← Back to site~120 pages · Print → Save as PDF · Margins: None · Background graphics on
Christoph HellmuthDesign Engineering Studio
Full edition · 100 working days · 120+ pages
Zero to Design System in 100 Days
The complete day-by-day playbook: one page per working day from UI chaos to an adopted system with tokens, components, Figma↔code parity, pilots, connected delivery, patterns, docs, and governance.
Christoph HellmuthDesign engineer · 16+ years · enterprise design systemschristophhellmuth.com
How to use this book
One day, one page, one deliverable.
This is the full edition. The 25-page executive guide compresses the same arc; this book expands every working day into a focused playbook page.
Who it's for: product companies with ~20-200 people, a real web product, a small design team, and engineers rebuilding the same UI patterns. You have leadership buy-in, or you are making the case.
How to read it: work in order. Phase 1 (days 1-14) is non-negotiable. Phase 2 is the build. Phase 3 makes the system real in production. Phase 4 is adoption. Skip a phase and you pay the inconsistency tax again within two quarters.
Pair with the audit
Run the 40-Point Design System Audit on day 1. Your reds become the backlog. This book tells you the order to fix them.
Time commitment: 0.5-1 FTE on the system plus ~2-4 hours/week from a sponsor. 100 working days is about 20 calendar weeks.
How to use this bookPage
Before day 1
Get these in place or wait.
The system won't survive politics. Sort sponsorship and access before you tokenize a single color.
1. Executive sponsor. One leader who will defend system time against feature pressure.
2. System owner (named). Even 20% of one person's role. "Everyone owns it" means no one owns it.
3. Design + engineering pair with authority to change the shared library and merge to main.
5. Scope boundary: one pilot surface, hitched to funded work.
Day 0 checklist
Sponsor confirmed · Owner named · Access granted · Pilot candidates listed · Kickoff scheduled
Your day 0 notes
Before day 1Page
Edition notes
Full edition · studio-authored
Prose is written by Christoph Hellmuth / Design Engineering Studio. Frameworks are attributed; publisher text is not reproduced.
What's inside: Pilot Scorecard and Measuring Spoon extraction, connected npm delivery, six-area coverage, Hot Potato collaboration, canon vs expanded governance, docs as a store, AI guardrails, and day-by-day deliverables.
Framework credits: Pilot Scorecard, Hot Potato, Measuring Spoon (Dan Mall, Design That Scales); six-area model and token pipeline (Sarrah Vesselov, Building Design Systems, 2nd ed.). See Appendix I.
121 pages: front matter, four parts, 100 daily playbooks, appendices. Short edition: 25-page executive guide.
· How to use this book 1
· Before day 1 2
· Credits and edition 3
I Part I · Align (days 1-14) 6
ContentsPage
Contents (continued)
II Part II · Foundations (days 15-49) 21
III Part III · Connect (days 50-77) 57
IV Part IV · Adopt (days 78-100) 86
Daily index by week
W01 Days 1-5 7
W02 Days 6-10 12
W03 Days 11-15 17
W04 Days 16-20 23
W05 Days 21-25 28
W06 Days 26-30 33
W07 Days 31-35 38
W08 Days 36-40 43
W09 Days 41-45 48
W10 Days 46-50 53
W11 Days 51-55 59
W12 Days 56-60 64
W13 Days 61-65 69
W14 Days 66-70 74
W15 Days 71-75 79
W16 Days 76-80 84
W17 Days 81-85 90
W18 Days 86-90 95
W19 Days 91-95 100
W20 Days 96-100 105
Appendices A-L: reference worksheets and frameworks · from page 110
ContentsPage
Part I
Align · Days 1-14
Audit, inventory, ROI, and stakeholder alignment before you build.
Charter and baseline
40-point audit
Figma and code inventory
Stakeholder interviews
ROI and Pilot Scorecard
Executive readout
Part I · AlignPage
Phase 1 · Align · Day 1Deliverable: Signed charter: goal, scope, roles, rituals, readout date
Write the charter before you touch a file
A design system without a charter is a component library with no mandate.
Today you name the goal, the pilot surface, and who owns what. Scope stays narrow: one product area, one quarter, one definition of done. Everyone signs the same page before audit work starts.
The charter lists roles (system owner, design lead, eng lead), weekly rituals, and what you will not build in Phase 1. That boundary stops scope creep before it starts.
Schedule the day 14 readout now. Put design and engineering leads in the room. A fixed date turns alignment from a wish into a deadline.
From the field
At Nord, Aurora was used by dozens of teams and millions of users. The first week was not components. It was a one-page charter that said which surfaces were in, which squads owned adoption, and when leadership would see proof. That page survived three reorgs because the goal was written down.
Today's output
Signed charter: goal, scope, roles, rituals, readout date
Day 1 · charter-kickoffPage
Phase 1 · Align · Day 2Deliverable: Access checklist complete, baseline screenshot gallery for pilot surface
Get access and capture the before state
You cannot measure drift until you can see every layer.
Request Figma edit access, repo read access, Storybook or staging URLs, and analytics for the pilot surface. List blockers in a shared doc and assign an owner to each one.
Screenshot the pilot surface at desktop and mobile widths. Capture light and dark if both exist. Label each shot with route, squad, and date. This is your baseline gallery.
Count visible patterns on the pilot: buttons, inputs, cards, nav variants. A rough number today beats a perfect inventory next week. You need a starting line.
Today's output
Access checklist complete, baseline screenshot gallery for pilot surface
Day 2 · access-baselinePage
Phase 1 · Align · Day 3Deliverable: Audit scores for token and component lenses with variant counts
Score tokens and components against reality
Most teams fail on foundations before they fail on governance.
Run the token lens of your audit: primitives, semantics, naming, theme coverage, export path. Note every place hex or raw spacing values appear inside component files.
Run the component lens next: variant count, state coverage, Figma-to-code parity, accessibility basics. Count button and input variants in Figma and in the repo. Write both numbers down.
Mark each item red, yellow, or green. Reds become Phase 2 backlog items. Do not debate fixes today. Measure first.
Today's output
Audit scores for token and component lenses with variant counts
Day 3 · audit-tokens-componentsPage
Phase 1 · Align · Day 4Deliverable: Audit scores for documentation and governance lenses
Audit documentation and governance gaps
A system nobody can find or change is already dead.
Score documentation: usage guidance, contribution rules, changelog, onboarding path for new designers and engineers. Can a new hire ship a screen using only published docs? Note where they would get stuck.
Score governance: ownership, review process, deprecation policy, version cadence. If three squads can merge UI changes with no review, say so plainly.
Combine docs and governance reds with yesterday's token and component reds. You are building one honest picture, not four separate reports.
Today's output
Audit scores for documentation and governance lenses
Day 4 · audit-docs-governancePage
Phase 1 · Align · Day 5Deliverable: Full audit scorecard with top 10 red items ranked by impact
Roll up the full audit score
One number per lens beats forty vague concerns.
Finish any remaining audit lenses: accessibility, tooling, adoption signals, AI guardrails if relevant. Score each item consistently using the same red-yellow-green scale.
Calculate a score per lens and an overall percentage. Compare against your target for day 14: leadership should see where you stand, not where you wish you stood.
Pull the top ten reds into a single list. Group them by theme: tokens, components, docs, governance, tooling. This list feeds the readout.
Today's output
Full audit scorecard with top 10 red items ranked by impact
Day 5 · audit-scoringPage
Phase 1 · Align · Day 6Deliverable: Figma inventory spreadsheet with variants, owners, and duplicate flags
Inventory every component in Figma
Figma is where design intent lives until code proves otherwise.
Open each team library and local file tied to the pilot surface. List every component and variant set. Record file path, component name, variant count, and last editor.
Flag duplicates: two buttons, three cards, five input styles. Note which squad created each and whether it uses variables or hardcoded values.
Add a drift column for items you already know differ from code. You will validate that column on day 7. Today, capture everything in Figma.
Today's output
Figma inventory spreadsheet with variants, owners, and duplicate flags
Day 6 · figma-inventoryPage
Phase 1 · Align · Day 7Deliverable: Code inventory spreadsheet merged with Figma rows and parity status
Inventory every component in the repo
The repo is the truth users actually get.
Search the codebase for shared UI: package folders, design-system directories, Storybook stories. List each component, its path, props API, and which product routes import it.
Match each code component to its Figma row. Mark parity: match, partial, missing in Figma, missing in code. Partial is the dangerous middle. Count how many fall there.
Note test coverage and accessibility lint status per component. A Button with no focus styles in code is a red item even if Figma looks fine.
From the field
On a Flyeralarm audit, we found eleven button implementations across two repos. Only three matched Figma. That single row in the spreadsheet convinced engineering leadership to fund consolidation. Inventory beats opinion every time.
Today's output
Code inventory spreadsheet merged with Figma rows and parity status
Your notes
Day 7 · code-inventoryPage
Phase 1 · Align · Day 8Deliverable: Designer interview notes with tagged quotes and linked examples
Interview designers on daily friction
Design pain shows up in rework hours, not slide decks.
Run three to five thirty-minute calls with designers who ship on the pilot surface. Ask what they duplicate, what they distrust in libraries, and where handoff breaks.
Use their words in your notes. "We rebuild buttons every sprint" is more useful than "alignment is hard." Capture specific screens and file links they mention.
Tag each quote to an audit theme: tokens, components, docs, governance. You are collecting evidence, not running a satisfaction survey.
Today's output
Designer interview notes with tagged quotes and linked examples
Day 8 · designer-interviewsPage
Phase 1 · Align · Day 9Deliverable: Eng and PM interview notes with adoption blockers listed
Interview engineering and product leads
If eng will not ship from the system, the system is optional.
Talk to two engineering leads and one PM who own the pilot surface. Ask what they refactor repeatedly, how they discover UI patterns, and what would make adoption realistic.
Press on release pressure: what would make them reach for a shared package versus copying a folder. Note CI, review, and versioning concerns in their language.
Compare eng and design answers. Mismatches (designers think the library is fine, engineers bypass it) are high-value reds for the readout.
