Design Principles
beta · Family standing-standards · Phase undefined · Sizes lean · ~6,850 tokens
The short standing list of value statements a team agrees once and then uses to frame and settle decisions about the experience of its product. Each principle carries a statement, what it rules out, the question a reviewer asks, what it rests on and a decision it settled. It is judged by whether a principle rules anything out, because a principle that rules nothing out is a slogan.
The short card. The reasoning behind every rule here, with the sources, is in
design-principles_companion.md. A fully worked instance is
design-principles_example.md. Where this card calls something the
library’s position, no source read for the bundle publishes it, and the card says so each time.
What this document decides. A design principles document settles how an interaction, a message or a screen should behave for the person using it when two good options compete. It does not settle what the team builds, funds or declines. That test is the library’s own inference, at medium-low confidence. No source read draws the line in so many words, and published sets cross it, so some principles will straddle. The Purpose and Scope section has a field where the document says which ones do.
What to call it. NN/g’s own longer name for this type begins with the word “product”, and it
collides with the sibling bundle product-principles. This bundle does not use that name. Its safe
aliases are “UX design principles” and “experience principles”.
When to use
Section titled “When to use”- The team keeps arguing the same trade-off between two good options, and nothing written says which way to lean. Jay Mistry’s stated goal for a set is to end repeated debates, not all debate.
- Design critique or review comes down to personal preference. GitLab’s critique sessions tell designers to “Connect feedback to user needs and business goals instead of personal preference”.
- A new product or surface has joined, and nobody can say whether the existing set covers it. Jared Spool’s test asks “Have You Evaluated it for This Project?”, and Purpose and Scope holds the answer.
- A set already exists and nobody cites it. Grade it with the rubric below, then rewrite the rows that score 0.
- A team is about to amend or replace a set and wants the reason on record. The Ownership and Changes section holds that record.
When NOT to use
Section titled “When NOT to use”Several documents get mistaken for design principles. The table says which one you likely need.
| You actually need | Because |
|---|---|
| Product principles | Those settle what the team builds, funds or declines. This document settles how an interaction, a message or a screen behaves. The line is the library’s, at medium-low confidence, and published sets cross it. When a principle straddles, say so in the Purpose and Scope field built for it. |
| A design system | NN/g says these principles are “not directions on how to design specific UI elements (like design patterns or standards)”. Component rules and pattern guidance belong to a design system. Purpose and Scope states the relation, because the sources split. |
| General usability heuristics or principles of good design | NN/g separates a product’s own principles from “usability guidelines, heuristics, or visual-design principles”. Heuristics are universal, and a team’s principles are specific to its product. |
| A team charter | A charter governs how a team works. A set of principles frames design decisions about the product. |
| A definition of done | Both are standing standards that work is judged against. The definition of done judges whether an increment is finished. Design principles judge which design choice to make. |
| A product vision | The product vision says where the product is going, and it sits upstream. A principle says how every design decision on the way should lean. |
| A design brief | The design brief hands one chosen concept to the people who design it. A brief may cite the principles, and the principles outlive every brief. |
Pick a variant
Section titled “Pick a variant”This bundle ships one size, lean, so there is no variant to choose. The reason is in
design-principles_companion.md section 4.
Three choices remain inside it.
- How many principles. Three to five is advice, not a finding, and the template ships three principle blocks. Add a fourth or fifth only when you can attach a real decision to it.
- How the design system relates. The set can sit inside the design system or beside it. State which, because the template prescribes neither.
- Which fields to leave honest. If a principle has no source or settled case yet, write N/A and one line of why. Do not invent a source or a case.
Run these tests on each principle before you grade
Section titled “Run these tests on each principle before you grade”- The opposite test. Mistry writes: “If you can’t imagine another successful product using the opposite approach, you don’t have a principle - you have a platitude.” Name the team or product that would reasonably choose the opposite. If you cannot, rewrite the statement.
- The say-no test. Spool asks “Does it Help You Say ‘No’ Most of the Time?” Write down a design the team can no longer ship because of the principle. That is the “What it rules out” field.
- The what-does-it-settle test. Ask what the principle decides. If it decides how an interaction, a
message or a screen behaves for its user, it belongs here. If it decides what the team builds, funds or
declines, it belongs in
product-principles. If it decides both, say so in the straddling field.
Quality rubric (self-grade)
Section titled “Quality rubric (self-grade)”Score each row 0, 1 or 2. A score under 15 out of 22 means several rows are thin or empty, so the set is not yet a standard a designer can lean on. The total does not say which rows failed. Read the rows that scored 0 first, because a missing conflict resolution (row 6) or a missing owner (row 8) leaves a designer with no way to choose or no one to ask, even in a set that scores well elsewhere. Grade what the author added to the blank template. The template already prints the field labels and the Review Trigger conditions, so a row that rewards their presence is satisfied before anyone writes a word.
