What We Try Not to Forget

21 principles for useful work.

These are things we currently believe. Some are about evidence. Some are about software. Some are about organisations. Some are simply useful ways to work.

They are not commandments. If better evidence comes along, we reserve the right to change our minds.

1. Show evidence, not just confidence. A strong claim is only as useful as the material underneath it. When somebody says something worked, failed, improved, declined or “never works for us”, we should be able to get back to the evidence that made that conclusion reasonable.

2. Marketing memory without evidence is just folklore. Conclusions survive surprisingly well after the reasons for believing them have disappeared. Over time, “we tested this once” becomes “we know this does not work”. Preserve the evidence or eventually you are just preserving a story.

3. Prefer the thing that happened to the story about it. Reports, recollections and summaries are interpretations of reality. They are useful, but they should not quietly become more authoritative than the underlying events, decisions and outcomes they describe.

4. A file without context is only half remembered. Storage preserves artefacts. Memory preserves what they meant: what changed, why it changed, what happened afterwards and what was learned.

5. A lesson already learned should not become a lesson relearned. Organisations repeatedly pay to rediscover things they once knew. If the evidence still exists, the next team should not have to start from zero.

6. A decision should carry its reasoning with it. Recording what was decided is not enough. Future teams need to know what evidence existed, what constraints mattered, what alternatives were considered and why the decision made sense at the time.

7. AI can interpret the record. It does not become the record. AI can connect, summarise and explain information. It should never erase the distinction between an interpretation and the evidence that supports it.

8. An honest gap is better than a confident guess. “We do not know” is useful information. Weak evidence, missing evidence and disagreement should remain visible rather than being polished into false certainty.

9. Make changing your mind cheap. Better evidence is only useful if the system allows beliefs, decisions and interpretations to change when the evidence changes. Design for revision rather than pretending the first answer will be permanent. Getting Real makes the same broader argument about keeping the cost of change low. getting-real.

10. Default to less. Every feature, field, workflow, report, dependency and setting creates future work. Add something when the value clearly outweighs the complexity it introduces, not merely because it might be useful one day. The book repeatedly ties less mass, fewer features and simpler systems to greater adaptability. getting-real.

11. Useful software should remove work. A technically impressive system that creates another administrative job has failed somewhere. The test is not how sophisticated the machinery is; it is whether somebody now has less searching, less reconstruction, less repetition or less uncertainty.

12. Memory should sit inside the work, not beside it. Do not create a second reality that everyone has to maintain manually. Capture useful context from the places where the work already happens whenever that can be done responsibly.

13. A useful system should compound, not just accumulate. More data is not automatically more knowledge. As the record grows, the system should become better at connecting decisions, outcomes, evidence and recurring patterns rather than simply becoming a larger archive.

14. Every extra layer becomes somebody’s future chore. Complexity rarely arrives announcing itself as complexity. It arrives as one harmless feature, one integration, one approval step or one more system to maintain. The maintenance bill appears later.

15. Infrastructure should leave the work more visible than itself. The machinery is doing its job when people can concentrate on the useful thing rather than the system underneath it. Good infrastructure should make the work easier to see, not demand attention for its own sake.

16. Your data should still be yours when you leave. A useful product should not depend on captivity. People should be able to understand, export and retain the information that belongs to them. Getting Real makes this principle explicit in its argument for open access to customer data and easy exit. getting-real.

17. Plain formats age well. Proprietary systems come and go. Simple, readable formats give information a better chance of remaining understandable long after the original software has disappeared.

18. If you cannot explain it clearly, you probably do not understand it yet. Good writing exposes fuzzy thinking. Clear language forces you to decide what actually matters and makes the resulting work easier for the next person to understand. Getting Real treats good writing as evidence of organised thinking, not merely a communication skill. getting-real.

19. A metric can show movement without explaining cause. Numbers are observations, not explanations. A dashboard can tell you that revenue fell, spend rose or conversion improved. It cannot, by itself, tell you why.

20. Put friction where mistakes are expensive. Easy things should stay easy. Publishing a note should not require ceremony. Changing production infrastructure, deleting evidence or making irreversible decisions should require more care.

21. Keep the evidence. Keep the context. Keep learning. The aim is not to preserve the past unchanged. It is to make yesterday useful to tomorrow without turning old conclusions into permanent rules.