1. LandOffer
  2. Blog
  3. All articles
  4. Design Systems Portfolio: Make Component States and Adoption Decisions Reviewable

Job search

Design Systems Portfolio: Make Component States and Adoption Decisions Reviewable

An original local five-story ReviewButton and a completed fictional adoption decision make component behavior, evidence limits and pilot scope reviewable.

9 min readLandOffer team

A design systems portfolio becomes easier to review when a reader can inspect a component's states, try its behavior and understand where you would adopt it. A polished gallery alone leaves important questions unanswered: what happens on a keyboard, when the action is unavailable, or when a consuming team needs something different?

Show a small implementation with explicit boundaries, then a completed adoption decision tied to that implementation. This article includes an original local Storybook demonstration and a fictional adoption record, checked on October 23, 2026. LandOffer publishes this guide; browse recently posted roles to identify which component, accessibility and collaboration responsibilities your target employers describe.

Wooden architectural doorway models and drafting materials arranged to compare several configurations.

Give the reviewer a component they can question

Choose a component small enough to explain completely. A text-labelled action button is a useful starting point because it exposes decisions about semantics, focus, unavailable actions, labels and layout. A catalogue of components does not explain how any individual control should behave. A narrow example can show the reasoning that a larger catalogue hides.

Storybook describes a story as a rendered component state, with multiple stories covering interesting supported states. That supports a portfolio organized around reviewable examples rather than a single attractive screenshot. Its Component Story Format uses module exports to describe examples; the demonstration here uses explicit render functions for its fixtures.

The accompanying local Storybook build (worked-assets/storybook/index.html, a local artifact) contains five stories: Default, Disabled, Loading, LongLabel and StateGallery. The last combines four examples so the differences can be inspected together. It is an original local demonstration using Storybook 10.6.3 and Vite 7.3.1, not an employer's library or a published package.

Its fictional ReviewButton changes a local status message when enabled. It sends no request and stores nothing. The loading example is a static fixture, not a running save operation. Those details belong beside the demo: otherwise a reviewer may assume a successful click means that validation, persistence and asynchronous recovery also work.

Make the intended action visible in the label. The W3C button pattern distinguishes an action from a navigation link and describes Enter and Space activation. This example uses a native HTML button with type button. It does not convert a styled div into a button or add a navigation destination to an action control.

Worked component review: five stories with bounded results

The source and review record are included with this article. The results below come from the October 23, 2026 local browser review recorded by the production team; they are not tests on an employer application. The following completed state grid separates what the fixture contains from what the local browser review established. Focus is an interaction state inspected in the gallery; it is not a sixth exported story.

Story or state Implementation in this example Completed local review and limit
Default Text-labelled enabled native button Enter changed the local status; after reload, Space also changed it. No save or network request occurred.
Keyboard focus Visible focus outline on the enabled control Tab exposed a blue outline. This was a visual keyboard check, not a screen-reader assessment.
Disabled Native disabled button A pointer click did not change status. Tab skipped this control in the gallery.
Loading Disabled button with a busy attribute and explicit static-state description Tab skipped it; the story states that no request is running. Asynchronous announcements and recovery were not exercised.
LongLabel Wrapping label constrained by available width At a 390-pixel mobile viewport the button measured 309 pixels wide and 76 pixels high; the document did not overflow.
StateGallery Four examples displayed together The manager exposed all five stories; the gallery allowed comparison without switching between separate routes.

This grid lets a reviewer choose a state, inspect its code and compare the observed result with the intended behavior. The long-label result says something specific about this fixture at one viewport. It does not establish support for every language, text size, browser or embedding layout. Likewise, skipping disabled controls is the observed consequence of this example's native disabled implementation; another unavailable-action design needs its own semantics and review.

The completed build and manual-review receipt (evidence/root-storybook-build-and-review.json, a local artifact) records the successful build, versions, warnings, source hashes and scope. Storybook's manager, focus and mobile screenshots are retained as evidence. An enabled activation and a non-activation both matter: a component's contract includes the actions it should prevent as well as those it should allow.

The untested work includes real submission, asynchronous completion and recovery, screen-reader behavior and automated accessibility checks. There is no real submit flow, error recovery, screen-reader review, automated accessibility scan or full conformance assessment. A reviewer can ask for those next without first discovering that the attractive “saving” state never performed a request. Honest scope makes the next discussion more useful.

Connect tokens and source to the behavior

A token list becomes more useful when its names lead to visible decisions. This fixture's CSS names the action background and text, hover background, focus outline, spacing and control radius. The same file defines the focus-visible outline and the unavailable appearance. A reader can follow a state from the story into the styling instead of accepting a disconnected palette slide.

The names are local conventions for this example. They do not prove support for a standard token interchange format, a complete theme architecture, dark mode or downstream package compatibility. Describe those capabilities only when your implementation and evidence support them. A local variable named for a role is a useful decision; it is not a full distribution strategy.

The story source (worked-assets/storybook-source/stories/ReviewButton.stories.js, a local artifact) and token file (worked-assets/storybook-source/stories/tokens.css, a local artifact) are included. The story creates the button, applies the disabled or static loading condition, and attaches a local click handler. The status text explicitly says nothing was sent or persisted. Keeping this code close to the state record makes the portfolio easier to inspect.