Today's output
Eng and PM interview notes with adoption blockers listed
Day 9 · eng-pm-interviewsPage
Phase 1 · Align · Day 10Deliverable: One-page ROI estimate: rework hours, onboarding drag, user-facing costs
Put a number on the cost of inconsistency
Leaders fund problems they can see in hours or revenue.
Estimate rework hours per month: duplicate components built, design QA cycles, bug fixes from visual drift. Use squad estimates from interviews. A conservative number you can defend beats an inflated one you cannot.
Add onboarding drag: how long until a new designer or engineer ships their first consistent screen. Multiply by hiring plans if you have them.
Frame user-facing cost where you can: support tickets tied to confusing UI, conversion impact on inconsistent flows. One Lighthouse project moved 34 to 77 after system work. Performance is part of ROI.
Today's output
One-page ROI estimate: rework hours, onboarding drag, user-facing costs
Day 10 · roi-exercisePage
Phase 1 · Align · Day 11Deliverable: Pilot scorecard with 3-5 candidates scored and primary pick drafted
Score candidate pilot surfaces
The right pilot is visible, bounded, and willing.
List three to five candidate surfaces: onboarding, settings, dashboard, checkout, admin. Score each on traffic, squad appetite, technical debt, and leadership visibility.
Weight willingness highest. A medium-traffic surface with a squad that wants change beats a hero page owned by a frozen team.
Pick a primary pilot and one backup. Write why each scored where it did. You will sort and confirm tomorrow.
Today's output
Pilot scorecard with 3-5 candidates scored and primary pick drafted
Day 11 · pilot-scorecardPage
Phase 1 · Align · Day 12Deliverable: Prioritized roadmap sorted by impact with named owners per workstream
Sort the roadmap by pilot impact
Priority is what you ship first on the pilot, not what sounds impressive.
Take the top ten audit reds and map each to a phase: tokens, components, docs, adoption. Assign every item to the pilot surface you selected yesterday.
Sort high to low by user impact and build effort. Quick wins (token export, one Button) sit beside larger fixes (form control family). Sequence so week 3 delivers visible change.
Name an owner per workstream: tokens, Figma, code, docs. Unowned items do not survive the readout.
Today's output
Prioritized roadmap sorted by impact with named owners per workstream
Day 12 · scorecard-sortPage
Phase 1 · Align · Day 13Deliverable: One-page executive summary and readout deck (10 slides max)
Prepare the executive readout
Leaders need three decisions, not forty slides.
Write a one-page executive summary: current audit score, top three priorities, pilot surface, ROI estimate, Phase 2 kickoff date. Lead with the pilot and the cost of waiting.
Build a ten-slide deck maximum: baseline screenshots, audit highlights, inventory duplicates, interview quotes, roadmap, ask. Every slide should support a decision.
Pre-wire contentious points with design and eng leads in a thirty-minute dry run. Fix wording where facts are wrong. Do not rehearse spin.
Today's output
One-page executive summary and readout deck (10 slides max)
Day 13 · readout-prepPage
Phase 1 · Align · Day 14Deliverable: Logged decisions, named owner, Phase 2 kickoff date on calendar
Run the live readout and lock decisions
Alignment is a meeting outcome, not a feeling.
Present audit results, ROI, and pilot recommendation to design and engineering leadership. Stay under forty-five minutes including discussion. Show baseline screenshots and duplicate counts live.
Get three decisions logged: approved pilot surface, named system owner, Phase 2 start date. If leadership debates whether you need a system, you are not done. Facts from fourteen days should close that door.
Send written notes within twenty-four hours: decisions, owners, next milestones. Phase 1 ends when the room agrees on the pilot and the calendar.
From the field
The best readout I ran ended with an eng director saying "I did not know we had four card components." That single inventory slide unlocked budget. Show the mess calmly. Then show the plan.
Today's output
Logged decisions, named owner, Phase 2 kickoff date on calendar
Your notes
Day 14 · live-readoutPage
Part II
Foundations · Days 15-49
Tokens, Tier 1-2 components, patterns, and library v0.1.
Token architecture
Button through Tabs
Hot Potato workshop
Principles and patterns
Midpoint readout
Publish v0.1
Part II · FoundationsPage
Phase 2 · Foundations · Day 15Deliverable: Token naming convention doc with examples and primitive vs semantic rules
Define the token naming convention
Names are the API your whole org will type for years.
Adopt a consistent pattern: category, role, optional variant. Examples: color-text-primary, space-inline-md, radius-control. Write the rules in a table with good and bad examples.
Ban ambiguous names like blue-500 in component code. Primitives may use scale numbers. Semantics must describe intent. Document the boundary in one page.
Share the convention in Slack or Teams and give squads forty-eight hours to comment. Two rounds of renames now beat six months of aliases later.
Today's output
Token naming convention doc with examples and primitive vs semantic rules
Day 15 · token-namingPage
Phase 2 · Foundations · Day 16Deliverable: Primitive color tokens in Figma and JSON with merged scales
Build the primitive color palette
Primitives hold raw values. Nothing else references hex directly.
Define neutral, brand, and semantic-hue scales: typically 9-11 steps per hue. Start with what the pilot surface already uses. Merge near-duplicates ruthlessly.
Record each primitive in Figma variables and a JSON stub in the repo. Same names in both places. A drift check on day 19 depends on this match.
Cap the palette at what you need for Phase 2. You can add steps when a real component demands it. Forty grays is not a system, it is a graveyard.
Today's output
Primitive color tokens in Figma and JSON with merged scales
Day 16 · primitive-palettePage
Phase 2 · Foundations · Day 17Deliverable: Semantic color alias table with light/dark maps and contrast checks
Map semantic color aliases
Semantics carry intent. Components never touch primitives.
Create semantic tokens that point to primitives: color-action, color-surface, color-text-primary, color-border-subtle. Define light and dark mappings in the same table.
Verify contrast pairs now: color-text-on-action against color-action must hit 4.5:1 for body text. Fix the palette before Button, not after.
Document usage in a two-column table: token name, when to use, when not to use. Designers and engineers should cite the same names in reviews.
Today's output
Semantic color alias table with light/dark maps and contrast checks
Day 17 · semantic-aliasesPage
Phase 2 · Foundations · Day 18Deliverable: Spacing, radius, and typography token sets in Figma and JSON
Define spacing, radius, and type scales
Rhythm and type do more visual work than most hero illustrations.
Set a spacing scale tied to a base unit (4px or 8px). Name inline and stack tokens: space-inline-sm, space-stack-lg. Map each to rem values in code.
Define radius tokens for controls, cards, and overlays. Define elevation shadows if the pilot uses them. Same names in Figma and CSS.
Lock type styles: font families, size steps, line heights, weights. Name them font-size-body, font-size-heading-sm, and so on. Pilot text styles should reference these only.
Today's output
Spacing, radius, and typography token sets in Figma and JSON
Day 18 · spacing-radius-typePage
Phase 2 · Foundations · Day 19Deliverable: Published Figma variables and working token export to CSS/Tailwind stub
Publish Figma variables and wire export
One export path feeds design and code from the same source.
Publish Figma variable collections for color, spacing, radius, and type. Apply them to pilot text styles and swatches. Designers should not pick hex from a legacy panel.
Set up export: Figma variables to JSON via Tokens Studio or native export, then Style Dictionary or a thin script to CSS custom properties and Tailwind theme extension.
Run the pipeline once end to end and commit the output. CI can wait until day 50. Today, prove the path works on your machine.
Token export path: one source, two consumers.
Today's output
Published Figma variables and working token export to CSS/Tailwind stub
Day 19 · figma-variables-exportPage
Phase 2 · Foundations · Day 20Deliverable: Button component in Figma and code with Storybook stories and usage note
Ship Button in Figma and code
Button is your first vertical slice. Treat it as the template.
Build Button variants in Figma on semantic tokens: primary, secondary, ghost, danger. Include default, hover, focus, active, disabled, loading. Wire all states to variables.
Implement the same API in React (or your stack): props for variant, size, disabled, loading. Match Figma spacing and type within one pixel. Add a Storybook story per state.
Write a short usage note: when to use each variant, minimum tap target, focus ring requirement. Button sets the definition of done for every component after it.
Today's output
Button component in Figma and code with Storybook stories and usage note
Day 20 · buttonPage
Phase 2 · Foundations · Day 21Deliverable: Link and Icon in Figma and code with shared focus and size scales
Ship Link and Icon components
Navigation primitives expose whether tokens and states actually work.
Build Link with inline and standalone modes, external flag, and visited styles on semantic colors. Match underline offset and focus ring to Button focus language.
Build Icon with a constrained size scale (16, 20, 24) and name mapping to your SVG set. Icons inherit color from parent text unless overridden.
Add Storybook docs showing Link inside copy, Icon beside label, and both together in a compact row. Check keyboard tab order on a real pilot route.
From the field
At the Into Design Systems AI Hackathon 2026, our team placed first with a plugin that caught swapped components before handoff. Link and Icon are where those swaps hide. Build them tight and parity checks get easier.
Today's output
Link and Icon in Figma and code with shared focus and size scales
Your notes
Day 21 · link-iconPage
Phase 2 · Foundations · Day 22Deliverable: Typography utility component and matching Figma text styles
Ship typography utilities
Text styles are components without a border.
Create Text or Typography primitives that map to your type tokens: body, caption, label, heading levels. Support weight and color props that reference semantics only.
Mirror the same styles as Figma text styles designers already use. Name them identically. A designer and engineer should say font-size-body and mean the same thing.
Document truncation, max width, and line clamp patterns if the pilot needs them. Add one Storybook page that shows every text style on color-surface and color-surface-raised.
Today's output
Typography utility component and matching Figma text styles
Day 22 · typography-utilitiesPage
Phase 2 · Foundations · Day 23Deliverable: Hot Potato session notes with parity fixes and one screen matched both sides
Run a Hot Potato workshop
Handoff breaks at the boundary. Hot Potato fixes the boundary in one room.
Bring one designer and one engineer. Designer builds Button and Link in Figma for a real pilot screen. Engineer implements the same screen in code. Swap every twenty minutes.
