia-l] Real World UI Design Failure

[Jonathan Baker-Bates]

On 12 January 2011 20:28, darin sullivan <darinqsullivan at gmail.com> wrote:

> Thanks Jonathan... That's great stuff and the primary reason that I've
> continued following discussions on this list :)
>
>
Thanks - and thanks also to you for replying! After a blizzard of emails
recently on this list about the conduct of the list itself, but until now
not a peep amount my posting, I was beginning to wonder whether information
design was now too boring to talk about!


> Would you please comment on how your group has proceeded to increase use of
> results sorting since this attempt?
>

We originally had the sort drop-down above the filters on the left hand side
(not shown on my screenshots). The hypothesis here, somewhat supported by
observation, was that people often think about sorting and filtering as very
similar things. So we "zoned" the sort above the filters thinking that
people would discover it more easily. But usage went down - quite possibly
for the reason you give.

So we put them to the right as you see them today. So far the test results
haven't reached statistical significance so I can't say if that has restored
their usage yet. If they are still down (results will be mine in about a
week), I'll try your idea of zoning it with the search results at the same
visual "level". In fact we can try that anyway in a bid to get to the "local
maximum".

As an aide that may be interesting to those who have not had the luxury of
designing using a mutli-variate testing platform, I can say it's a bit of a
mixed blessing. The obvious Achilles heel of bugs invalidating test results
(which happens rather a lot when you're testing complex UI that deal with
complex back-ends) is one thing, but the issue of interpretation can also be
a gigantic problem. This gets exponentially worse the more granular stuff
you test, as the more fraught subsequent design decisions become, and
secondary maintenance issues start to proliferate (we recently conducted 8
weeks of lab studies, during which time the site changed underneath us
several times). I could tell stories of the most Byzantine nature imaginable
when you're looking at stats while running 20+ concurrent tests in 40
countries, all reporting pretty different results across several metrics!

But on the other hand, the occasional universally clear result of a test can
sure settle any arguments :-) Would that were a more frequent event is all I
can say!

Jonathan