Answer in brief
A design rule is worth keeping when it can be traced to a research finding, states what it is for and can be checked by someone who was not in the room. Sort research into data, findings and insights, turn the strongest insights into a few principles that choose one value over another, then write each rule with its intended outcome, its reason, an example, a counter-example and a test. Anything that fails those checks is a preference and should be labelled as one.
Research ends in a folder unless someone writes the rule
A design team finishes a round of research with a great deal of material: interview recordings, an audit of every leaflet and screen the organisation has produced, screenshots of competitors, survey answers. A month later the brand guidelines are written, and it is often hard to point to a single line in them that exists because of that material. The research informed the mood. It did not become rules.
The opposite failure is just as common. Guidelines fill with instructions nobody can trace: a minimum logo size, a forbidden colour pairing, a preferred style of photography. Some rest on evidence and some record what the lead designer liked that year, and the document gives the reader no way to tell which is which.
Closing that gap is what this guide is about. Most advice on how to write design guidelines begins with structure and tone; the steps here begin earlier, with the evidence. Research is sorted by how far it has been analysed. The strongest conclusions become principles. Each principle produces rules that name something visible, and each rule carries its reason, an example, a counter-example and a test.
Data, findings and insights are three different things
Nielsen Norman Group, the user-experience research firm, separates research output into three levels. Data is the unanalysed collection of observations: transcripts, notes, metrics, survey output. Findings describe patterns in that data, what happened but not why. Insights explain the patterns in context and point to an opportunity the organisation can act on.
The distinction matters for a design team because rules written from the wrong level fail in predictable ways. A rule built on one data point, a single memorable quote, is an anecdote with authority. A rule built on a finding alone treats a symptom. For example, “customers could not find the price” says what happened, not whether the cause was size, position, colour or wording.
So before writing anything, label the material. For each candidate conclusion ask how many sources support it and whether you know the cause. To reduce bias, NN/g recommends triangulation: several sources of data, several ways of analysing them and more than one person doing the analysis. An audit that agrees with the interviews and with the support inbox is ground you can build a rule on.
A principle picks a side and says why
An insight tells you what is true about your audience. A principle says what you will do about it whenever two good things conflict. NN/g defines design principles as value statements that frame design decisions and support consistency in decision making across teams. It stresses that they are specific to a product or service, unlike general rules of good design, and that they are not directions for designing individual elements.
The article lists what makes a principle usable. It takes a stand on which value matters, ideally by naming the value it wins against. It says why that value matters to the people served. It is concise and memorable, and the set is small: NN/g warns that with ten or more, many are forgotten. And the principles do not contradict one another. Its own example for its website is “Clarity over popularity”.
For a graphic design team the same form works. “Legibility before atmosphere” is a principle if research showed readers abandoning beautiful, low-contrast pages. “Clean and modern” is not, because it gives up nothing and nobody would argue the opposite. A useful test is to ask what the principle forbids. If the answer is nothing, it is a slogan.
From a principle to something a reviewer can see
A principle guides judgement. A rule can be checked against a finished piece of work. The step from one to the other is to name three things: the element the rule governs, the condition in which it applies, and the limit that must not be crossed. Without all three, a reviewer has to guess.
Take the principle about legibility. A rule that follows from it might read: body text is never placed over photographs; captions over images sit on a solid panel. The element is body text, the condition is a photographic background, the limit is “never”. A second rule might set a minimum contrast for all text on colour. These are illustrations, not recommendations, but they show the shape.
Write each rule so that a person who missed every workshop could apply it. Avoid adjectives that need the author present to interpret them: “generous” margins, “subtle” gradients, “appropriate” imagery. Where a number is possible, give the number. Where it is not, give the comparison: wider than the column gutter, lighter than the body text, no more than one per page.
Four parts that make a rule checkable
A good public model for writing rules comes from an unexpected place. GS1, the standards body behind barcodes, publishes a guideline on how product packaging should be shown in small images on phones. It is a document about graphic design decisions, and every individual guideline in it is presented in the same four parts.
The first is the intended outcome: what should be true once the rule has been followed. The second is the explanation: why the outcome is included and how it helps the person looking. The third is a possible approach to implementation, offered as the best advice at the time of writing, with the note that other methods may suit other circumstances. The fourth is how to test whether the outcome has been achieved, regardless of the method used.
That structure does two useful things for a brand guideline. It separates the goal from the technique, so a designer who finds a better technique is still compliant. And it forces the author to write the test. A rule whose author cannot describe how to check it is not ready to publish. Some of GS1’s tests are admitted to be subjective, and the document says so, which is more honest than pretending to precision.
