Sense and decide
Pre-registration
Write down what would count as success before a pilot runs, so nobody can call whatever happened a win afterward.
Borrowed from pre-registration in science, where researchers record their predictions and analysis plan before they collect data. Read the original source.
Ongoing habit · Minutes to start · Anyone on the team can start it
Try this first
Write one line before the pilot. What result would make this a success?
Use it when
Any outcome risks being spun as success after the fact.
Skip it when
It suits tests meant to inform a decision. Pure exploration may not have a criterion yet, though naming what you are looking for still helps.
How to introduce it
Before a pilot, write down what result counts as success and how you will judge it. An AI tool trial is the usual case. Commit before you see the outcome. This kills the trick of running a trial and then moving the definition to fit the result.
How to show up
Make them define success in advance and write it down. Then hold them to it when the result disappoints. The rewriting afterward is what you are guarding against.
How long it takes
A short step before a pilot to define and record success.
What makes it hard
People want room to maneuver and resist a standard they might miss. That resistance is the thing to work through. After a bad result the pull to redefine success is strong, and the written commitment is what holds.
What it looks like when it's working
Pilots carry a standard set in advance and get judged against it, including the ones that fail. Every trial ending in success means the definition is written afterward.
How long until it sticks
A few pilots before pre-committing to success criteria becomes routine.
How you know it stuck
Someone asks what would make this a success, and writes it down, before any test runs.
The idea behind it
Decide what success means after you see the result and any result becomes one. Committing first is what lets a test tell you anything.
Where it comes from
Adversarial collaboration. Write down what a pilot must show to count as a win before running it. The antidote to declaring any outcome a success. State your hypothesis and how you'll analyze it before seeing the data, so you can't rationalize the result after the fact.