Most strategies are shown after they work. The ones that didn’t are left out. That is survivorship bias, and it makes a published result look better than the process that produced it.
If you only see the study that passed, you can’t tell how many were tried, what changed between them, or whether the final design was fitted to the data. The result may be real. You have no way to check.
So we publish every study, in order, whatever the outcome. A failed study gets the same layout, the same place in the ledger and the same four answers as one that works:
- What we tested.
- What we expected.
- What happened.
- What it forces us to change.
The fourth answer is the point. A failure that changes nothing is noise. A failure that forces a specific change is evidence. It shows what the design had to survive, and it lets you check that the next study dealt with it rather than worked around it.
There is a second reason. Each design is sealed before its study runs and published with it. If the only studies you ever saw were passes, you would have to trust that no sealed design was adjusted after the fact. With the failures on the record, the sequence is visible: design, result, change, next design.
This has a cost. Failed studies are unflattering, and they stay public. We think that is the price of being checkable.
The same rule applies once the book runs. What went wrong is published next to what went right, with the same prominence.