| # | Criterion | 0 | 1 | 2 |
|---|---|---|---|---|
| 1 | Scope is bounded | No product is named, or the scope covers “everything we make” | The product and its readers are named, but nothing says which kinds of decision the set does not settle, or how it relates to a design system | You can point at the sentence naming the product, the people who apply the set and the relation to the design system, and at one kind of decision the set refuses to settle and where that decision goes instead |
| 2 | Principles rule something out | The rules-out field is blank, restates the statement, or names “bad experiences” | Some principles name a choice a team could have made, and others rule out only what nobody would choose | For every principle you can point at a design the team could have shipped and now cannot, and a colleague who did not write it can state the opposite approach and say who might reasonably choose it |
| 3 | Questions are askable | The reviewer question restates the statement, asks whether the design is good, or is blank | Each principle has a question, but two reviewers could answer it differently about the same screen | For every principle, a reviewer who did not write the set can put the question to a proposed design and reach an answer they could defend in a critique |
| 4 | Principles rest on something | The field reads “best practice” or is blank | A source is named but a reader cannot find it or it does not support the statement, or the field says plainly that no source exists yet | You can open the interview round, decision record or stated value the field names, and it supports the statement it sits under |
| 5 | Cases are real | The field says “many”, describes a case no one can place, or is blank | A case is described, but the same outcome would have followed without the principle, or the option it turned down is not named, or the field says plainly that no decision has met the principle yet | You can point at the critique, ticket or decision record where the principle settled something, and the option it turned down is named |
| 6 | A conflict is resolved | The section says every principle matters equally, or is blank | Two principles are named and a winner is stated, but the kind of decision is “everything” or no reason is given | You can point at two principles, a named kind of decision where they collide, which one wins there and why, and a designer could apply it to the next decision of that kind without asking the author |
| 7 | Use has a place | The section says designers should keep the principles in mind, or is blank | Moments of use are named, but the form of a citation, or what a reviewer may do when a decision sets a principle aside, is missing | You can point at the moments where the set is cited, at the line a decision record carries, and at what a reviewer may do when a principle is set aside, and someone who was not in the room could write that line |
| 8 | Owner is reachable | “The team”, or no owner | A role is named without a person, or the place to propose a change or who decides is missing | You can point at the role and person who own the set, the place where anyone proposes a change and who decides, and you could message the owner today |
| 9 | Changes keep reasons | The log is blank, or each entry is a date and the word “updated” | Entries say what changed but not what fired the change, or a retirement leaves no reason | The log opens with the date the set was first agreed. Each later entry says what changed in a sentence a stranger could follow and names the Review Trigger row that fired it. A retired principle keeps its reason and sits in an archive, not out of the document |
| 10 | Triggers carry owners | The Review Trigger table is blank, or every owner is “the team” | Some conditions have a role and an action, but at least one has neither, or the only observable condition is a date | Every condition has a role as owner and an action that person takes, such as reopening the set or writing a dated line in the log, and any condition that has already fired points at its line in the change log |
| 11 | Direction, not craft | Several principles name a component or a pattern, or settle what the team builds or declines, or the set holds no principles, or the straddling field is blank | The statements are clean, but the straddling field says “None” with no reason, or a principle that straddles the product-principles line is not named in it | At least one principle is present, and you can apply the what-does-it-settle test to each: it decides how an interaction, a message or a screen behaves, and the straddling field names every principle that also decides what is built, or says “None” and gives the reason |
Every cell above describes evidence, not a count. The test for each cell is whether someone could satisfy it without improving the document. If they could, the cell is written wrong. The rubric does not grade how many principles the set holds, because the number is advice and not a finding.
Rows 4 and 5 grade two fields that published sets rarely carry. An honest “none yet” scores 1 on both rows, so a new set can clear the line without inventing a source or a settled case. Grade with someone who did not write the set. The blank template should score 0 on every row.
Named anti-patterns (the usual wrecks)
Section titled “Named anti-patterns (the usual wrecks)”Each entry gives the failure, the test that catches it and the rubric row it costs. The arguments behind them, and the disputes among sources, are in the companion’s sections 6 and 7.
- The slogan. A statement with no decision it would change. Test: could a designer disagree with it? Costs row 2.
- The platitude. A principle whose opposite nobody would choose. Test: Mistry’s opposite test. Costs row 2.
- Too many. The figures sources give are heuristics, and none cites a measurement. Test: can you attach a real decision to each one? Costs row 5.
- A UI direction dressed as a principle. A statement that names a component belongs to the design system. Test: does it name a component or a pattern? Costs row 11.
- Values mistaken for principles. Brignell: “Values shape culture. Principles shape decisions.” Test: what decision does it settle? Costs row 2.
- No stated order when two collide. Brignell: “If everything is equally important, nothing actually guides the decision.” Test: does the table name a winner in a named kind of decision? Costs row 6.
- Overlapping principles. Spotify’s audit named “Overlapping concepts” as a defect. Test: do two principles settle the same decision? Merge or retire one. Costs rows 5 and 9.
- A set nobody cites. The library’s inference is that a set with no stated moment of use risks being read once and not applied. Test: could someone who was not in the room write the citation line? Costs row 7.
- A standard read as a rule book. GOV.UK says “This isn’t a straitjacket or a rule book.” Test: does a decision cite the principle it leaned on and the one it set aside, with a reason? A citation so mechanical that it becomes a checkbox has lost that reading. Costs row 7.
- A set that never changes, or changes with no record. Test: does the log keep the reason for each retirement, and does each trigger name an owner? Costs rows 8, 9 and 10.
- Borrowing another team’s set. Only the NHS page read states a reuse licence, and Productboard’s prompt reserves its rights. Test: would the opposite test hold for your product? Write the team’s own words. Costs row 2.
- A set that restates the current strategy. The library’s own caution, not a sourced one. Test: would the principle survive a change of strategy while its evidence holds? Costs row 4.
- Presenting the library’s positions as received practice. Published sets carry the statement and, in three of the five tabulated, the reviewer’s question. What it rules out appears only in a vendor’s AI prompt (Productboard). What it rests on appears in no set read. A decision it settled has partial precedent in GitLab Pajamas’ “Example in practice” and GOV.UK’s linked case studies. The role names in Ownership and the table form of the Review Trigger are also the library’s. A filled set may keep them and should not describe them as a standard.
Pairing with your process
Section titled “Pairing with your process”This bundle ships in the standing-standards family as a foundation: a standard that a team’s
experience decisions are judged against. The reasoning is in
ADR 0089 (design principles joins standing-standards as a foundation).