Log every mismatch as you go: spacing, token name, missing state, wrong prop. Fix naming and docs on the spot where you can. Escalate reds to the backlog.
End with one merged screen in Figma and code side by side. Schedule the next potato for Input after day 24. Repeat until boring.
Today's output
Hot Potato session notes with parity fixes and one screen matched both sides
Day 23 · hot-potatoPage
Phase 2 · Foundations · Day 24Deliverable: Input component in Figma and code with accessibility states verified
Ship Input with full state coverage
Forms expose every weak token and missing focus style you skipped.
Build Input in Figma: default, focus, error, disabled, read-only. Bind border, fill, and text colors to semantic tokens. Include label and helper text slots in the frame.
Implement Input in code with the same states, keyboard focus ring, and aria-invalid for errors. Wire to the same spacing and radius tokens as Button.
Add Storybook examples for login and search patterns from the pilot. Test with screen reader and keyboard only before you call it done.
Today's output
Input component in Figma and code with accessibility states verified
Day 24 · inputPage
Phase 2 · Foundations · Day 25Deliverable: Textarea and Label in Figma and code plus form field pattern doc
Ship Textarea and Label
A form field is a system, not a box.
Build Textarea sharing Input tokens and focus language. Add resize behavior and character count if the pilot needs it. Keep error and disabled states identical to Input.
Build Label as a standalone component: required marker, optional hint, htmlFor wiring. Label should work with Input and Textarea without custom CSS in product routes.
Publish a form field pattern doc: Label + Input, Label + Textarea, error text below, helper text rules. You now have a complete field stack for the pilot.
Today's output
Textarea and Label in Figma and code plus form field pattern doc
Day 25 · textarea-labelPage
Phase 2 · Foundations · Day 26Deliverable: Checkbox + Radio in Figma and code with shared Field wrapper, full state matrix, and keyboard spec in Storybook
Ship Checkbox and Radio as one selection family
Grouped inputs share focus, label, and error wiring; build them together or you pay the tax twice.
Checkbox and Radio look simple until you count the states: unchecked, checked, indeterminate, disabled, read-only, error, and focus-visible on a hit target that is too small on mobile. Map every state in Figma first, then mirror the same prop names in React.
Use a shared Field wrapper so label, description, and error text attach once. Radio groups need a fieldset, legend, and roving tabindex. Checkbox groups need a clear group label even when options stack vertically.
Keyboard contract: Space toggles checkbox. Arrow keys move within a radio group. Tab moves between groups, not every option. Test with VoiceOver and NVDA before you call either component done.
Today's output
Checkbox + Radio in Figma and code with shared Field wrapper, full state matrix, and keyboard spec in Storybook
Day 26 · checkbox-radioPage
Phase 2 · Foundations · Day 27Deliverable: Select component with native + custom paths, size variants, and keyboard/open-state tests in Storybook
Build Select with native fallback and custom parity
Select is where design intent dies in handoff unless Figma, code, and keyboard behavior match.
Start with a native select styled to your tokens. It covers mobile, autofill, and screen readers on day one. Layer a custom listbox only when you need search, multi-select, or rich options.
Document size variants, placeholder behavior, disabled options, and empty states. The trigger, chevron, and menu panel must use semantic spacing tokens, not one-off padding.
Typeahead and open-on-focus are optional. Focus trap inside the listbox is not. Close on Escape, restore focus to the trigger, and announce the selected value to assistive tech.
Today's output
Select component with native + custom paths, size variants, and keyboard/open-state tests in Storybook
Day 27 · selectPage
Phase 2 · Foundations · Day 28Deliverable: Switch + Field error components with i18n-length stress tests and DE/EN error copy examples in docs
Add Switch and Field error with EU copy in mind
Error text and toggle labels that work in English often break in German; plan for longer strings now.
Switch is not a fancy checkbox. It commits a setting immediately, so show on/off state, loading, and disabled clearly. Pair it with visible text: never rely on color alone.
Field error is a system primitive, not a one-off red line. It carries the message, links to the input via aria-describedby, and survives validation on blur and on submit. Write error copy in a direct voice: what failed and how to fix it.
For EU products, stub DE and EN error strings in Storybook. German validation messages run 30 to 40 percent longer. If your error row is single-line and fixed height, it will clip before you reach production.
From the field
At Nord, with dozens of teams shipping across EU markets, a checkout validation string in German wrapped past its container and pushed the primary button below the fold on mobile. We fixed the layout, not the translation. Field error became a flex row with multi-line support and a min-height token, not a single-line afterthought.
Today's output
Switch + Field error components with i18n-length stress tests and DE/EN error copy examples in docs
Your notes
Day 28 · switch-field-errorPage
Phase 2 · Foundations · Day 29Deliverable: Form smoke-test story in Storybook, pass/fail checklist completed by design and eng
Run a keyboard-only form smoke test today
If you cannot complete the pilot form without a mouse, the form stack is not shippable.
Assemble every control you have shipped so far into one Storybook story or a staging route: text inputs, select, checkbox, radio, switch, labels, and field errors. Use real validation rules from the pilot surface.
Run the test yourself first: Tab through, fix focus order, submit with Enter, recover from errors. Then ask one engineer and one designer to repeat it cold. Log every friction point in a shared doc.
Pass criteria: visible focus on every control, logical tab order, errors announced and readable, submit blocked until required fields pass. Anything less blocks Tier 2 sign-off.
Today's output
Form smoke-test story in Storybook, pass/fail checklist completed by design and eng
Day 29 · form-smoke-testPage
Phase 2 · Foundations · Day 30Deliverable: Alert + Badge with four severities, dismiss pattern, and do/don't usage notes in Storybook
Ship Alert and Badge for status without noise
Feedback components earn trust when severity is obvious and dismiss behavior is predictable.
Alert carries four intents: info, success, warning, error. Each maps to semantic color tokens and includes an icon, title, body, and optional action. Do not use alert styling for marketing banners.
Badge is for compact metadata: count, status, or label on a parent. Keep contrast at WCAG AA on every semantic background. Cap count display at 99+ so layout does not explode.
Dismissible alerts need a named close control, focus management on dismiss, and a policy: does the message return on refresh or stay gone? Document the default in your usage notes.
Today's output
Alert + Badge with four severities, dismiss pattern, and do/don't usage notes in Storybook
Day 30 · alert-badgePage
Phase 2 · Foundations · Day 31Deliverable: Card component with header/body/footer slots, density variants, and one settings-panel recipe story
Make Card the default container for grouped content
One Card primitive kills dozens of bespoke bordered boxes across squads.
Card is a surface token, border, radius, and optional elevation. Slots: header, body, footer. Interactive cards get hover and focus-visible; static cards do not pretend to be buttons.
Define density variants: compact for tables and settings, comfortable for marketing modules. Padding comes from spacing tokens, not hand-tuned values per screen.
Ship one composed example in Storybook: settings row with title, description, and action in the footer. That recipe is what squads copy, not the raw box.
Today's output
Card component with header/body/footer slots, density variants, and one settings-panel recipe story
Day 31 · cardPage
Phase 2 · Foundations · Day 32Deliverable: Divider, Stack, and Grid in Figma + code with gap tokens, responsive examples, and lint note for raw margins
Add Divider, Stack, and Grid as layout primitives
Layout primitives turn one-off spacing hacks into a rhythm squads can reuse.
Divider handles horizontal and vertical rules with semantic border tokens. Offer inset and full-bleed variants so lists and menus do not fight the spec.
Stack is flex with tokenized gap and alignment props. Grid covers the two or three column layouts you see in every settings and dashboard screen. Ban magic margin strings in pilot PRs starting today.
Document breakpoint behavior once. Stack that wraps at sm, Grid that collapses to one column: same rules in Figma auto-layout docs and in code props.
Today's output
Divider, Stack, and Grid in Figma + code with gap tokens, responsive examples, and lint note for raw margins
Day 32 · divider-stack-gridPage
Phase 2 · Foundations · Day 33Deliverable: Skeleton variants for text, media, and card blocks with reduced-motion support and one pilot route wired
Ship Skeleton so loading states match final layout
A skeleton that jumps when content loads is worse than a spinner.
Skeleton mirrors real component geometry: text lines, avatar circles, card blocks. Tie shimmer or pulse to your motion tokens and respect prefers-reduced-motion.
Pair skeleton stories with their loaded counterparts in Storybook. Designers and engineers should see the before and after on the same page.
Use skeletons on the pilot surface for one slow route this week. Measure perceived wait, not just bundle size. At Flyeralarm we cut reclamation friction ~90% once error and loading states told users what was happening instead of leaving them guessing.
Today's output
Skeleton variants for text, media, and card blocks with reduced-motion support and one pilot route wired
Day 33 · skeletonPage
Phase 2 · Foundations · Day 34Deliverable: Modal with focus trap, scroll lock, size variants, and destructive confirmation example in Storybook
Build Modal with focus trap and escape hatches
A modal that loses focus or scrolls the page behind it will fail review every time.
Modal stacks overlay, dialog panel, header, body, and footer actions. Size variants map to max-width tokens. Trap focus inside while open; return focus to the trigger on close.
Block background scroll, close on Escape, and keep a visible close control. Destructive confirmations use a danger button token, not a surprise red link in the corner.
Document when to use Modal vs. drawer vs. inline expansion. If the content is long or multi-step, Modal is probably wrong.
Today's output
Modal with focus trap, scroll lock, size variants, and destructive confirmation example in Storybook
Day 34 · modalPage
Phase 2 · Foundations · Day 35Deliverable: Tooltip + Popover with shared positioning, keyboard specs, and do/don't table in docs
Ship Tooltip and Popover with clear interaction contracts
Tooltips show; popovers invite interaction. Mix them up and accessibility breaks fast.
Tooltip: hover and focus on the trigger, brief text, no interactive children inside. Delay open slightly; close on Escape. Never put essential instructions only in a tooltip.
Popover: click or explicit trigger, can hold links and buttons, positions with collision detection. Manage focus inside when opened from keyboard.
Both need arrow offset tokens and a shared positioning utility so you are not maintaining twelve one-off popper configs by day 40.
