Write Design Rules People Can Use Without You
By Aldridge Dagos, operations software engineer
If your design guide only works when you sit beside the person using it, the guide has not left your head.
A teammate opens the file, follows the listed colors and sizes, and produces a page that obeys every visible rule. It still looks wrong. The image competes with the title, the title has no pace, and the accent appears everywhere because the guide never says what deserves it.
The creative lead steps in, repairs the draft, and explains the choices during review. The page improves. The guide learns nothing. Next week, the same rescue happens again.
This dependency does not come from a teammate lacking taste. It comes from granting permanent status to the surface of one good example while leaving its judgment temporary and private.
A design guide should remember how a direction earned the right to repeat. The lead starts with taste and tests the choice until a real artifact can serve as the gold standard. The guide records what made the choice hold, production repeats it, and new evidence decides whether that answer stays.
That last decision matters most. A guide turns into a cage when old instructions keep their authority after the work has disproved them.
Living guide 01
Creative judgment becomes memory, then returns for revision
- 01 Judgment Creative lead sets a direction
Taste begins in a person.
- 02 Experiment Drafts test the choice
Possibilities stay temporary.
- 03 Gold standard A real example proves it
The result becomes the reference.
- 04 Reasoning The guide records why
Choices gain usable boundaries.
- 05 Production People repeat the direction
Work no longer waits for rescue.
- 06 Revision Feedback changes the memory
The guide stays alive.
- Fixed
- Identity rules that require approval to change
- Evolving
- Judgment backed by examples and reasons
- Temporary
- Experiments that have not earned reuse
Taste gets the first vote
Take a familiar design choice. A creative lead places an oversized issue number beside an article title. The number gives the archive a strong vertical edge and lets the title carry the subject without another decorative image above it.
Nobody reaches that choice by averaging a checklist. The lead looks at the writing, tries several weights, rejects the timid versions, and selects the one that gives the page its character. Taste supplies the first direction.
At this stage, the large number is an experiment. It may look convincing on the first article because the title happens to fit on two lines. The image and a generous desktop canvas may hide every difficult spacing decision.
Calling it a rule now would confuse confidence with proof, so the lead needs to put the idea under pressure. Try a longer title. Remove the hero image and see whether the number still holds the composition. Check the archive beside the article page. Narrow the viewport until the title wraps and the date begins competing for space.
When I review an issue-number treatment, I do not approve it from the wide view alone. I put it beside a difficult title and reopen it at 375 pixels, where weak spacing choices stop hiding inside extra room.
Some parts will survive. Perhaps the number keeps the archive recognizable across every subject. Other parts will change. On a small screen, it may need to move above the title instead of squeezing beside it. That is not a failure of the idea. It tells you which part carried the identity and which part only suited the first canvas.
This is the moment when taste becomes teachable. “Make the number large” only describes appearance. “Use the issue number to establish the reading edge, then move it above the title when a side-by-side layout weakens the title” gives the next person judgment they can carry into another layout.
You still need the example. Reason without an artifact can float into interpretation. The tested article becomes the gold standard because it shows the direction holding through the difficult content, not because it was the first draft the lead liked.
An experiment needs a status and an owner
Design guides become confusing when they give every note the same authority. A logo clearance rule and a crop idea from yesterday should not look equally permanent. Give every note one of three states: fixed, evolving, or temporary.
Keep the fixed state small. The approved mark, type family, core palette, focus treatment, and minimum contrast may belong there because the brand or the reader depends on them. Because changing one of those rules requires deliberate review, a casual draft should never rewrite it.
Most creative direction belongs in the evolving state. Composition, image weight, density, motion, and pacing depend on the content. The guide can explain the current judgment and show proven examples without pretending the choice will survive every future case.
New ideas need an unmistakable temporary label. If the team tries a tighter crop or a louder title, the draft should name who is testing it and what question the experiment should answer. Otherwise, another teammate may copy it tomorrow and turn an open question into accidental policy.
The creative lead owns promotion. Teammates can bring examples, failures, and better language. They can show that a choice works across the actual range of material. They should not have to guess when an experiment became part of the brand.
Ownership does not mean the lead approves every production draft. It means one person carries the decision to change the shared direction. Without that role, rules enter through repetition. The team sees something twice, assumes approval, and spreads it before anyone has tested its boundary.
If you lead the work, give experiments an answer. Move a choice that holds into evolving or fixed guidance, but keep promising work in test and delete failures before other people copy them. A permanent “maybe” state forces every producer to decide alone.