Update the set when a Review Trigger condition fires, not on a calendar, because no source read states a
revision interval. A done criterion may say that work was reviewed against the principles, and the
principles own the content of that review.
What the research did not settle
Section titled “What the research did not settle”The companion’s section 6 lists the open disputes: conflict, ranking, generic principles, the right count and the two kinds of principles. Sections 2, 5 and 9 list what the research did not find, among it who started the practice, any tie to a delivery methodology, and any research on regulated products or a solo author. The template is untested for both of the last two.
When it is good enough
Section titled “When it is good enough”When a designer who did not write the set can take the next real choice between two good options, name the principle that decides it, say what that principle rules out, and say which principle wins if a second one pulls the other way. And when the page names who owns the set and what event reopens it.
Then delete every HTML comment. Keep the source_template and source_template_version frontmatter
fields, because they are the only record of which template and version produced the document.
The artifacts
Section titled “The artifacts”design-principles_template-lean.md · ~6,850 tokens
---title: "{{product_name}} Design Principles"doc_type: design-principlessize: leanproduct_name: "{{product_name}}"owner: "{{owner}}"status: draftdoc_version: "{{doc_version}}"created: "{{date}}"updated: "{{date}}"related_links: []source_template: design-principlessource_template_version: 0.1.0---
<!--LEAN DESIGN PRINCIPLES. This bundle ships one size, deliberately. No source the research read publishes afuller and a shorter version of one set. The only condensed forms found are a PDF GitLab links and an NHSposter, and the research opened neither. NN/g warns that "having 10 or more leads to many beingforgotten", and Brignell writes "The most effective examples use 3-5 principles." So the document isshort on purpose, and a longer variant would model the weight the research rejected. The sources thatadd detail add it per principle, not per document, and so does this template. Confidence in the singlesize is medium-high. See design-principles_companion.md section 4 (Variants and sizing).
A DESIGN PRINCIPLES DOCUMENT IS A SHORT STANDING LIST OF VALUE STATEMENTS that a team agrees once andthen uses to frame and settle decisions about its product's experience. This family reads that as astandard decisions are judged against. It is a standard applied by judgement, not by checkbox: GOV.UKsays "This isn't a straitjacket or a rule book." and Brignell writes "Unlike rules, which prescribe aspecific outcome, principles guide judgement." See design-principles_companion.md section 6 (Debates andcontested boundaries).
SOME OF THIS STRUCTURE IS THIS LIBRARY'S OWN POSITION, NOT RECEIVED PRACTICE.No source the research read publishes a template for this type, so every section here is assembled frompublished sets and method articles. Each principle below carries five fields. Published sets carry thestatement and, in three of the five tabulated, the reviewer's question. What it rules out appears only ina vendor's AI prompt (Productboard). What it rests on appears in no set read. A decision it settled haspartial precedent in GitLab Pajamas' "Example in practice" and GOV.UK's linked case studies. Where asection is the library'sposition, its comment says so. No source shows that keeping such a documentimproves outcomes. The research found no source on solo practitioners or regulated products, so thetemplate is untested for both.
WHAT THIS IS NOTIt is NOT a design system. NN/g says these principles are "not directions on how to design specific UIelements (like design patterns or standards)", so on this library's reading, component rules andpattern guidance belong to a design system. It is NOT a set of general usability heuristics, becauseNN/g separates a team's own principles from "usability guidelines, heuristics, or visual-designprinciples".It is NOT a list of what the team builds, funds or declines. The line with the sibling product-principlesbundle is this library's, not the sources', and published sets cross it. The working test, at medium-lowconfidence, is to ask what a principle settles. If it settles how an interaction, a message or a screenshould behave for its user, it belongs here. If it settles what the team builds, funds or declines, itbelongs in product principles. Some principles straddle, and this document should say so rather thanhide them. The last field of Purpose and Scope is where it says so.Do not call this type by NN/g's longer name, which begins with the word "product". That name collideswith the sibling bundle.
All names, dates and figures in the GOOD and WEAK illustrations below are invented.
HOW TO FILL THIS IN1. Read the comment under each heading: WHAT it wants, WHY it matters (with a pointer into design-principles_companion.md for the deep reasoning), guiding questions to ASK, a GOOD and a WEAK example, and the TRAP to avoid. The When Principles Conflict, Ownership and Changes and Review Trigger tables also carry PRIORITY and ROW HINT.2. Replace each {{placeholder}} with your content. Draft Purpose and Scope first, then The Principles, then When Principles Conflict, then the rest. Within The Principles, start from a decision the team has argued about, ask which option a principle would have chosen, then write what it ruled out.3. If a section, or one row inside a table, does not apply, write "N/A" and one line of why, rather than leaving it blank or deleting the section. An unexplained blank and a skipped item look identical to a later reader.4. This is a standing document, revisited on the Review Trigger below, not written once and forgotten. Before you first share it: self-grade against design-principles_guide.md, then DELETE every HTML comment. They are guidance, not content. Keep the source_template and source_template_version frontmatter fields: they are the only record of which template and version produced this document.-->
# {{product_name}} Design Principles
## Purpose and Scope
<!-- WHAT The product or service these principles cover, who applies them, how they relate to the design system, which kinds of decision they do not settle, and which principles, if any, also settle what the team builds. WHY NN/g says "Design principles are specific to the product or service that is being designed." Brignell makes the same point: "Design principles are unique to an organisation, team or project." A set with no stated scope leaves a new product with no way to know whether the set covers it, and that reading is this library's. Spool's test asks the question directly: "Have You Evaluated it for This Project?" Blackie separates industry heuristics, team principles and business or program principles, which implies that a set should say which kind it is. The relation to the design system is this library's position, because the sources split. GitLab says "Our design principles are outlined in the Pajamas Design System and guide all product design work." The U.S. Web Design System says "These design principles are one set of shared values, and the design language of the U.S. Web Design System is