From the field
On a Flyeralarm print-config flow, critical bleed guidance lived only in a hover tooltip desktop users never saw on touch devices. We moved specs into a popover with a labeled trigger; support tickets on file setup dropped sharply. If the info is required to complete the task, it cannot live in a tooltip.
Today's output
Tooltip + Popover with shared positioning, keyboard specs, and do/don't table in docs
Your notes
Day 35 · tooltip-popoverPage
Phase 2 · Foundations · Day 36Deliverable: Tabs component with horizontal/vertical variants and Tier 2 completion checklist signed off
Finish Tier 2 with Tabs that work from keyboard
Tabs close the overlay set; get roving focus and active panel wiring right once.
Use the tabs role pattern: tablist, tab, and tabpanel. Roving tabindex on arrows; Enter or Space activates. Active panel is the only one in the tab order for its contents.
Support horizontal and vertical orientations and overflow scroll on small screens. Disabled tabs are focusable but not activatable, or removed from the list; pick one rule and document it.
Mark Tier 2 complete in your tracker: every form control and overlay from the priority list has Figma, code, Storybook, and a keyboard note.
Today's output
Tabs component with horizontal/vertical variants and Tier 2 completion checklist signed off
Day 36 · tabsPage
Phase 2 · Foundations · Day 37Deliverable: Five design principles with do/don't pairs, published in Storybook and Figma library
Run a 90-minute principles workshop with squads
Principles turn token choices into decisions your team can defend under pressure.
Facilitate with sticky notes: ask what the system optimizes for (speed, density, clarity, brand) and what it refuses to do. Aim for five principles max, each with a do and a don't example from your product.
Write them in plain language, not poster slogans. 'We favor clarity over decoration' beats 'Delight at every touchpoint.'
Publish the final list at the top of your Storybook docs and in Figma library description. Reference them in PR reviews when someone proposes a one-off.
Today's output
Five design principles with do/don't pairs, published in Storybook and Figma library
Day 37 · principles-workshopPage
Phase 2 · Foundations · Day 38Deliverable: Pattern backlog ranked by frequency with two Phase 3 briefs and named owners
Scope Phase 3 patterns before you build more components
Components are the bricks; patterns are what squads actually need to ship screens.
Audit the pilot surface for repeating compositions: settings panels, empty states, data tables, onboarding steps. Score each by frequency and pain.
Map candidates to the six areas: layout, styles, components, regions, content, usability. If it is only a new button variant, it is not a pattern.
Pick two patterns for Phase 3 week one. Write a one-page brief per pattern: user job, components used, open questions, owner.
Today's output
Pattern backlog ranked by frequency with two Phase 3 briefs and named owners
Day 38 · patterns-scopingPage
Phase 2 · Foundations · Day 39Deliverable: Three-times request log template, first week's entries, and review cadence on the calendar
Apply the three-times rule to every expansion request
The third repetition is your signal to systematize; before that, local solutions are cheaper.
When a squad asks for a new variant, ask where else it appeared. One-off: extend with an escape hatch. Twice: note it in the backlog. Third time: promote to the library with a proper API.
Log requests in a lightweight table: requester, surface, count, decision. Review weekly with your sponsor so scope creep stays visible.
Say no with a path: 'Not in v0.1; here's how to compose with Card and Stack today.' That keeps trust while you protect the canon.
Today's output
Three-times request log template, first week's entries, and review cadence on the calendar
Day 39 · three-times-rulePage
Phase 2 · Foundations · Day 40Deliverable: Storybook docs pass for all Tier 1-2 components, one recipe page, changelog page, Lighthouse run logged
Upgrade Storybook from component gallery to product docs
Docs are adoption infrastructure; if usage notes are missing, squads will fork.
Every Tier 1 and Tier 2 entry needs: when to use, when not to use, props table, accessibility notes, and links to Figma. Add one recipe page that composes five components into a realistic screen.
Enable autodocs or equivalent so props stay in sync with TypeScript. Add a changelog page that lists what shipped this week.
Run Lighthouse on the static Storybook build. You are not shipping marketing pages, but sub-80 performance on docs hurts trust. I have seen doc sites climb from 34 to 77 once image lazy-loading and font subsetting were fixed.
Today's output
Storybook docs pass for all Tier 1-2 components, one recipe page, changelog page, Lighthouse run logged
Day 40 · storybook-docsPage
Phase 2 · Foundations · Day 41Deliverable: Midpoint demo script, Loom backup, and pre-read email draft to squad leads
Prep the midpoint demo like a release, not a slide deck
Squad leads adopt what they see working in their language, not what you promise on a roadmap.
Build a 15-minute script: open Storybook, walk the pilot form smoke test, show Figma library parity, show one real PR using the package. No voice-over of future quarters.
Pre-record a Loom fallback in case live demo fails. Test screen share, font loading, and npm install from the same registry squads use.
Send a pre-read 24 hours ahead: what's in v0.1, what's explicitly out, how to file requests. Collect three questions you expect and answer them in the demo.
Today's output
Midpoint demo script, Loom backup, and pre-read email draft to squad leads
Day 41 · midpoint-demo-prepPage
Phase 2 · Foundations · Day 42Deliverable: Readout notes with tagged issues, severity, owners, and office hours scheduled
Run the squad readout and collect honest feedback
Midpoint is a contract reset: show what exists, hear what's blocking adoption, assign owners.
Present live Storybook and the Figma library side by side. Ask two questions: what would you use Monday, and what would you still build locally? Write answers in the room, not in a retro doc weeks later.
Capture parity gaps, missing tokens, and doc holes as tagged issues with severity. Critical means blocks pilot migration; nice-to-have waits for Phase 3.
End with named actions: who fixes parity, who updates docs, when v0.1 publishes. Schedule a 30-minute office hours slot the week after publish.
From the field
During Aurora at Nord, we ran a readout with squad leads across dozens of teams. One team had rebuilt Modal locally because ours lacked a destructive variant. They were right. We added the variant that week, deprecated theirs in the next sprint, and migration jumped because we treated their fork as signal, not disobedience.
Today's output
Readout notes with tagged issues, severity, owners, and office hours scheduled
Your notes
Day 42 · squad-readoutPage
Phase 2 · Foundations · Day 43Deliverable: Parity fix PRs for all critical readout items and designer sign-off on top five components
Burn down Figma-to-code parity gaps from the readout
Two-pixel drift across twenty components reads as 'the system is not ready.'
Sort readout issues by pilot impact. Fix spacing, typography, and state mismatches before you touch net-new API surface.
Use side-by-side screenshots or visual regression if you have it. If not, designer sign-off on five high-traffic components is the minimum bar.
Update your parity rule in the README: Figma and Storybook match within 2px on spacing, exact on color tokens. Link the rule in every PR template.
Today's output
Parity fix PRs for all critical readout items and designer sign-off on top five components
Day 43 · parity-fixesPage
Phase 2 · Foundations · Day 44Deliverable: Duplicate inventory with migrate/wrap/delete decisions and at least one pilot PR migrating to system imports
Deprecate duplicate components on the pilot surface
Two buttons in the repo means every squad picks the wrong one half the time.
Inventory local copies in the pilot app: buttons, inputs, cards, modals. Tag each as migrate, wrap, or delete. No silent duplicates past publish day.
Ship codemods or find-replace recipes where imports are mechanical. For messy ones, pair with the owning squad for one migration PR they can copy.
Add deprecation notices in old components pointing to the system import. Set a removal date six weeks out and message it in the changelog.
Today's output
Duplicate inventory with migrate/wrap/delete decisions and at least one pilot PR migrating to system imports
Day 44 · deprecate-duplicatesPage
Phase 2 · Foundations · Day 45Deliverable: v0.1.0 published to npm, Figma library released, changelog posted, git tag and Storybook deploy linked
Publish library v0.1 to npm and Figma
v0.1 is a credible foundation: tokens, Tier 1-2 components, docs, and a changelog.
Version 0.1.0 on npm with locked peer dependencies documented. Publish the Figma library to the team workspace with a release note comment.
Changelog entries: what shipped, what is deprecated, known gaps, how to install. Tag the git release and link Storybook deploy.
Post in the team channel with install command, Figma link, and office hours time. Ask for one adopter to speak at the next guild. Adoption starts with one visible win.
Today's output
v0.1.0 published to npm, Figma library released, changelog posted, git tag and Storybook deploy linked
Day 45 · publish-v0-1Page
Phase 2 · Foundations · Day 46Deliverable: Up to two overflow fixes shipped or formally deferred with documented workarounds
Use buffer day one for overflow components and edge cases
Something always slips: today is for the stragglers that block pilot migration.
Reopen the readout backlog. Pick the top two items that are not quite defects but still block a squad: missing size variant, icon slot, or loading state on a Tier 2 piece.
Timebox each fix to half a day. If it needs more, downgrade to Phase 3 and document the workaround.
Ship a patch release 0.1.1 if you change public APIs; otherwise bundle doc fixes into the same week.
Today's output
Up to two overflow fixes shipped or formally deferred with documented workarounds
Day 46 · overflow-bufferPage
Phase 2 · Foundations · Day 47Deliverable: One infrastructure debt item resolved and documented for the team
Pay down debt that slows every future component
Fix the pipeline cracks now or every Phase 3 pattern ships slower.
Audit token export, Storybook build time, CI visual checks, and Figma publish friction. Pick the one infrastructure pain that cost you the most hours in Phase 2.
Spend today on that single upgrade: faster Style Dictionary run, autodocs fix, or a shared test util for focus trap. No new components.
Write a short ADR or README section so the next contributor does not rediscover the same trap.
Today's output
One infrastructure debt item resolved and documented for the team
Day 47 · debt-bufferPage
Phase 2 · Foundations · Day 48Deliverable: Champion roster, kit shared, and pre-Phase 3 sync on the calendar
Recruit champions and arm them to sell v0.1 locally
Systems scale through people who answer Slack questions before you do.
Identify one design and one engineering champion per pilot squad. They get early access, direct Slack line to you, and responsibility to file crisp issues.
Give them a champion kit: install snippet, Figma link, migration recipe, three-minute Loom of the form smoke test, and FAQ for the objections you heard at readout.
