Collection
Make the next change small enough to teach you something
Turn an uncertain improvement into a bounded change with a visible signal and a cheap path back.
Explore in any order.
What’s included
1 checklist · 5 protocols
6 ideas
Use a stretch goal to expose what must change
A bigger target is useful only when it changes how you think about the system. Treat the stretch as a probe: it should reveal a constraint, assumption or lever, not demand heroic effort for its own sake.
Read the idea ProtocolCompare two approaches to learn, not to decorate a decision
A comparison is useful when it answers a question. Decide what you want to learn, keep the alternatives as comparable as practical, record the result, and be honest about whether you ran an experiment or merely observed two different cases.
Read the idea ChecklistAutomate a report only with validation and a visible failure state
A scheduled report that silently publishes bad data is worse than a manual report that admits it is late. Automate the path and the checks together: inputs, transformations, freshness, totals, delivery and failure handling.
Read the idea ProtocolTaste, change one thing, then taste again
When a dish feels wrong, several fixes at once make it hard to know what helped. Identify the problem, make one small adjustment that fits the dish, let it integrate, and taste again before changing something else.
Read the idea ProtocolTrack one signal that can change the plan
A dashboard is not progress. Choose an observable signal close enough to the real outcome that a change in it would make you do something differently. Record it at useful checkpoints and connect each review to a decision.
Read the idea ProtocolPause before a reactive choice
A short pause can create enough distance to see whether a decision needs action, one missing fact or deliberate delay. The useful part is not a sacred breathing count; it is interrupting an automatic response long enough to choose the next move.
Read the idea