another." This template asks the author to state the relation and prescribes neither. The last two fields are where the line with product principles is first drawn, and that line is this library's. Published sets cross it: one of NN/g's own example principles is "Prioritize user needs over business constraints.", and USWDS's principle on earning trust asks about data privacy and continuous-integration testing. So some principles will straddle the line. The last field asks the author to name them and say why, or to write "None" with a reason, and it is the library's position. No source read asks for it. Deep dive: design-principles_companion.md section 3 (Anatomy > Purpose and Scope). ASK Which product or products does this set cover, and who applies it: designers only, or designers and product managers and engineers? Does a design system exist, and does this set sit inside it or beside it? Which kinds of decision does this set not settle? Which principles, if any, also decide what the team builds, and why? GOOD "These principles cover the Linden Transit journey planner, in its app and on its website. The designers, the two product managers and the front-end engineers apply them. They sit beside the Linden design system, which holds the components, and are linked from its front page. They do not decide which routes or features the team builds. That belongs in the product principles." For the last field: "Principle 4, Work without a signal, also settles what the team builds, because working offline means funding an offline timetable, and that is a product choice. This set settles only how the screen behaves when there is no signal." WEAK "These principles apply to everything we make." (names no product, no reader and no design system, and gives a new product no way to tell whether it is covered) For the last field: "None." (gives no reason, so a reader cannot tell whether the author tested each principle or never looked) TRAP A scope that drifts toward what the team should build or decline. That has strayed into product principles. It is a library judgement and no source states it. A close cousin is writing "None" in the last field without applying the what-does-it-settle test to each principle. -->
**Product or service covered:** {{product_covered}}
**Who applies this set:** {{set_appliers}}
**Relation to the design system:** {{design_system_relation}}
**Decisions this set does not settle:** {{out_of_scope_decisions}}
**Principles that also settle what the team builds:** {{straddling_principles}}
## The Principles
<!-- WHAT Three to five principles. Each carries five fields: a statement, what it rules out, the question a reviewer asks, what it rests on, and a decision it settled. This section is the document. Copy the block below to add a fourth or fifth principle. WHY NN/g's definition makes principles the thing that "frame design decisions", so each has to be specific enough to frame one. The three to five figure is advice, not a finding. Brignell writes "The most effective examples use 3-5 principles.", Kholmatova's slides say "not more than 3-5", and Mistry writes "Limit to 3-4 maximum." None of them cites a measurement. Each field has a source or is labelled. STATEMENT: all five sets tabulated carry one. WHAT IT RULES OUT: Productboard's AI prompt, a vendor's prompt and not a published set, has a field headed "What it rules out:", and Spool's test asks "Does it Help You Say 'No' Most of the Time?" No published set read carries the field, so asking for it is this library's position. Brignell writes "A slogan describes intent. A principle guides a choice." Mistry's opposite test is the check: "If you can't imagine another successful product using the opposite approach, you don't have a principle - you have a platitude." THE QUESTION A REVIEWER ASKS: GitLab's pages say "When designing any feature, ask:", IBM gives four to six yes-or-no questions per principle, and USWDS gives "Key considerations". WHAT IT RESTS ON: Spool's first test asks "Does It Come Directly From Research?", and the NHS states a provenance for its set. No published set read carries this field, so it is this library's addition. A DECISION IT SETTLED: GitLab gives each principle an "Example in practice", GOV.UK links case studies, and Spotify reports "The more examples, the better". Asking for one real case on every principle, and for the option it turned down, is this library's addition. The wording should be the team's own. The NHS page states a reuse licence, the Open Government Licence v3.0, and Productboard reserves its rights. The other pages read state no licence in the text read, which is not proof that none exists. Deep dive: design-principles_companion.md section 3 (Anatomy > The Principles) and section 6 (Debates and contested boundaries). ASK Which decision has the team argued about, which option did this principle choose, and which option did it turn down? What can the team no longer do once it adopts the principle, and could a designer disagree with the statement? What question would a reviewer put to a proposed design? What research, earlier decision or stated value does it come from, and which real case did it settle? GOOD Principle 1: "Show the next departure first." Statement: "Whenever a rider opens the app at a stop, the first thing on the screen is the time of the next vehicle for the stop they are most likely at." Rules out: "A home screen that opens on news, promotions or a map the rider must pan before any time appears." Reviewer question: "If the rider looks at this screen for two seconds, can they say when the next vehicle leaves?" Rests on: "The invented March rider interviews, in which riders opened the app at a stop and left when no time appeared." Decision it settled: "We turned down keeping the service-alerts banner above the departure time and moved it to a one-line note under it, after a critique showed the banner pushing the time off the first screen." WEAK Principle 1: "Delight the user." Rules out: "Bad experiences." Reviewer question: "Is it delightful?" Rests on: "Best practice." Decision it settled: "Many." (a slogan: nobody would choose the opposite, so it rules nothing out, and no field names a real decision or a real source) TRAP A principle that rules nothing out, which no source read defends: the Slogan or decision tool debate in the companion treats it as a failure. A close cousin is a UI direction in disguise: a statement that names a component belongs in the design system, as NN/g's separation shows. A last slip is copying a published principle and its wording from another team's set. A set written for another product may not pass the opposite test. -->
**Principle 1: {{principle_1_name}}**
- **Statement:** {{principle_1_statement}}- **What it rules out:** {{principle_1_rules_out}}- **The question a reviewer asks:** {{principle_1_reviewer_question}}- **What it rests on:** {{principle_1_rests_on}}- **A decision it settled:** {{principle_1_decision_settled}}
**Principle 2: {{principle_2_name}}**
- **Statement:** {{principle_2_statement}}- **What it rules out:** {{principle_2_rules_out}}- **The question a reviewer asks:** {{principle_2_reviewer_question}}- **What it rests on:** {{principle_2_rests_on}}- **A decision it settled:** {{principle_2_decision_settled}}