Examples show the rule, counter-examples show its edge
People learn a rule faster from a pair of pictures than from a paragraph. An example shows what compliance looks like. A counter-example shows where compliance stops, and that boundary is what a designer under deadline actually needs. The GS1 guideline does this throughout: it sets conformant images beside non-conformant ones and explains, case by case, why only the first passes.
Choose counter-examples from real work, not from caricature. A logo stretched to twice its width teaches nothing, because nobody was going to do that. A layout that follows every written rule and still fails the principle teaches a great deal. The audit from the research phase is the best source: it already contains the organisation’s own near misses.
Annotate both. Under the example, name the rule it satisfies. Under the counter-example, name the rule it breaks and the smallest change that would fix it. When a case is truly borderline, say so and record the decision. A guideline that admits its grey areas is consulted more often than one that pretends to have none, because it matches what designers meet in practice.
Three questions that separate a rule from a taste
Every guideline contains preferences, and there is nothing wrong with that, provided they are labelled. The trouble starts when a preference is written in the voice of a rule. Three questions sort the two. Can the statement be traced to an insight from the research? Would two reviewers reach the same verdict when applying it to the same piece? Is there evidence that it helps in more than one situation?
The third question is the one design systems have formalised. The GOV.UK Design System’s contribution criteria ask for evidence that a proposed component or pattern would be useful for many teams or services, and that it does not replicate something already there. Before publication it must have been tested in user research with a representative sample of users, including those with disabilities. Where something has been tried in only a few contexts, the team says it will be clear about which ones.
The GS1 guideline states the same idea as a proportion: the more prescriptive an individual guideline is, the stronger the research base it builds on. That is a workable policy for any team. A firm “never” needs strong evidence. A “prefer” can rest on less. A statement with no evidence at all can stay in the document, under a heading that calls it house style, where nobody will mistake it for a finding.
Rules stay alive only if decisions cite them
A set of principles and rules earns its keep in reviews. NN/g’s advice is to use principles to explain and justify decisions: a trade-off is settled not by someone’s say-so but because a named principle tips the balance towards one of the options. The same habit works for rules. A review comment that cites a rule number is a different conversation from one that begins “I feel”.
Keep a short log beside the guideline. Each entry records a case where a rule was applied, bent or found wanting, and what was decided. A rule that is broken for good reasons three times is telling you something: it is too strict, too vague or aimed at the wrong element. That log is new research, and it costs nothing to collect.
Then review on a schedule. For each rule ask whether its evidence still holds, whether its test still works and whether anyone has used it. Remove what has gone stale, promote house-style notes that have gathered evidence, and date every change. The aim is a document a newcomer can trust: every line in it either points to something the team learned or says plainly that it is a choice.
Practical checklist
- Label every research conclusion as data, a finding or an insight before any of it is used to justify a rule.
- Write a small set of principles that each prefer one value over another and state the reason for the preference.
- Give each rule an element, a condition and a limit, plus its intended outcome, its explanation and a test.
- Pair every rule with one conforming example and one realistic counter-example taken from the audit.
- Move statements without evidence into a section marked as house style, and review all rules on a fixed schedule.
Questions and answers
What is the difference between a design principle and a design rule?
A principle is a value statement that helps a team choose between two good options, for example clarity before decoration. A rule is a specific, checkable instruction that follows from a principle and names an element, a condition and a limit. Principles guide judgement in situations nobody predicted; rules let a reviewer confirm that a finished piece complies.
How many design principles should a team have?
Few enough to remember without looking them up. Nielsen Norman Group advises against long lists and notes that with ten or more, many are forgotten. A handful that each take a clear position is more useful in a review than fifteen that try to cover everything.
How do you turn user research into design guidelines?
Sort the material first: raw observations, patterns across them, and explanations of those patterns in context. Write principles from the explanations, then derive rules that name what is governed and where the limit lies. Give every rule its reason, an example, a counter-example and a way to test it, and keep a note of which research supports it.
How can I tell whether a guideline is evidence or personal taste?
Ask whether it can be traced to something learned in research, whether two people would judge the same work the same way using it, and whether it has helped in more than one case. A guideline that passes is a rule. One that fails may still be worth keeping, but it should be marked as house style so that readers know it is a choice and not a finding.
Should brand guidelines include examples of what not to do?
Yes, as long as they are realistic. A counter-example marks the edge of a rule, which is where most real decisions are made. Take them from the organisation’s own past work, annotate which rule each one breaks and show the smallest change that would bring it into line. Exaggerated mistakes that nobody would make add length without adding guidance.