should measure as much as possible with the least risk. For example, we might
run a market test:
✦ After an iterative UCD process of design consulting, exploratory testing,
field studies, and validation testing has resulted in extensive changes to the
UI.
✦ During a relatively level sales period; in e-commerce it’s not recommended
to test new ideas in November and risk Christmas sales, even if only part of
the target audience experiences a new site design.
Measuring external ROI is usually more difficult than measuring internal ROI,
because so much of the improvement takes place outside the UCD group.
Nevertheless, careful observation of confounds, supplemented ...