**Principle 3: {{principle_3_name}}**
- **Statement:** {{principle_3_statement}}- **What it rules out:** {{principle_3_rules_out}}- **The question a reviewer asks:** {{principle_3_reviewer_question}}- **What it rests on:** {{principle_3_rests_on}}- **A decision it settled:** {{principle_3_decision_settled}}
## When Principles Conflict
<!-- WHAT A table that says which principle wins when two pull apart, and in which kind of decision. WHY NN/g's method has a step titled "Identify Any Common Tradeoffs" and tells the author to "Take a stand on which value is important". Brignell writes "Real principles reveal priorities.", "Sometimes consistency conflicts with clarity." and "If everything is equally important, nothing actually guides the decision." THIS SECTION TAKES A SIDE IN A LIVE DISPUTE, AND THE SIDE IS THIS LIBRARY'S POSITION. NN/g also says a principle should "Not conflict with one another", and Blackie lists "Non-conflicting:" among his criteria. The two camps may mean different things by conflict: NN/g may mean that two principles should not contradict each other, and Brignell that a useful principle exposes a tension. No source says so, so that reading is at medium-low confidence. This template asks the author to name, for a kind of decision, which principle wins. A resolution can take several forms. Salesforce "stack-ranked" its four principles. Keith recommends the form "X, even over Y." Atlassian, as the principles.design mirror records it, writes each principle as one thing before another, as in "Trusted fundamentals before comprehensive patterns". As this research read them, GOV.UK, the NHS and USWDS state no priority order, though the NHS list is numbered. Ranking is welcome and not required. A resolution scoped to a kind of decision survives longer than a global order, and that reading is the library's. Deep dive: design-principles_companion.md section 3 (Anatomy > When Principles Conflict) and section 6 (Debates and contested boundaries). ASK Which two principles are most likely to collide? In which kind of decision do they collide? Which one wins there, and why? Is the answer different in another kind of decision? PRIORITY Resolve the most likely collision first, and do not rank every pair. One resolved conflict in a named kind of decision is worth more than a complete table nobody applies. That is this library's advice, not a sourced rule. ROW HINT A good row names the two principles, names the kind of decision, says which principle wins there and gives the reason. A weak row says both matter equally. GOOD | Show the next departure first, against Explain every delay in full | Layout of the home screen and notification text | Show the next departure first | A rider at a stop needs the time before the cause, so the explanation moves one tap away. | WEAK | All of them | Everything | Use judgement | They are all important. | (no pair, no kind of decision, no winner) TRAP Saying that all principles are equally important, or ranking every pair so the table becomes a document nobody applies. A set that never says which one wins leaves the argument to whoever is in the room. -->
| Principles in tension | Kind of decision | Which wins | Why ||---|---|---|---|| {{conflict_principles}} | {{conflict_decision_kind}} | {{conflict_winner}} | {{conflict_reason}} |
## Using Them in Decisions
<!-- WHAT Where the set is applied, such as design critique, design review and proposals for a new pattern, how a decision cites a principle, and what a reviewer may do when a decision sets one aside. WHY The sources say what a set is for before they say how to apply it. The NHS says "Use them to get started on a project and to help with making decisions." NN/g advises teams to "use the principles to explain and justify design decisions". GitLab's critique sessions tell designers to "Connect feedback to user needs and business goals instead of personal preference" and to "Review the complete design principles in Pajamas". Salesforce used its ranked principles in design review, and the practice spread to marketing and product management. Brignell names principles nobody refers to as a failure mode. This library infers that a set with no stated place where it is cited risks being read once and not applied, and no source read ties the failure to that cause. The list of places and the form a citation takes are this library's position, and no source read prescribes either. The last field rests on a reading. USWDS calls its set "an evaluative lens for design and implementation decisions" and GitLab says "Principles act as a reusable standard for teams to measure their work." The family reads judged against as judgement: a decision cites the principle it leaned on and the one it set aside, and a reviewer may accept the reasoning without accepting the outcome. Deep dive: design-principles_companion.md section 3 (Anatomy > Using Them in Decisions). ASK In which two or three moments is a principle cited? How is the citation written down? When a decision sets a principle aside, what does the reviewer do? GOOD "Applied in the fortnightly design critique, in design review before any screen ships, and when someone proposes a new component. A decision record carries one line naming the principle it leaned on and the one it set aside, with the reason. A reviewer may accept the reasoning without accepting the outcome, and says which." WEAK "Designers should keep the principles in mind." (names no moment, no form of citation and no way to tell a decision that used them from one that did not) TRAP A set that is read at adoption and never cited again, or a citation so mechanical that it becomes a checkbox. This library reads a standard judged against as judgement, not compliance. -->
**Where the set is applied:** {{where_applied}}
**How a decision cites a principle:** {{how_decisions_cite}}
**When a decision sets a principle aside:** {{when_set_aside}}
## Ownership and Changes
<!-- WHAT The named owner, how anyone proposes a change, who decides, and a dated change log. WHY This is where the set's own staleness is handled. GOV.UK credits a body, the Government Digital Service and the Central Digital and Data Office. Pajamas says "Everyone is welcome to suggest improvements by opening an issue", and the NHS offers "Feed back or share insights on GitHub". GitLab's handbook names designers as "the DRIs for design decisions", but names no owner or revision trigger for the principles themselves. No source read names a role that decides a change to a set, so the role names this template asks for are the library's position. The dated record is sourced in practice, not in method. GOV.UK's update history has three entries in thirteen years, from 3 April 2012 to 2 April 2025, and the NHS page carries "Updated: April 2023". The 2025 addition followed a cross-government working group: its authors wrote "We facilitated 2 workshops", and afterwards "we're now looking for someone to create an 11th poster, and to update the other 10", so a change had a cost beyond the sentence. Sets are replaced as well as amended. Spotify replaced a 2013 set of six with a set of three after an audit, and Spool writes that "old principles will need to go into storage". So record a retirement as a change and keep the reason. No source read states a revision interval, and sets appear to change rarely and on an event. That is at low to medium confidence, since three dated entries on one page are a thin base. Deep dive: design-principles_companion.md section 3 (Anatomy > Ownership and Changes). ASK Who owns this set, by role and by name? Where does anyone propose a change? Who decides? When was the set first agreed, and what has changed since, including anything retired? PRIORITY Start the log with the date the set was first agreed, so the first row is never blank. ROW HINT A good row has a date, says what changed in a sentence a stranger could follow, and names what fired it: a Review Trigger row, or "first agreed". A weak row is a date and the word "updated". GOOD | 2025-03-18 | Principle 4 retired and moved to the archive: an audit found it overlapped principle 2, and nobody could recall it unprompted | Review Trigger: an audit finds principles overlapping or forgotten | Design lead | WEAK | 2025-03-18 | Updated | | | TRAP A set with no owner and no condition that reopens it, which has no way to change on purpose. A change log that shows only additions hides replacement. -->
