General Process Best Practices (GPBP)

GPBP1

Determine a good process, then automate or create a checklist.

Prefer scripts (programs) to manual processes.

GPBP2

Get someone else to check results, compliance, and assumptions.

Adopt ☞ reproducible research practices; do code reviews.

GPBP3

Document everything but do not assume people will read docs.

Data dictionaries, README files and .tdda files.

GPBP4

Have context-specific prompts and reminders when & where relevant.

Write good prompts; check consistency; validate data.

GPBP5

Use clear labeling and disambiguation.

Label code and data with version numbers or datestamps.

GPBP6

Adopt consistent conventions and standards.

e.g., naming conventions; coding standards.

GPBP7

Perform root-cause analysis when failures occur; address causes.

Create regression tests; use data checks at all stages; write logs.

GPBP8

“Design out” failure points and failure modes where possible.

Use pre-commit checks, ☞ linting.

Monitoring analytical processes and run reference tests.

GPBP9

Build robust systems that assume mistakes and faults will occur.

Check data at all stages;

Use ☞ defensive programming and version control.

GPBP10

Build systems that minimize the harm when failures occur.

Use alerting, monitoring, data checks, and version control.

GPBP11

Use redundancy for key processes, tests, and checks.

GPBP12

Make it easy to do the right thing and hard to do the wrong thing.

Incentivize reporting of failures.

Establish a “no blame” culture.

Train people properly.

Publish and learn from mistakes.

GPBP13

Maintain alertness.

Constantly ask: How is this misleading data misleading me?