Promotion requires a reason, not a vote count
An idea does not earn the guide because three people like it or because one page shipped with it. It earns reuse when the team can point to what improved, the conditions it survived, and the edge where it should stop.
Return to the issue number. The test shows that it strengthens archive scanning, keeps its place across light and dark images, and survives titles of several lengths. On mobile, placing it above the title protects the reading width. The lead now has a case for promotion.
The guide should capture that case in plain language. It names the job of the number, links to the gold-standard page, shows the wide and narrow arrangements, and marks the condition that causes the layout to switch. The exact pixel value belongs in the component. The reason and boundary belong in the guide.
This separation matters. A component library carries the working part and its supported states. The design guide carries why the team uses it and how to judge a new case. Link the two so code and judgment can change together without copying either into the other.
The same standard applies before another teammate or an AI tool repeats the direction. Give them the established example, reason, and boundary, then ask them to work within what those materials establish and mark any exception.
AI can generate another arrangement, compare a draft with the standard, or apply the approved component across many pages. It should not look at one experimental draft and decide that the brand now speaks that way. The creative lead establishes the direction through tested work and documents it first.
Promotion also needs a removal plan. If the new pattern replaces an older one, mark the older direction and find the live components that still depend on it. A guide that shows both without status leaves the team with competing instructions.
Repetition is another test
Documentation does not finish the design. It changes who can participate in producing it.
A teammate uses the approved article pattern on the next issue. This time, review does not begin with the creative lead fixing the obvious structure. The draft already belongs to the same family. The useful conversation starts with the subject: does this title need more room, does the image deserve less weight, and did the current rule preserve the right emphasis?
That is what repeatable production should buy. It moves review away from repairing forgotten basics and toward judging the new material.
The teammate also becomes a source of evidence. They may discover that the number works for ordinary issues but overpowers a short announcement. An editor may supply a title that breaks the mobile arrangement. A developer may find that a documented state has no matching component behavior.
Do not treat these reports as disobedience. Production reaches combinations the gold standard could not contain. The guide should make it easy to return an exception with the page, condition, and observed result attached.
One exception may deserve a local answer. The same rescue appearing across several drafts points somewhere else. The rule may lack a boundary, the component may contradict the note, or the direction may no longer fit the work.
The creative lead reviews that evidence and decides what changes status. They can add a missing condition, create a second approved arrangement, demote the old rule to a historical note, or retire it completely. The guide stays authoritative because the lead records that decision and updates the shared example.
Let production edit the memory
A frozen guide can preserve a brand’s appearance while making the work worse. People follow an old crop because it still says approved. They force a new article into a layout built for shorter writing. After enough exceptions, the creative lead spends review time granting permission to ignore the document.
That is the cage: a rule survives after it stops helping.
A living guide keeps a visible return path. When production exposes a problem, the evidence goes back to the person who owns the direction. The lead decides whether the case is local or whether the shared memory should change. Then the guide, example, and component move together.
You should be able to tell what happened to an old instruction. Mark when the team retired it and why. Point to the new standard. Anyone maintaining an older page can understand the difference without treating history as current advice.
This record also protects experiments. The team can try a bold idea without pretending it will last. If the test fails, remove it cleanly. If it reveals a better direction, the lead can promote the useful part and leave the rest behind.
Now the next draft arrives with an issue number the guide never anticipated. It is unusually short, and the approved scale overwhelms the title. The teammate flags the case instead of quietly shrinking everything. The lead reviews the page, changes the rule to account for short titles, and updates the component example. The draft after that begins from the new answer.
A creator should not have to rescue every draft for the guide to work.
Frequently asked questions
What should trigger a review of fixed rules?
Review them on a set cadence and when a brand, accessibility, legal, or production change challenges one. The cadence prevents neglect. A real conflict should trigger review immediately instead of waiting for the calendar.
Can two gold standards coexist?
Yes, when they answer distinct content or channel conditions and the guide explains how to choose between them. Two unexplained standards force each person to choose by personal preference. Two bounded standards give the team useful range.
What if the guide and the component disagree?
Pause new use and ask the creative lead and component owner to resolve the mismatch together. Decide which source reflects the approved direction, update the other, and record the change so the conflict does not return.
How should a new creative lead inherit ownership?
Review the fixed rules, open experiments, recent retirements, and unresolved production exceptions together. The outgoing lead should explain the reasons that still matter. The incoming lead then names which decisions they accept and which need a new test.