The day · Page 43
The parallel build: a variant with receipts
For you only
Run your first day the classic way: one build per team, six hours, the whole pair on one prototype. It is simpler to facilitate, and your first day has enough moving parts. This chapter is the upgrade for your second one, and it changes only the middle of the day.
The shape: everyone starts alone. Two hours, one prototype per person, same brief. Then every solo build goes through a fast round of synthetic user testing, AI users clicking through it and thinking out loud. Then each team sits down with its builds and the findings, and decides what the afternoon is for: adopt the stronger one, merge the best parts, or start a third idea both people now believe in. From 12:30 it is the classic flow again, one prototype per team, polished for the real users at 15:30.
Why bother? Because this is one of the few workshop moves with real experimental receipts.
In 2010, Steven Dow and his colleagues at Stanford had 33 people design real web banner ads. Half worked serially: one design, refined five times. Half worked in parallel: three designs first, feedback, then narrowing. Then the ads ran in a real online campaign. The parallel ads earned about 12% more clicks over the full 15 days, and 69% more in the first five. Professionals rated them higher. They were measurably more diverse. And the parallel designers' confidence grew, while the serial group's stayed flat.
The team version came a year later. Dow's group put 41 pairs to work and varied one thing: whether partners brought multiple designs to the table or just one. The share-multiple pairs won on every axis: clicks in a live campaign, expert ratings, diversity, ideas borrowed from each other. The finding I care about most is quieter: rapport rose only in the pairs that shared several designs, and fell in the others. When there are many ideas on the table, you critique the idea. When there is one, you critique its author.