Marcus Andrey Study
Ler em português Library

Framework · Design systems · 3 min read

Pattern tiers

Four tiers for design-system patterns (core, recommended, domain-specific, experimental) that make it explicit where teams must be consistent and where they are free to differ.

  1. Experimental
    Rule
    Labelled, time-boxed and measured.
    For example
    AI-suggested field values in one app
  2. Domain-specific
    Rule
    Owned by the domain; may graduate to core later.
    For example
    A lookup only one domain needs
  3. Recommended
    Rule
    Use unless you have a documented reason.
    For example
    Filtering, bulk edit, empty states
  4. Core
    Rule
    Must use, consistent everywhere.
    For example
    Buttons, inputs, tables, navigation shell

Label every pattern with its tier, so nobody has to ask.

Four pattern tiers; each step up asks teams for more consistency.

Use it when

  • Domain teams ask "do we have to use this?" and the honest answer is "it depends", but nobody has written down on what.
  • You are setting up a contribution process and need to decide what a new pattern becomes when it is accepted.
  • A team built a good pattern for its own domain, and other teams start to want it too.

How to use it

  1. List the patterns you have. Collect components and patterns from the library and from the domains, including the ones that live only in one team's files. Output: one inventory list.
  2. Put each pattern in a tier. Core (buttons, inputs, tables, navigation shell) must be used and consistent everywhere. Recommended (filtering, bulk edit, empty states) is used unless there is a documented reason. Domain-specific is owned by the domain. Experimental is labelled, time-boxed and measured. Output: a tier next to every pattern.
  3. Write the rule for each tier. Say what "must use" and "documented reason" mean in practice, who can approve an exception, and where exceptions are recorded. Output: a short rules page in the design-system docs.
  4. Label tiers where people work. Show the tier on each component's documentation page and in the library, so designers see it without asking. Output: a visible tier label on every pattern.
  5. Review graduations regularly. In council reviews, check whether domain-specific patterns are now needed by other domains (they may graduate to core) and whether experiments have ended. Output: a list of patterns moved up, moved down or retired.

Worked example

An internal operations platform with 6 product teams.

Pattern Tier Why
Buttons, inputs, data table, navigation shell Core Operators switch between apps all day; differences cost time and cause errors.
Filtering, bulk edit, empty states Recommended Most teams need them the same way; one team documents why its filter differs.
Customs-code lookup Domain-specific Only one domain needs it. Owned by that domain's designer.
AI-suggested field values Experimental Labelled as experimental, tested in one app for one quarter, measured on errors.

After two quarters, a second domain needed the customs-code lookup pattern. The council moved it toward the recommended tier.

Watch out for

  • Everything is core. If every pattern is mandatory, teams stop asking and start working around the system. Keep core small.
  • "Documented reason" with no place to document. If there is no exception log, "recommended" quietly becomes "optional". Name where reasons are written.
  • Experiments that never end. An experimental pattern without a time box becomes permanent by accident. Set an end date and a measure.
  • Hidden tiers. If the tier lives only in the lead's head, the question "do we have to?" keeps coming back. Label every pattern.

Template

Copy this into your design-system docs as the tier rules page.

Pattern tiers

Core            Must use, consistent everywhere.
                Exceptions: none without council approval.
                Patterns:

Recommended     Use unless you have a documented reason.
                Record reasons in:
                Patterns:

Domain-specific Owned by the domain. May graduate to core.
                Owner per pattern:
                Patterns:

Experimental    Labelled, time-boxed, measured.
                End date and measure per pattern:
                Patterns:

Next tier review:

"See on Amazon" links are affiliate links. As an Amazon Associate I earn from qualifying purchases, at no extra cost to you.

Saved only on this device.

↑ ↓ to move · Enter to open · Esc to close