Beliefs
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.
These are not commandments. If better evidence comes along, we reserve the right to change our minds.
- 01
If you’re sure, show us why.
“We know this doesn’t work” isn’t very useful if nobody can find the test, the numbers, the circumstances or the outcome. Confidence isn’t evidence.
- 02
Marketing memory without evidence is just folklore.
“We tried generics and they don’t work” may once have been completely reasonable. Five years, three agencies and twelve retellings later, nobody knows what was actually tested. The conclusion survived. The evidence didn’t.
- 03
What actually happened beats the story we tell about it.
Reports, decks, summaries and recollections are interpretations. They’re useful, but they shouldn’t quietly become more authoritative than the decisions, evidence and outcomes they were supposed to describe.
- 04
Keeping the file isn’t the same as keeping the knowledge.
Organisations are brilliant at storing things and terrible at remembering why they mattered. The deck can survive for ten years while the context disappears in six months.
- 05
Stop paying to learn the same thing again.
Organisations repeatedly spend money answering questions they’ve already answered. New team, new agency, new strategy, same experiment. If we learned something once, we should at least be able to find out what we learned and why.
- 06
“We decided X” isn’t enough.
Decisions need their reasoning. What did we know? What were we trying to achieve? What alternatives did we reject? What constraints mattered? Without that, a perfectly sensible decision can look completely irrational two years later.
- 07
AI can read the record. It doesn’t get to rewrite history.
Let AI connect things, summarise them and help us understand them. But a convincing answer doesn’t become true because a model produced it. We should always be able to get back to the evidence.
- 08
“We don’t know” is a perfectly good answer.
Missing evidence is information. Contradictory evidence is information. Uncertainty is information. Filling the gaps with something plausible just gives you a nicer-looking problem.
- 09
Make it cheap to change your mind.
We’re going to be wrong. Requirements will change. Better evidence will appear. The product should make revision normal rather than treating every previous decision as sacred. Keeping the cost of change low is one of the strongest recurring ideas in Getting Real.
- 10
Start by leaving things out.
Every feature sounds reasonable when you discuss it on its own. Then you end up with forty reasonable features and one horrible product. Features, fields, settings and integrations should have to earn their way in. Getting Real repeatedly makes the same case for doing less and letting features prove their value.
- 11
If the software creates more work, something’s gone wrong.
Nobody needs another system to feed, clean, organise and apologise for. Good software should remove searching, reconstruction, repetition or administration. Preferably several of them.
- 12
Don’t make people stop working so they can document the work.
The useful context is already being created in emails, reports, conversations, decisions and tools. Capture it there where you can. A separate knowledge system everybody has to remember to maintain will eventually become another abandoned knowledge system.
- 13
More data can just mean a bigger mess.
A system doesn’t become more useful because the database got larger. As information accumulates, it should become easier to understand relationships, decisions, outcomes and patterns. Otherwise we’ve just built a more expensive cupboard.
- 14
Complexity always sends the bill later.
One more integration. One more workflow. One more setting. One more approval. They rarely look expensive when they arrive. Someone still has to understand, maintain and eventually remove them.
- 15
Nobody should be impressed by the plumbing.
Infrastructure exists to support the work. If everyone has to understand the machinery before they can get something useful done, we’ve made the machinery too important.
- 16
If leaving means losing your data, it was never really yours.
People should be able to get their information out in a form they can understand and use somewhere else. Lock-in is not a retention strategy we’re interested in. Getting Real makes the same case explicitly: easy in, easy out, and let people take their data with them.
- 17
Boring file formats have an excellent survival rate.
Clever platforms disappear. Companies get bought. Products get shut down. Plain text keeps opening. We’ll take understandable and portable over clever and trapped.
- 18
If we can’t explain it plainly, we probably haven’t understood it properly.
Jargon can hide a surprising amount of fuzzy thinking. Writing forces decisions. What happened? Why does it matter? What are we actually saying? Getting Real makes essentially the same connection between good writing and organised thought.
- 19
The number isn’t the explanation.
Revenue fell. CPA rose. Conversion improved. Fine. That tells us what moved. A dashboard doesn’t automatically tell us why it moved, what somebody changed beforehand or whether the two things are even connected.
- 20
Make easy things easy and dangerous things slightly annoying.
Writing a note shouldn’t require a process. Neither should fixing a typo. Deleting evidence, changing production infrastructure or making something irreversible deserves a little friction.
- 21
Keep the evidence. Keep the context. Keep learning.
We’re not trying to preserve old conclusions forever. Quite the opposite. We want enough of the record to know when an old conclusion still makes sense, when it doesn’t, and why we changed our mind.