Case Studies
Every study here is a real experiment, start to finish.
Same method, different industries.
Each one walks a single experiment end to end: the problem, the variables that were tested, the design, and what the numbers said. Every one names its source, and the data comes from published experiments or our own work.
Some of them can also be loaded into Blueprint, so you can examine the results yourself.
Featured experiment
Which of Your Nineteen Good Ideas Actually Works?
A Fortune 500 financial firm had nineteen ideas for improving a credit card mailer and no way to tell which ones worked. Testing them one at a time would have taken 475,000 prospects. Using a highly efficient structured experiment instead, they tested all nineteen at once with 100,000 prospects — and got a better answer.
Plackett–Burman20 runsMarketingRead the experiment
One Mailing: More Orders, Lower Fees, Higher Profit
A screening test beforehand had narrowed nineteen ideas to five. This follow-up experiment tested four pricing levers in all sixteen combinations — and found that the combination which resulted in the largest customer response was not the one the company should pursue.
Marketing · Oct 2026Read the experiment
Your Average Part Size Is Fine. So Why Don't the Parts Fit?
Sixteen carefully selected runs, instead of a full set of 64, found the two variables that set average shrinkage of injection molded parts. A third variable didn't affect the average at all, yet was driving the rework. The scrap problem was never the average size of the parts — it was the consistency between them.
Manufacturing · Sep 2026Read the experiment
Maximum Cleaning Power, Minimum Cost: Optimizing a Vacuum Brush Roller
Two of the three surviving variables cost almost exactly the same to change, but one bought nearly four times more cleaning performance per dollar. A 2³ factorial with five replicates optimized cleaning performance while keeping costs down.
Consumer Products · Sep 2026Read the experiment
Which Variables Matter? Screening a Vacuum Brush Design in 12 Configurations
A consumer product team believed all seven of their brush-roller variables affected cleaning performance of a new household vacuum cleaner — they just didn't know which ones mattered most. Twelve configurations settled it: two variables carried 92% of the variation, and one they were certain about turned out to do nothing at all.
Consumer Products · Sep 2026Read the experiment