**Owner (role and person):** {{owner_role_and_person}}
**How anyone proposes a change:** {{how_to_propose}}
**Who decides a change:** {{who_decides}}
| Date | What changed | What fired it | Who decided ||---|---|---|---|| {{change_date}} | {{change_summary}} | {{change_trigger}} | {{change_decider}} |
## Review Trigger
<!-- WHAT A table of conditions, owners and actions: what event makes this set wrong or incomplete, who notices, and what they do about it. The conditions are things an owner can observe, not dates. WHY The standing-standards family requires every member to carry this mechanism, because members fail by going quietly out of date. No design principles source names a trigger, so the table form is this library's position, and each condition below is sourced where a source exists. A new cross-cutting concern with people behind it: GOV.UK added a principle in 2025 after a cross-government working group formed around the topic. An audit finds principles overlapping or forgotten: Spotify's audit named "Overlapping concepts" as a defect, and NN/g warns that "having 10 or more leads to many being forgotten". A new product or surface joins the scope: Spool's test asks "Have You Evaluated it for This Project?" A debate recurs and no principle settles it: the library's own position, with Mistry's stated goal to end repeated debates, not all debate. The audience or the strategy changes: the library's own inference, which no source read states. No calendar interval is sourced, so none is required. A team may add one as a floor if it wants one, and should label it as its own choice. Deep dive: design-principles_companion.md section 3 (Anatomy > Review Trigger). ASK What could the owner observe that shows this set is wrong? Who notices each condition, by role? What do they do about it: reopen the set, run a short audit, or write the change into the log? PRIORITY List the conditions an owner is most likely to see first, and write each as an event someone can observe, not as a date. ROW HINT A good row states a condition the owner could observe, names a role as owner, and states an action, such as reopening the set or writing a dated line in the change log. A weak row says "review regularly" and names no one. GOOD | An audit finds two principles overlapping, or one nobody can recall | Design lead | Merge or retire one, write the reason into the change log and move a retired principle to an archive section, not out of the document. | WEAK | Review regularly | The team | Update the principles. | TRAP Writing a date as the only trigger, or naming one condition and treating it as the whole mechanism. The strongest rows are the ones that have already fired once, so tie a row to a line in the change log when it does. -->
| Condition | Owner | What They Do About It ||---|---|---|| A new cross-cutting concern with people behind it | {{concern_owner}} | {{concern_action}} || An audit finds principles overlapping or forgotten | {{audit_owner}} | {{audit_action}} || A new product or surface joins the scope | {{new_scope_owner}} | {{new_scope_action}} || A debate recurs and no principle settles it | {{recurring_debate_owner}} | {{recurring_debate_action}} || The audience or the strategy changes | {{audience_change_owner}} | {{audience_change_action}} |---title: "Acme Analytics Design Principles"doc_type: design-principlessize: leanproduct_name: "Acme Analytics"owner: "Marcus Webb (Design Lead, Reporting)"status: activedoc_version: "1.1.0"created: "2025-10-14"updated: "2026-02-03"related_links: - "../user-persona/user-persona_example.md (the Recurring Analyst persona, 2026-01-05; the evidence behind principles 1, 2 and 3)" - "../product-vision/product-vision_example.md (the product vision, 2026-01-14; holds where the product is heading, a decision this set does not settle)" - "../product-strategy/product-strategy_example.md (the FY26 product strategy, 2026-01-28; the change that fired the 2026-02-03 revision)"source_template: design-principlessource_template_version: 0.1.0---
> **Worked example.** A filled `design-principles`, the one size this bundle ships, for the same Acme> Analytics product used throughout this library's examples. The `standing-standards` family contract> weakens the chaining on purpose, because a standing standard belongs to a team rather than to a moment> in a story, so this document sits beside the thread and not at one point on it. It is dated 2026-02-03,> the day of the last line in its change log, and the body below cites only documents that existed by> then: the [Recurring Analyst persona](../user-persona/user-persona_example.md) of 2026-01-05, the> [product vision](../product-vision/product-vision_example.md) of 2026-01-14 and the> [FY26 product strategy](../product-strategy/product-strategy_example.md) of 2026-01-28. It comes well> before the [Saved Views design brief](../design-brief/design-brief_example.md) of 2026-06-01, whose> design team would work under a set like this one. The sibling bundle `product-principles` settles what> a team builds, funds or declines, and this document leaves that to it.>> Read it alongside [`design-principles_guide.md`](design-principles_guide.md), the rubric it was graded> against. Three things are visible in it. A principle has been retired and archived with its reason, not> deleted. One Review Trigger row has fired and points at its line in the change log. One conflict is> resolved in a named kind of decision. Published sets carry the statement and, in three of the five> tabulated, the reviewer's question. What it rules out appears only in a vendor's AI prompt> (Productboard). What it rests on appears in no set read. A decision it settled has partial precedent> in GitLab Pajamas' "Example in practice" and GOV.UK's linked case studies. The role names in Ownership> and Changes and the table form of the Review Trigger are also this library's own position, not received> practice. A set of three is not a target: the number is advice, and this team holds three> because it could attach a real decision to each.>> All figures are illustrative. They are internally consistent with the other examples in this library> and are not drawn from a real company. The persona's interview counts are repeated exactly, because> both documents describe the same research.