For reproduction, the included source folder has package.json, its lockfile and the Storybook configuration. Use an isolated copy, install its locked dependencies with npm ci, then run npm run build after reviewing the package's output path. The recorded build ran in a separate local runtime and succeeded with a Node deprecation warning and a large-chunk advisory. Do not describe a clean production pipeline or a reader's successful installation from that one result.

Controls and automated checks are separate features. These stories use fixed render functions; no interactive args or Controls configuration is demonstrated. State which controls the reviewer can change in your demo. Keep the wider Storybook feature list separate.

A small footbridge model beside separate material samples and an architect's decision notebook.

Worked adoption decision: pilot two controls and leave three unresolved

The second asset is a completed fictional adoption decision, DS-01 (worked-assets/adoption-decision.md, a local artifact). It connects the demonstrated component to a bounded integration choice. All tools, people and counts are invented teaching premises, not observed company adoption or user research. The decision deliberately stops short of claiming that the component solved an organization's design problems.

The stipulated problem is five save controls with inconsistent labels, spacing and unavailable states across two fictional tools. The selected scope is Workbench's two review-form save controls. Roster's three candidates remain untouched. Here is the completed decision record in a form a reviewer can challenge.

Decision field Completed DS-01 disposition
Problem boundary Five stipulated save controls vary across two fictional tools; no measured user confusion or financial loss is claimed.
Alternatives considered Retain independent controls; pilot a shared control in one flow; replace all actions immediately.
Selected option Pilot the shared ReviewButton in Workbench's two review-form save controls after review evidence exists.
Excluded uses Navigation, icon-only actions, menus, destructive confirmation and real submission or persistence.
Fictional ownership Jo reviews behavior and regression evidence; Sam reviews labels and visual states; the consuming-flow owner approves integration.
Acceptance work Inspect stories, visible focus, Enter/Space, unavailable actions, narrow labels and contextual meaning; review contrast and behavior in the consuming flow.
Version and reversal Keep demonstration version 0.1.0; a consuming flow may return to its previous version if integration changes behavior. No production rollback was exercised.
Completed disposition Proceed with a bounded fictional pilot, retain the other three candidates as unresolved, and report evidence rather than invent a successful rollout.

Two of five is a planned scope count. It is not measured 40% adoption. Adoption would need an actual definition, consuming inventory, observation date and evidence that the controls were integrated and used. A portfolio can explain why two candidates were selected without pretending the plan has already happened.

The excluded uses show why the pilot is small. A navigation link has a different function from an action button. A destructive confirmation introduces consequences this fixture does not model. Menus and icon-only actions add interaction and naming questions. Replacing all of them with one attractive shape would erase the distinctions the portfolio is meant to explain.

DS-01 also separates ownership from endorsement. Naming fictional reviewers explains the handoff; it does not imply real colleagues approved the work. For a genuine project, state your own contribution, who could accept the integration, and which decision remained outside your authority. Remove private customer information and internal documents before sharing anything publicly.

Report accessibility evidence at the level you checked

A keyboard check is meaningful evidence, but its scope must remain visible. Here, Enter and Space activation, a visible focus outline, disabled non-activation and narrow-width wrapping were manually reviewed in the local fixture. Those results support specific statements about the reviewed build. They do not justify a blanket “accessible design system” label.

Storybook's accessibility documentation describes automated rendered-DOM checks against heuristics and provides an accessibility addon. This example did not install or run that addon or Vitest accessibility checks. Do not turn documented availability into a test result for your own portfolio.

If you add automated checks later, record the tool, version, rules, context, results and incomplete items. Keep manual review of meaning and interaction alongside them. A clean automated result cannot answer every question about a full application, and a component reviewed in isolation still needs evaluation in its consuming flow.

Show an unresolved issue as a precise boundary rather than an apology. “Static loading only; request completion and announcement not tested” tells a reviewer where the work ends. “Fully accessible” makes a much larger claim without explaining how it was established. The narrower statement is easier to verify and lets a reviewer ask about a specific next check.

Assemble a short path through the evidence

Put the reader's decision first: what the component does, which states are supported, and why the adoption scope is limited. Then provide the state gallery, source, observed results and DS-01. A reviewer should not need to search through an unrelated visual archive to find the behavior behind a component thumbnail.

Keep one clear route from the case-study page to the demo and back to the decision. Mark local artifacts as local examples; the accompanying build has not been deployed as a public portfolio. Before using a public link in an application, verify its actual access and permissions rather than assuming your local preview will work for a recruiter.

Choose one component in your portfolio and complete its state grid and adoption decision before adding another gallery page. Then compare recent roles on LandOffer.ai and attach the artifact that demonstrates a responsibility named in three relevant listings.

Download the worked example files

Download the companion example files. The archive contains the original story source, configuration, exact dependency lock and adoption memo discussed above, with portable rebuild instructions. Precompiled Storybook and third-party font assets are excluded. Review its README and source before running code; fictional inputs and local test results retain the limits stated in this article.

Sources and Further Reading