Schedule a 30-minute champion sync before Phase 3 starts. Their job next week: one migrated screen each.
Today's output
Champion roster, kit shared, and pre-Phase 3 sync on the calendar
Day 48 · champions-bufferPage
Phase 2 · Foundations · Day 49Deliverable: Phase 2 scorecard, sponsor one-pager, and Phase 3 week-one agenda circulated
Close Phase 2 with metrics, gaps, and a Phase 3 kickoff
Day 49 is the handoff: honest scorecard, no victory lap for work still local.
Update your tracker: components in Figma, components in npm, pilot screens migrated, open parity issues, Storybook Lighthouse score, weekly active importers if you can measure them.
Write a one-page Phase 2 retrospective for your sponsor: what shipped, what slipped, what Phase 3 must cover first. Tie at least one metric to business language: fewer duplicate PRs, faster form build, or reduced a11y defects.
Preview Phase 3 week one: first pattern build, governance cadence, expanded universe policy. Squads should know what changes on day 50 before they log off.
Connected npm delivery, docs, migration, and production proof.
Package and semver
Token CI pipeline
Storybook and recipes
Pilot migration sprint
Production deploy
Part III · ConnectPage
Phase 3 · Connect · Day 50Deliverable: Package scaffold with build pipeline and @your-org/design-system entry point
Scaffold the npm package
A design system that ships as a ZIP is shelfware by Friday.
Today you turn the library into a product. Create a dedicated package (monorepo workspace or standalone repo) with a clean entry point, ESM and CJS builds, and a publish script wired to your internal registry or npm org.
Export tokens, primitives, and Tier 1 components from one barrel file. Keep side effects out of the root import so tree-shaking works in consuming apps.
Tag this as v0.1.0 internally even if you do not publish publicly yet. Connected delivery starts when another repo can install you, not copy your folder.
Today's output
Package scaffold with build pipeline and @your-org/design-system entry point
Day 50 · npm-scaffoldPage
Phase 3 · Connect · Day 51Deliverable: Peer dep matrix in README plus one live import in a pilot app
Document peer deps and wire the first import
Peer dependencies are the contract between your system and every product app.
List React, React DOM, and any styling runtime as peerDependencies with a tested version range. Document them in a short README section: what versions you support and what breaks outside that range.
Install the package in one pilot app today. Replace a single local Button with the import. Run the app. Fix whatever breaks before you touch a second component.
Log the import path pattern your squads should copy: import { Button } from '@your-org/design-system'. One pattern, everywhere.
Today's output
Peer dep matrix in README plus one live import in a pilot app
Day 51 · peer-deps-importPage
Phase 3 · Connect · Day 52Deliverable: CHANGELOG.md, release script, and v0.1.0 tagged release
Ship semver and a real changelog
Version numbers are promises. Your changelog is how squads trust them.
Adopt semver strictly: patch for fixes, minor for backward-compatible additions, major for breaking API changes. Write a CHANGELOG.md entry for v0.1.0 covering everything in Phase 2.
Add a release script that bumps version, updates the changelog, and tags git. Even internal packages deserve release discipline.
Send a short release note to the pilot squad channel. Tell them what changed, what to upgrade, and who to ping if something breaks. Connected npm delivery means squads hear about releases before they hit production.
Today's output
CHANGELOG.md, release script, and v0.1.0 tagged release
Day 52 · semver-changelogPage
Phase 3 · Connect · Day 53Deliverable: CI job that exports Figma variables and fails on token drift
Automate token export in CI
Manual token sync is where Figma and code drift apart.
Add a CI job that exports Figma variables on every merge to main. Fail the build if the exported JSON differs from what is committed. No silent drift.
Wire the job to your design token source file in the repo. Designers merge variable changes in Figma; engineers review the PR with the diff.
Document the flow in three steps: edit variables in Figma, export triggers on merge, consuming apps pick up changes on their next package upgrade.
Today's output
CI job that exports Figma variables and fails on token drift
Day 53 · token-ciPage
Phase 3 · Connect · Day 54Deliverable: Style Dictionary config producing CSS, Tailwind, and JSON outputs
Run tokens through Style Dictionary
One token source should feed CSS, Tailwind, and JSON without copy-paste.
Configure Style Dictionary to transform your token JSON into CSS custom properties, a Tailwind theme extension, and platform JSON for mobile or other consumers.
Map primitives to semantics in the config, not in component files. Components reference semantic tokens only.
Run the transform in CI after the Figma export step. Commit the generated outputs or generate them at build time, but never hand-edit them.
Today's output
Style Dictionary config producing CSS, Tailwind, and JSON outputs
Day 54 · style-dictionaryPage
Phase 3 · Connect · Day 55Deliverable: CI lint rule blocking hardcoded hex in component files
Ban hardcoded hex in CI
If hex slips back into component files, your token architecture is theater.
Add an ESLint or stylelint rule that flags hex colors, rgb(), and raw pixel spacing in component source files. Allow exceptions only in token definition files.
Run the lint in CI on every PR to the system repo. Block merge on violations.
Fix any existing violations today. Most teams find 10 to 30 one-off colors hiding in legacy components. Clean them before the pilot migration starts.
Today's output
CI lint rule blocking hardcoded hex in component files
Day 55 · lint-hex-banPage
Phase 3 · Connect · Day 56Deliverable: Code Connect files for Tier 1 core components, published in Figma
Link core components with Code Connect
Designers should see real code when they inspect a component in Figma.
Create Code Connect files for Tier 1 components: Button, Link, Input, Card. Map Figma properties to code props so the snippet in Dev Mode matches what engineers import.
Publish the links to your Figma library. Run a 15-minute session with one designer: have them click through three components and confirm the snippets look right.
Prioritize the components your pilot squad uses weekly. Perfect parity on five components beats partial coverage on twenty.
From the field
On a Nord-scale project, designers stopped asking 'which Button is real?' once Code Connect went live. Dev Mode showed the same import path engineers used. Slack threads about component names dropped within a week. The map finally matched the territory.
Today's output
Code Connect files for Tier 1 core components, published in Figma
Your notes
Day 56 · code-connectPage
Phase 3 · Connect · Day 57Deliverable: Storybook docs pages for all Tier 1 components with usage notes
Document Tier 1 in Storybook
Every Tier 1 component gets a docs page before you ask squads to adopt it.
Add autodocs pages for Button, Link, Text, Icon, Card, Stack, and Divider. Include props table, all states, and a short usage note: when to use, when not to.
Show token references in the docs, not raw values. Link to the semantic token name so designers and engineers speak the same language.
Add a 'Related components' section on each page. Adoption starts when people can navigate from Button to Link without opening Slack.
Today's output
Storybook docs pages for all Tier 1 components with usage notes
Day 57 · storybook-tier-1Page
Phase 3 · Connect · Day 58Deliverable: Storybook docs for all Tier 2 components with a11y and keyboard notes
Document Tier 2 in Storybook
Form controls and overlays need docs that cover keyboard paths, not just screenshots.
Complete docs for Input, Textarea, Select, Checkbox, Radio, Switch, Label, Alert, Badge, Tooltip, Modal, and Tabs. Every interactive component shows focus-visible and disabled states.
Add accessibility notes: expected keyboard behavior, ARIA roles, and contrast pairs. Copy the pattern from your Button docs.
Run a keyboard-only walkthrough of a sample form built from these docs. If you cannot complete it without a mouse, fix the component before you document it.
Today's output
Storybook docs for all Tier 2 components with a11y and keyboard notes
Day 58 · storybook-tier-2Page
Phase 3 · Connect · Day 59Deliverable: Settings panel recipe page with Figma frame and Storybook story
Write the settings panel recipe
Squads do not build screens. They build settings panels, tables, and wizards.
Document a settings panel pattern: page shell, section headers, form rows, save/cancel actions. Compose it from Tier 1 and Tier 2 components only.
Include a Figma frame and a live Storybook story side by side. Show spacing tokens, not magic numbers.
Add copy guidance for labels, helper text, and error messages. Content is part of the system.
Today's output
Settings panel recipe page with Figma frame and Storybook story
Day 59 · settings-recipePage
Phase 3 · Connect · Day 60Deliverable: Data table and onboarding wizard recipe pages in Storybook
Write table and onboarding recipes
Two more flows cover most of what product teams actually ship.
Document a data table pattern: toolbar, column headers, row actions, empty state slot, and pagination footer. Note which pieces are Tier 3 and which compose from Tier 1.
Document an onboarding wizard: step indicator, content area, primary/secondary actions, and skip path. Three steps is enough for the recipe.
Nest component links inside each recipe. Nobody walks into a store to buy loose bricks.
Today's output
Data table and onboarding wizard recipe pages in Storybook
Day 60 · table-onboarding-recipePage
Phase 3 · Connect · Day 61Deliverable: 'Build your first screen' onboarding guide with Loom walkthrough
Publish the first screen guide
New hires should ship a screen on day one, not week three.
Write a 'Build your first screen' guide: install the package, import tokens, compose a Card with a form, deploy to staging. Target 60 minutes for a mid-level engineer.
Include troubleshooting for the three errors you see most: peer dep mismatch, missing CSS import, and wrong token path.
Record a 10-minute Loom walking through the guide. Link it from the README and the pilot squad channel.
Today's output
'Build your first screen' onboarding guide with Loom walkthrough
Day 61 · first-screen-guidePage
Phase 3 · Connect · Day 62Deliverable: Restructured docs IA with products-first navigation
Restructure docs around products, not parts
Lead with recipes and flows. Nest components inside what teams actually build.
Move component reference pages under the recipe that uses them most. Button lives inside Settings, not at the top level.
Add a one-line description under each nav item that answers 'when would I open this?' Docs are a store, not a warehouse.
Today's output
Restructured docs IA with products-first navigation
Day 62 · docs-ia-productsPage
Phase 3 · Connect · Day 63Deliverable: Search usability report with fixes shipped and FAQ page
Run a docs search pass
If squads cannot find the answer in 30 seconds, they will ask in Slack.
Ask three people who did not build the system to find: how to install, how to build a settings page, and how to handle empty states. Time them. Log every dead end.