# Acme Analytics Design Principles
## Purpose and Scope
**Product or service covered:** The dashboards and the report-authoring screens of the Acme Analyticsweb application, and the emails and in-product notifications that the application sends about them. Thepublic marketing site is not covered.
**Who applies this set:** Marcus Webb's Reporting product design team applies it to every screen itdesigns. The product managers who commission a change, and the Design Systems group that reviewspattern proposals, also apply it, and the front-end engineers meet it only through those two.
**Relation to the design system:** The set is a separate file from the shared design system, which theDesign Systems group maintains for components and patterns. The two are joined at one point. Thedesign system's contribution form for a new or changed pattern has a required field that asks whichprinciple the proposal serves, and the answer must name one from this set. A statement that names acomponent belongs in the design system, so none of the principles below does.
**Decisions this set does not settle:** Which problems Acme builds for and which it declines, which the[FY26 product strategy](../product-strategy/product-strategy_example.md) holds. Where the product isheading, which the [product vision](../product-vision/product-vision_example.md) holds. How a componentlooks or behaves on its own, which the design system holds. Whether a number is correct: engineeringowns that, and these principles settle only how a screen presents a number that is already right.
**Principles that also settle what the team builds:** Principle 3, Say what to try next. Whether anearer alternative exists to offer is partly a decision about what the product does, which belongs to theproduct team, and this set settles only that the screen says so. Principles 1 and 2 settle how a screenpresents a figure and a label, and neither decides what Acme builds.
## The Principles
**Principle 1: Say what a number covers**
- **Statement:** Wherever a screen shows a figure, the same screen says in words which records the figure counts, for which period and as of when. The reader should not have to open a control to find out.- **What it rules out:** A headline figure whose scope lives only in filter controls that are collapsed, or in a hover tip. A chart titled with the measure alone, such as "On-time delivery", with no region and no week beside it. A glanceable wall display could reasonably choose the opposite, because nobody carries its number into a meeting. Acme's reader does.- **The question a reviewer asks:** If someone read this figure aloud at a meeting, could a listener who never saw the filters say what it covers?- **What it rests on:** The [Recurring Analyst persona](../user-persona/user-persona_example.md) of 2026-01-05. It records that Elena Cho, a customer's regional operations manager, presented three hubs' on-time numbers in October 2025 where she meant four, and corrected herself in front of her VP. It also records how the miss comes about: the dashboard opens on the company-wide default, and she rebuilds the region, week and hub filters by hand each Monday from a personal spreadsheet. The persona does not record what the figure on her screen said about its scope. The design team's reading is that the figure itself should have told her which hubs it counted. That reading is a judgement, not a finding from the interviews.- **A decision it settled:** At the design critique of 2025-11-25 the team decided how a headline tile should show its filters. It added a caption under each figure, which read, on sample data, "All four hubs, week of 3 November, as of 06:00 Monday". It turned down a small badge reading "3 filters applied", because the badge would have told the reader that filters existed without saying what they were.
**Principle 2: Use the word the reader uses**
- **Statement:** When the product names something that a person at an operating company already has a word for, the screen uses that word first. The system's own term appears only when the person asks to see it.- **What it rules out:** A column or table name, such as "fct_shipments.delivered_flag", as the main label for what the person calls "delivered on time". A heading that explains a feature in the vocabulary of the data model. A query tool for developers could reasonably choose the opposite, because its reader wants the schema's names. Acme's reader has no query background.- **The question a reviewer asks:** Would the person say this label out loud at their own weekly review, in these words?- **What it rests on:** The same persona. It describes Elena Cho as having no dedicated analyst and no query background, and it says that a team hearing the word "analyst" builds for someone who can write a query, and that she cannot.- **A decision it settled:** At the design critique of 2025-12-09 the team reviewed the labels in the filter panel. It labelled the region control "Region" and kept the table's column name available when the person asks for it. It turned down showing both names on every label, because that doubled the text a reader had to scan before reaching the control.
**Principle 3: Say what to try next**
- **Statement:** When the product cannot give a person what they asked for, the screen says in one sentence what it could not do, and offers the nearest thing it can do from that same screen.- **What it rules out:** An empty chart with no words. A message that reads only "No results found" or quotes an error code. A dead end that sends the person to someone else before they can continue. A product sold to teams that keep an analyst on hand could reasonably choose the opposite and say "ask your analyst". Acme's reader has no dedicated analyst.- **The question a reviewer asks:** If this screen cannot give the person what they came for, can they act on what it says without leaving the screen or asking anyone?- **What it rests on:** The same persona's finding that when the dashboard cannot answer a question, the only path is to ask someone else, a data analyst, and wait. Eight of the twelve interviewees raised it, and the question often goes unasked. The interviews ran in November and December 2025.- **A decision it settled:** At the design critique of 2026-02-03, where this principle was added, the team applied it to a chart whose filters match no records. It turned down the empty chart and the sentence "No results found". It chose a sentence saying that the filters matched no records, with the previous week offered as the next thing to look at. **This principle straddles a line.** Whether a nearer alternative exists to offer is partly a decision about what the product does, which belongs to the product team. This principle settles only that the screen has to say so.
