These are small, honest case studies from building Mac and iPhone apps in public from Japan. The numbers are not inflated and the misses stay in the record.
We use Rank to choose a Yuzool page, record the baseline, ship one controlled change, and return after 7, 14, and 28 days.
Dispatch Kun is treated as a distribution experiment: explain one job, share one link, listen for the language people use, and improve the next release.
Each page, public reply, and product update is a small test. We track attention, intent, evidence, and revenue, then publish what changed.
Customer language is more useful than a polished claim. These are the words behind the product decisions, support replies and next screenshots.