Fix titles, add aliases, and cross-link related pages based on what they searched for versus what they found.
Add a search index if your docs site supports it. Map common Slack questions to doc URLs in a short FAQ page.
From the field
I ran this pass on Aurora at Nord. Three engineers averaged four minutes to find the Modal docs. After renaming pages and adding aliases, the same task took under 40 seconds. Slack questions about imports dropped the following week.
Today's output
Search usability report with fixes shipped and FAQ page
Your notes
Day 63 · search-passPage
Phase 3 · Connect · Day 64Deliverable: DataTable component with sort, skeleton, and empty state slot
Build the data table component
Build only what the pilot surface needs. Skip the enterprise grid until a third squad asks.
Ship a DataTable with sortable columns, a loading skeleton, and an empty state slot. Use your table recipe from day 60 as the spec.
Support keyboard navigation for row focus and sort toggles. Document the API in Storybook with three real column configurations.
Resist feature creep: no inline edit, no column resize, no export-to-CSV until a funded ticket demands it.
Today's output
DataTable component with sort, skeleton, and empty state slot
Day 64 · data-tablePage
Phase 3 · Connect · Day 65Deliverable: Navigation component matching pilot app shell, documented in Storybook
Build navigation for the pilot surface
Nav is product-specific. Build the shell your pilot actually uses.
Ship a Sidebar or TopNav matching the pilot app layout. Include active state, collapsed mode, and mobile breakpoint behavior.
Use semantic tokens for backgrounds, borders, and text. No one-off colors in the nav shell.
Document which parts are configurable (menu items, icons, badges) and which are fixed (spacing, typography, focus ring).
Today's output
Navigation component matching pilot app shell, documented in Storybook
Day 65 · navigationPage
Phase 3 · Connect · Day 66Deliverable: EmptyState component with three variants and approved copy
Ship empty state patterns
Empty screens are where inconsistent UI hides in plain sight.
Build an EmptyState component: illustration slot, headline, body copy, and primary action. Ship three variants: no data, no results, and error.
Write approved microcopy for each variant. UX writing is part of the system, not a squad afterthought.
Add empty states to your table recipe and onboarding recipe. Show them in Storybook with real copy, not lorem ipsum.
Today's output
EmptyState component with three variants and approved copy
Day 66 · empty-statesPage
Phase 3 · Connect · Day 67Deliverable: Deprecation tags in code and Figma plus published deprecation log
Tag deprecated components in code and Figma
Deprecation warnings in code beat surprise Slack threads every time.
Add @deprecated JSDoc tags to legacy components the pilot will replace. Point each tag to the canonical replacement and the target removal version.
Mark matching Figma components with a 'Deprecated' variant or a clear naming prefix. Designers should see the warning before they use the old pattern.
Publish a deprecation log: old name, replacement, removal target, and owner. One table, linked from the changelog.
Today's output
Deprecation tags in code and Figma plus published deprecation log
Day 67 · deprecation-markersPage
Phase 3 · Connect · Day 68Deliverable: Legacy component inventory with replacements and effort estimates
Audit legacy components on the pilot surface
You cannot migrate what you have not inventoried.
Walk the pilot screens file by file. List every component that is not from the system: name, location, replacement, and effort estimate (S/M/L).
Flag snowflakes that should stay local versus patterns that belong in the system after the 3x rule. Be honest about what not to migrate.
Share the inventory with the pilot squad lead. Agree on scope for the migration sprint starting day 70.
Today's output
Legacy component inventory with replacements and effort estimates
Day 68 · legacy-auditPage
Phase 3 · Connect · Day 69Deliverable: Pilot migration kickoff deck with scope, owners, and success metrics
Kick off the pilot migration sprint
The Measuring Spoon Cycle starts when product UI feeds the system, not the other way around.
Hold a 30-minute kickoff with the pilot squad: scope, timeline, success metrics, and who owns design QA on each PR.
Set three metrics: lines of custom CSS removed, system components imported, and time to ship versus baseline. Track them daily.
Assign migration slices by screen, not by component type. One engineer owns settings, another owns the table view. Ship vertical slices.
From the field
Dan Mall's Measuring Spoon Cycle is the rhythm behind every system that lasts: make product UI, extract what repeats, reuse on the next screen. I have run this on funded pilot work at Nord scale. Extraction during the sprint beat a big-bang library launch every time.
Today's output
Pilot migration kickoff deck with scope, owners, and success metrics
Day 69 · pilot-kickoffPage
Phase 3 · Connect · Day 70Deliverable: Pilot app shell migrated to system navigation and layout components
Migration day 1: replace the shell
Start with layout and navigation. Swap the frame before the furniture.
Replace the pilot app shell: navigation, page layout, and spacing primitives. Import from the package, delete local CSS as you go.
Run the app after every screen swap. Do not batch ten changes and debug blind.
Log CSS lines removed and components imported in a shared tracker. Visible progress keeps the squad motivated.
Today's output
Pilot app shell migrated to system navigation and layout components
Your notes
Day 70 · migration-day-1Page
Phase 3 · Connect · Day 71Deliverable: Settings and form screens migrated with a11y spot check passed
Migration day 2: migrate forms and settings
Settings screens are the highest-density test of your Tier 2 work.
Replace form controls and the settings panel using your day 59 recipe. Every input, label, and error state comes from the package.
Run keyboard-only and screen reader spot checks on the migrated settings flow. Fix focus order issues before moving on.
Update the migration tracker. Target 50% of pilot screens on system components by end of day.
Today's output
Settings and form screens migrated with a11y spot check passed
Day 71 · migration-day-2Page
Phase 3 · Connect · Day 72Deliverable: Table and empty states migrated, legacy files removed, 90%+ coverage
Migration day 3: tables, empty states, and cleanup
Finish the pilot surface. Delete the legacy components you replaced.
Migrate the data table view and wire empty states using day 64 and day 66 components. Remove unused legacy files from the repo.
Run a grep for hardcoded hex in the pilot app. Fix or file tickets for anything outside the system repo.
Target 90%+ of pilot screens on system components. Document the remaining snowflakes with a reason each one stays local.
Today's output
Table and empty states migrated, legacy files removed, 90%+ coverage
Day 72 · migration-day-3Page
Phase 3 · Connect · Day 73Deliverable: Migration PR with screenshots, design QA checklist completed
Run design QA on the migration PR
Design QA is a PR checklist, not a meeting that happens after merge.
Open a PR with the full migration diff. Attach before/after screenshots for every changed screen.
Walk the design QA checklist: system components only, tokens not hex, focus visible, contrast OK, Storybook updated if APIs changed.
Designer and engineer review together in one session. Log fixes as PR comments, not a separate feedback doc.
Today's output
Migration PR with screenshots, design QA checklist completed
Day 73 · design-qa-prPage
Phase 3 · Connect · Day 74Deliverable: Migration merged to staging with smoke test report
Merge to staging and smoke test
Staging is where you catch the surprises that CI cannot see.
Merge the migration PR to staging. Run a full smoke test: every migrated screen, every interactive state, mobile and desktop.
Compare Lighthouse scores before and after. Note bundle size and CSS footprint changes.
Fix staging blockers same day. Do not carry known bugs into production deploy.
Today's output
Migration merged to staging with smoke test report
Day 74 · merge-stagingPage
Phase 3 · Connect · Day 75Deliverable: Production deploy with before/after metrics report for sponsor readout
Deploy to production and measure
Production proof is the only proof that matters for your sponsor.
Deploy the migrated pilot surface to production. Monitor error rates and key user flows for 24 hours.
Capture before/after metrics: custom CSS lines removed, system components imported, Lighthouse score, and time to ship the migration versus baseline.
On one project I led, Lighthouse went from 34 to 77 partly from this kind of discipline: fewer one-off components, less CSS chaos. Performance is a design system outcome whether your roadmap admits it or not.
Today's output
Production deploy with before/after metrics report for sponsor readout
Day 75 · production-deployPage
Phase 3 · Connect · Day 76Deliverable: Before/after metrics sheet with pilot surface screenshots
Measure what changed in production
Numbers close Phase 3. Without before/after proof, adoption stays a nice story.
Your pilot surface is live on system components. Today you stop guessing and start counting. Pull a baseline from day 71 (or your best estimate) and compare to what shipped yesterday.
Track the metrics that matter to your sponsor: custom CSS lines removed, hardcoded hex count, package imports, time to ship the feature, Lighthouse or Core Web Vitals if you have them. On one project I led, Lighthouse went 34 to 77 partly from killing one-off components and CSS chaos.
Fill one row per metric in your worksheet. Screenshot the pilot surface before and after. These numbers become your Phase 4 pitch to squads 2 and 3. If you cannot show relief, fix the pilot before you roll out.
Today's output
Before/after metrics sheet with pilot surface screenshots
Day 76 · before-after-metricsPage
Phase 3 · Connect · Day 77Deliverable: Phase 3 retro notes + AI usage section in docs
Close Connect with a retro and AI guardrails
Phase 3 ends when the system is production-real and machine-readable, not when the Figma file looks finished.
Run a 45-minute retro with the pilot squad: what worked, what broke, what docs were missing. Capture every gap as a doc PR or backlog item due within one week. No parking-lot items without an owner.
Phase 3 done when: pilot is in production on system components, token export is automated, docs answer "how do I build X?" without Slack. Tick those boxes honestly. Gaps you skip here become adoption blockers in Phase 4.
Add AI readiness to your docs today: semantic token names, component states in Storybook, do's and don'ts with rationale, deprecation status visible. One prompt can generate a dozen plausible screens that all disagree. Your system is the constraint that makes AI reliable.
Today's output
Phase 3 retro notes + AI usage section in docs
Your notes
Day 77 · phase-3-retro-ai-readinessPage
Part IV
Adopt · Days 78-100
Multi-squad rollout, governance, metrics, and handover.
Champions and office hours
Squad migrations
RFC and PR checklist
Enablement and metrics
Day 100 handover
Part IV · AdoptPage
Phase 4 · Adopt · Day 78Deliverable: Rollout kickoff deck + champion roster with squad assignments
Kick off rollout and name your champions
Adoption is a people problem with a calendar. Start with a kickoff, not a Slack announcement.