## When Principles Conflict
| Principles in tension | Kind of decision | Which wins | Why ||---|---|---|---|| Use the word the reader uses, against Say what a number covers | The label of a figure whose everyday name differs from what the product counts, such as "on time" | Say what a number covers. The reader's word may stay as the heading, with the exact count written under it | A familiar word that hides a different count lets the reader repeat the figure with confidence in the wrong meaning. A longer and less familiar phrase costs a few seconds of reading |
## Using Them in Decisions
**Where the set is applied:** At the Reporting design team's critique, which runs every two weeks andis the main place. In the Design Systems group's review of a pattern proposal, through the form fieldnamed above. And whenever someone writes or edits the wording of an email or a notification theproduct sends.
**How a decision cites a principle:** The pull request or design ticket that ships the change opens itsdescription with a line beginning "Principle:", naming the one that decided it. A second line beginning"Weighed against:" appears only when another principle pulled the other way, and it gives the reason ina sentence. An example: "Principle: Say what a number covers. Weighed against: Use the word the readeruses, because the everyday word hid the count."
**When a decision sets a principle aside:** The designer who proposes the change states the reason atthe critique, and the room replies in one of two ways. Either the reason holds and the change goesahead, or the reason is a fair argument for a worse screen and the change goes back for another pass.The ticket records which reply it was. A principle set aside twice in the same kind of decision goes toMarcus Webb, who decides whether it needs a row in When Principles Conflict or a rewrite.
## Ownership and Changes
**Owner (role and person):** The Design Lead, Reporting: Marcus Webb.
**How anyone proposes a change:** By opening a ticket labelled "principles" in the design team's queue,or by raising it at the critique. Anyone at Acme may do either.
**Who decides a change:** Marcus Webb decides. Dana Okoro, VP Product, approves any change that adds orretires a principle, because that changes what every screen is judged against.
| Date | What changed | What fired it | Who decided ||---|---|---|---|| 2025-10-14 | First agreed, with three principles: Say what a number covers, Use the word the reader uses, and Keep the expert's hands fast. The When Principles Conflict row, which resolves Use the word the reader uses against Say what a number covers, was part of that first agreement | First agreed | Marcus Webb, approved by Dana Okoro || 2026-02-03 | Keep the expert's hands fast retired to the archive below, and Say what to try next added in its place. The rests-on fields of principles 1 and 2 now cite the Recurring Analyst persona of 2026-01-05, which did not exist when they were agreed. Their wording is unchanged | Review Trigger: the audience or the strategy changes. The FY26 product strategy of 2026-01-28 names the unaccompanied analyst, the person the persona calls the Recurring Analyst, as its target segment | Marcus Webb, approved by Dana Okoro |
**Archived principle: Keep the expert's hands fast.** Agreed on 2025-10-14 and retired on 2026-02-03.Its statement read: when a person has already told the product exactly what they want, nothing standsbetween them and the result, with no confirmation step, no walkthrough and no second screen. It ruledout a confirmation dialog on a routine action and a tutorial overlay that blocks the page. It wasretired because it served the person who already knows what to ask, and the FY26 strategy names adifferent person as its target, one for whom the missing knowledge is the obstacle and the extra step isnot. It stays here with its reason, so that a later change of strategy can bring it back knowingly, andso that nobody proposes it again as though it were new.
Principles 2 and 3 rest on the persona's evidence rather than on a strategy choice. They would survive achange of strategy for as long as the persona holds. What would retire them is evidence that the personano longer describes the reader, such as research showing that most of Acme's readers now keep an analyston hand or a query background. The set is reread when the target segment changes, which is what theReview Trigger row below records.
## Review Trigger
| Condition | Owner | What They Do About It ||---|---|---|| A new cross-cutting concern with people behind it | Marcus Webb | Asks the people raising the concern which decisions they keep losing. Drafts a principle only if two options keep recurring in those decisions. Otherwise writes the concern and the date into the change log and leaves the set alone || An audit finds principles overlapping or forgotten | Marcus Webb | Asks each designer to list the principles from memory. A principle that two designers cannot recall, or that settles the same decisions as another, is put to the next critique, which rewrites it or retires it. Whichever happens is dated in the log with the reason, and a retired text moves under the log || A new product or surface joins the scope | Priya Nair (PM, Reporting) | Raises the new surface at its first critique, before a screen is designed. Marcus Webb then either names it in Purpose and Scope or writes down why the set does not cover it || A debate recurs and no principle settles it | Marcus Webb | When a trade-off comes up at a second critique, writes the choice that was made as a candidate row in When Principles Conflict, or as a candidate principle, and dates it in the change log || The audience or the strategy changes | Dana Okoro (VP Product) | Tells Marcus Webb when a strategy or a target segment changes. Marcus Webb then reads each principle against the new audience at the next critique and rewrites or retires any that favours a group no longer targeted. **This row has fired once, on 2026-02-03, and the change log above records it** |
No calendar date is a trigger. A date would only call for a read of the set on a day when nothing hadchanged.Provenance
Section titled “Provenance”The reasoning, the history and every source, in the repository:
- Companion - the long-form argument: why these sections, where the sources disagree, and what the bundle refuses to claim
- History - what changed in this bundle, and when
- Research log - every source consulted, with what each one actually supports
- Catalog metadata - the machine-readable record this page is generated from
Catalog record: 6 sections across 1 format(s), methodology generic, typically owned by Design lead.