Schedule a 30-minute rollout kickoff with squad leads, your sponsor, and the system owner. Show the pilot metrics from day 76. Name the problem: UI inconsistency costs rework hours every sprint. Name the path: one feature per squad, system components only, office hours twice a week.
Each migrating squad gets one design champion and one engineering champion. Champions are not volunteers by default. They are named by the squad lead, with 20% of sprint capacity protected for migration work. Champions attend office hours and route questions to docs.
Publish a rollout calendar: squad 2 migrates days 81-85, squad 3 days 86-88, governance live by day 94. Deadlines create focus. Vague "we'll adopt eventually" creates shelfware.
Today's output
Rollout kickoff deck + champion roster with squad assignments
Day 78 · rollout-kickoff-championsPage
Phase 4 · Adopt · Day 79Deliverable: Squad 2 migration brief (one page, signed off by squad lead)
Write the squad 2 migration brief
A one-page brief beats a 40-slide deck. Squads need scope, deadline, and a clear definition of done.
Meet squad 2's champions for 30 minutes. Pick one funded feature surface to migrate (billing form, settings panel, onboarding step). Not the whole app. One wedge, same playbook as the pilot.
Write the brief: feature name, current pain, target components from the package, acceptance criteria (tokens only, no one-offs, a11y bar met), deadline (day 85), and who reviews the PR. Attach links to pattern recipes, not just component docs.
Get a verbal commit from the squad lead before you leave the room. If they cannot protect capacity, escalate to your sponsor today. Phase 4 dies in sprint planning fights you did not show up to.
Today's output
Squad 2 migration brief (one page, signed off by squad lead)
Questions that stay in Slack never become doc updates. Office hours turn friction into fixes the same week.
Create a recurring 30-minute Zoom invite, twice weekly, for the next four weeks. No agenda. Open floor. Post the link in #design-system and pin it. Champions are required attendees for their squad's migration window.
During each session, capture every question in a shared doc. If the answer exists in docs, send the link and fix the discoverability gap. If it does not exist, open a doc PR before Friday. Questions answered only in Slack are adoption tax.
Track attendance and top themes. By day 85 you want repeat questions trending toward zero. That is how you know docs are doing their job.
Phase 4 · Adopt · Day 81Deliverable: Squad 2 component audit spreadsheet + migration branch with layout shell on tokens
Squad 2 day 1: audit and branch
Migration starts with inventory, not code. Know what you are replacing before you import anything.
Squad 2 engineers run an audit on their target feature: list every custom component, hardcoded color, and local CSS file. Compare against the system package. Flag gaps (missing variants, no table pattern) as RFC candidates, not blockers.
Create the migration branch. Install or bump the design system package to the latest semver. Replace the app shell layout first: spacing, typography, and color tokens before you touch feature logic.
Daily checkpoint: audit spreadsheet complete, branch open, package imported, zero custom hex in the layout shell. Post status in the squad channel by end of day.
Today's output
Squad 2 component audit spreadsheet + migration branch with layout shell on tokens
Day 81 · squad-2-day-1-auditPage
Phase 4 · Adopt · Day 82Deliverable: PR with structural layout migrated to system components
Squad 2 day 2: structure and navigation
Get the frame right before you fill it. Page structure on system primitives sets the pattern for everything below.
Replace page headers, sidebars, and section containers with system layout components. If you do not have a layout primitive yet, compose Stack, Grid, and Card from Tier 1. Do not invent a one-off page wrapper.
Wire navigation and breadcrumbs to system patterns. Match Figma spacing to token values, not eyeballing. Designer and engineer pair for 60 minutes today if parity is off.
Daily checkpoint: feature shell renders on system layout only, no local margin/padding overrides, Storybook links added to the PR description for every component used.
Today's output
PR with structural layout migrated to system components
Day 82 · squad-2-day-2-structurePage
Phase 4 · Adopt · Day 83Deliverable: PR with form layer migrated; RFC opened for any missing states
Squad 2 day 3: forms and inputs
Forms are where drift hides. One custom input today becomes twelve button variants by next quarter.
Swap every input, select, checkbox, and button for system components. Use documented form patterns from your docs, not ad hoc label markup. Validation states must match Storybook: error, disabled, focus-visible.
Delete local form CSS as you go. If a state is missing from the system, open an RFC instead of forking a local variant. Composition beats customization.
Daily checkpoint: all form controls on system components, zero hardcoded colors in form CSS, keyboard tab order verified.
Today's output
PR with form layer migrated; RFC opened for any missing states
Day 83 · squad-2-day-3-formsPage
Phase 4 · Adopt · Day 84Deliverable: Design QA sign-off comment on migration PR
Squad 2 day 4: polish and design QA
The last 10% is where trust is won or lost. Polish is not optional.
Run design QA on the migration branch: spacing, typography, hover/focus states, empty states, loading states. Use the PR checklist even if it is not published yet. Designer signs off in the PR comment thread.
Fix a11y issues today, not in a follow-up ticket. Contrast, focus rings, and aria labels are part of definition of done. Run axe or your linter in CI if you have it.
Daily checkpoint: design QA sign-off recorded, a11y scan clean on migrated surface, no open P1 visual bugs.
From the field
On a billing migration, the squad lead tried to ship with a custom date picker "just for this flow." We spent 20 minutes in office hours mapping it to the system DateInput plus a documented range pattern. They saved two days of maintenance and became the team that stopped asking for exceptions.
Today's output
Design QA sign-off comment on migration PR
Your notes
Day 84 · squad-2-day-4-polishPage
Phase 4 · Adopt · Day 85Deliverable: Squad 2 feature live in production + updated metrics row
Squad 2 day 5: merge and measure
Ship it. A merged PR in production beats a perfect branch in review.
Merge squad 2's migration PR. Deploy to production or your staging equivalent with real traffic. Announce in #design-system with before/after screenshots and the package version used.
Update the metrics worksheet: add squad 2's custom CSS reduction, import count, and ship time vs. their baseline. One more row of proof for the sponsor deck.
Daily checkpoint: feature live in production on system components, metrics row filled, retro note captured (one thing that was hard, one doc gap to fix this week).
Today's output
Squad 2 feature live in production + updated metrics row
Squad 3 moves faster because the playbook exists. Reuse the brief template, do not reinvent the process.
Hand squad 3 the same brief format squad 2 used. They pick one feature surface, name champions, and commit to a day 88 ship date. Show squad 2's before/after as proof.
Run the component audit on day 86. Squad 3 should spot fewer gaps because docs improved during squad 2's migration. Any repeat gap is a doc bug: fix it today.
Daily checkpoint: squad 3 brief signed, audit complete, migration branch open with package installed.
Today's output
Squad 3 migration brief + audit spreadsheet
Day 86 · squad-3-kickoffPage
Phase 4 · Adopt · Day 87Deliverable: Squad 3 migration PR at 80%+ completion
Squad 3 day 2: core migration push
Two days to ship because the system earned trust on squad 2. Move fast, compose hard, RFC anything missing.
Squad 3 runs layout, forms, and primary UI in one focused push. Pair design and engineering for at least one Hot Potato session: pass the screen back and forth the same day.
Champions route blockers to office hours, not side DMs. System owner triages gaps: compose existing components, extend a variant, or defer to RFC. No silent forks.
Daily checkpoint: 80%+ of target surface on system components, blockers logged with owners, PR open with draft description.
Today's output
Squad 3 migration PR at 80%+ completion
Day 87 · squad-3-migratePage
Phase 4 · Adopt · Day 88Deliverable: Squad 3 feature live + two-squad adoption summary
Squad 3 day 3: ship and celebrate
Two squads in production ends the "pilot only" phase. You have a system now.
Finish design QA, merge, and deploy squad 3's feature. Same announcement format as squad 2: screenshots, version, link to pattern docs used.
Update metrics: two squads shipped, total package imports across repos, combined CSS reduction. This is the slide your sponsor cares about.
Hold a 15-minute win post in #engineering: name the champions, thank the pilot squad, point to office hours for squad 4 when they are ready.
Today's output
Squad 3 feature live + two-squad adoption summary
Day 88 · squad-3-shipPage
Phase 4 · Adopt · Day 89Deliverable: RFC template committed to repo and linked from docs
Publish the RFC template in the repo
Governance starts as a one-page template, not a committee.
Copy the RFC template into your repo under docs/governance/rfc-template.md. One page: problem, proposal, alternatives, adoption plan (which squads in 30 days), decision line for the owner.
Link it from your docs homepage and CONTRIBUTING.md. Squads propose changes through RFCs; the owner approves, defers, or rejects within five business days.
Canon vs. expanded universe: core tokens and Tier 1-2 are canon. Squad contributions live in expanded universe until the 3x rule hits (third squad needs it, a11y bar met, tests pass, owner assigned).
Today's output
RFC template committed to repo and linked from docs
Day 89 · rfc-templatePage
Phase 4 · Adopt · Day 90Deliverable: PR checklist in repo + linked from contributing guide
Ship the PR review checklist
Design QA in PRs catches drift before it ships. Make the checklist impossible to ignore.
Publish a PR checklist for design QA: system components only, tokens not hex, focus visible, contrast OK, keyboard path works, Storybook updated if API changed. Store it as .github/PULL_REQUEST_TEMPLATE.md or a required CI comment bot.
Train champions to use it on every migration PR. Owner reviews system PRs with the same list. Consistency here is how you stop "just this once" from becoming policy.
Add a "deprecated component used?" line with a link to the deprecation log. Surprises belong in changelogs, not production.
Today's output
PR checklist in repo + linked from contributing guide
Day 90 · pr-checklistPage
Phase 4 · Adopt · Day 91Deliverable: Flow/Systems week calendar for next 8 weeks (shared with eng manager)
Calendar Flow week and Systems week
System work loses every sprint fight unless it is on the calendar with the eng manager's buy-in.
Flow week: ship product UI on the system with the squad. Systems week: extract, document, version, test. No feature pressure. Alternate if your team is embedded in product squads.
Book the next two months with your eng manager and sponsor. Protect at least one systems week per month for the owner. Without explicit alternation, your roadmap becomes a graveyard of good intentions.
Share the calendar in #engineering. Name who owns each week. Flow without systems rots the library. Systems without flow loses political cover.
From the field
I watched a strong system die because every sprint was Flow week. Components shipped in product PRs but nobody extracted or documented them. Six months later, three squads had three different tables. Calendar the alternation or accept the tax.
Today's output
Flow/Systems week calendar for next 8 weeks (shared with eng manager)
Your notes
Day 91 · flow-systems-calendarPage
Phase 4 · Adopt · Day 92Deliverable: RFC triage notes with decisions on all open proposals
Run your first RFC triage
Triage is where governance becomes real. Review proposals weekly, decide fast, document everything.
Schedule 30 minutes with the system owner and one champion. Review every open RFC: approve, defer, or reject. Deferrals need a revisit date. Rejections need a one-line reason and a suggested alternative (usually: compose existing components).
Approved RFCs get an owner and a target release. No approved RFC sits unassigned. This is the rhythm that replaces hallway negotiations.
Log decisions in a CHANGELOG or governance doc. Transparency builds trust faster than perfect components.
Today's output
RFC triage notes with decisions on all open proposals
Day 92 · rfc-triagePage
Phase 4 · Adopt · Day 93Deliverable: First external RFC merged + changelog entry published
Merge the first external RFC
The first squad-contributed change proves the system is shared infrastructure, not your side project.
Take one approved RFC from triage through merge this week. It can be small: a new Button size, a Tooltip variant, a documented empty state. Size matters less than the process working end to end.
Contributor opens the PR, owner reviews against the RFC, design QA runs the checklist, changelog entry ships with the release. Semver bump even for minor fixes.
Announce the merge in #design-system: who contributed, what changed, how to adopt. This is the story other squads will copy.
Today's output
First external RFC merged + changelog entry published
Day 93 · rfc-mergePage
Phase 4 · Adopt · Day 94Deliverable: Org-wide adoption update (draft for sponsor send)
Send the org-wide adoption update
If leadership does not hear the win, the next budget fight starts from zero.
Draft a short update for your sponsor to send org-wide (or post in the all-hands channel): pilot plus two squads live, metrics snapshot, how to get help (office hours, docs link, RFC process).
Include one quote from a squad lead about time saved or rework avoided. Qualitative proof matters alongside the spreadsheet.
End with honest next steps: what is in canon, what is deprecated, what is coming in the next release. No hype. Competent and specific wins budgets.
Today's output
Org-wide adoption update (draft for sponsor send)
Day 94 · org-commsPage
Phase 4 · Adopt · Day 95Deliverable: Enablement session recording + updated onboarding path in docs
Record the enablement session
Live training scales once. Record it so new hires stop pinging you in week one.
Run a 60-minute session: how to install the package, build a screen from docs, open an RFC, run design QA on a PR, and read the changelog. Demo one pattern recipe start to finish.
Record and upload. Link from your onboarding doc and the "build your first screen" guide. Slides are optional; a screen share of the real workflow is better.
Collect questions. Every question without a doc answer becomes a PR due within five days.
Today's output
Enablement session recording + updated onboarding path in docs
Day 95 · enablement-recordPage
Phase 4 · Adopt · Day 96Deliverable: Champion handover notes with per-squad next-month commitment
Hand governance to the champions
You are training replacements, not building dependency on yourself.
Meet each squad's champions for 30 minutes. Walk through: office hours format, RFC triage, PR checklist, how to escalate to the owner, when to pull in the sponsor.
Champions commit to one action per squad next month: a doc PR, a migration, or an RFC review. Write it down.
System owner remains the decision maker. Champions are the local glue. That split is how governance survives when you step back.
Today's output
Champion handover notes with per-squad next-month commitment
Day 96 · champion-handoverPage
Phase 4 · Adopt · Day 97Deliverable: Top 5 doc gaps closed + updated question log
Close the top doc gaps from migrations
Every migration surfaces doc debt. Fix the top five before day 100 or squads stop trusting the docs.
Pull the question log from office hours and migration retros. Rank gaps by frequency. Top five get closed this week: new pattern recipe, clearer install steps, missing component state, broken Code Connect link, outdated deprecation note.
Assign each gap to an owner (champion or system owner). Due date: before day 100 handover. No gap survives without a ticket.
Re-run the "new hire builds a screen" test after fixes. If they stall, the gap is still open.
People succeed, not systems. Measure whether squads ship faster and feel relief.
Fill every row in your metrics worksheet: day 1, day 50, day 100, and target. Button variants in production, Figma-code parity count, % pilot on system, squads shipped, rework hours, new-hire time to first screen, Slack "which component?" volume, hardcoded hex count, package imports, squad lead relief interviews.
Add one qualitative paragraph: what changed in how teams work, not just what changed in the repo. PMs referencing the system in specs is a metric too.
Flag metrics that missed target with a one-line root cause and a backlog item. Honest gaps beat polished fiction.
Phase 4 · Adopt · Day 99Deliverable: Sponsor review deck + written decision on next-quarter owner capacity
Present the sponsor review
Day 99 is for decisions, not demos. Show proof, name the owner, ask for what happens next quarter.
Build a 10-slide deck (or one strong doc): charter goals vs. outcomes, before/after metrics, two squad wins, governance now live, open risks. Lead with the number your sponsor cares about most.
Present to your sponsor for 30 minutes. Ask three questions: continue funding the owner role? expand to squad 4? align one system metric to next quarter's product OKR?
Leave with a written decision on owner capacity for the next 90 days. No decision is a decision to let the system drift.
Today's output
Sponsor review deck + written decision on next-quarter owner capacity
Day 99 · sponsor-reviewPage
Phase 4 · Adopt · Day 100Deliverable: Handover doc + 90-day backlog with top 3 prioritized items
Hand over and publish the 90-day backlog
Day 100 is not the finish line. It is the day the system survives without you.
Run a handover session with the named owner: roadmap, release cadence, office hours schedule, RFC triage rhythm, champion roster, sponsor escalation path. Transfer repo admin and calendar ownership.
Publish a 90-day backlog: Tier 3 gaps, deprecation targets, doc improvements, CI lint for hardcoded hex, squad 4 migration date. Prioritize top three items with owners and dates.
Phase 4 done when: two squads shipped on the system, one external RFC merged, owner can run the next quarter without you. Sign off with the sponsor. Then celebrate. You earned it.
From the field
The best day 100 I have seen ended with the owner running office hours alone while I watched from the chat. No questions stumped them. The docs answered it. That is the whole job.
Today's output
Handover doc + 90-day backlog with top 3 prioritized items
Day 100 · handover-90-day-backlogPage
Appendix A
Token architecture reference
Primitives hold raw values. Semantics hold intent. Components never reference primitives directly.
It doesn't have to be this way. Most failures are governance and sequencing, not talent.
Appendix F · Failures (6-10)Page
Appendix G
AI guardrails
AI scales output. Your system scales correctness.
Semantic token names machines can read
Component states and rules in Storybook
Do's/don'ts with rationale
Deprecation status visible to tools
I placed 1st at the Into Design Systems AI Hackathon 2026 with a Figma plugin that smart-swaps mismatched components. The system was the constraint that made AI reliable.
Appendix G · AIPage
Appendix H
Master 100-day checklist
✓
Checkpoint
Phase
Pilot Scorecard complete; host committed
1
Audit readout done; ROI estimated
1
Tokens + Tier 1-2 in Figma and code
2
npm package; pilot migrated
3
2+ squads shipped; RFC merged
4
Owner trained; day 100 metrics reviewed
4
Appendix H · ChecklistPage
Appendix I
Framework attributions and further reading
Ideas referenced in this book; prose is studio-authored.
Pilot Scorecard, Hot Potato, Measuring Spoon: Dan Mall, Design That Scales
Six-area model, token pipeline: Sarrah Vesselov, Building Design Systems (2nd ed.)
Three Times rule, Flow/Systems weeks, canon tiers: Mall; adapted for this playbook
Score each candidate surface 0-10. Sum and sort. Politics becomes sequence.
Factor
What you score
0-10
Components needed
How much of the system will this force you to build?
Patterns needed
Flows, not just atoms
Business value
Revenue, retention, exec visibility
Feasibility
Stack, access, timeline
Receptiveness
Will the host squad use what you extract?
Scope clarity
Bounded surface vs endless platform
Independence
Ship without blocking on other teams?
Publicity
Will leadership see the win?
Appendix J · Pilot ScorecardPage
Appendix K
Measuring Spoon and Hot Potato
Two rhythms that keep systems tied to product work.
Measuring Spoon Cycle: make product UI on a pilot, extract what repeats, reuse on the next product, repeat. Your 100-day plan is one long loop, not a big-bang library launch.
Hot Potato: designer and engineer in the same session. Pass work back and forth the same day. Engineer drives implementation early. Handoff barely exists.
Five pilot types: Raiders (refactor live UI), Facelift (visual refresh), Speed Run (greenfield fast), Surrogate (system team ships for reluctant squad), Perimeter (marketing, auth, settings).
Appendix K · PilotsPage
Appendix L
Connected delivery checklist
Infrastructure installs. Libraries get copied.
Published as npm install @your-org/design-system
Semver releases with changelog and migration notes
Deprecation warnings in code, not surprise Slack threads
Token change ripples to Storybook and apps on upgrade
At least one product app importing the package (not copy-pasted files)
Code is the territory
Figma is the map. Users touch production. Connected delivery makes design intent survive handoff.
Appendix L · ConnectedPage
What to do next
Run it yourself, or bring a senior pair of hands.
This guide is enough for a capable internal team with protected time. If you are moving faster, or you have stalled before, I help scaling product teams scale, govern, and ship.
16+ years in product design and design systems
Led Aurora at Nord/NordVPN; used by dozens of teams and millions of users
Design engineering in React, TypeScript, Tailwind
1st place, Into Design Systems AI Hackathon 2026
Design System Audit
€2,500-4,500
Your Phase 1 in 1-2 weeks. Fee credited toward a build.