Design Systems · Step 1 · 5 min read
The interface inventory
Before you design a system, collect what already exists: one screenshot of every unique button, field, colour and message across your products. How Brad Frost's interface inventory works, how Tidewater ran one across two products, and how the result becomes the first draft of Harbour and the argument for it.
The interface inventory is the cheapest, most convincing first step of any design system. You don't start by drawing new components. You start by collecting the ones you have, side by side, until everyone in the room can see the eleven buttons for themselves. It takes a few hours, needs no tools beyond screenshots, and does two jobs at once: it shows you what the system needs, and it shows everyone else why.
What you'll be able to do after reading:
- Explain what an interface inventory is and why it persuades
- Run one across one or more products, step by step
- Turn the result into a first list of tokens and components
- Avoid the traps that make an inventory useless
1. What it is
Brad Frost defined it in 2013: "An interface inventory is a comprehensive collection of the bits and pieces that make up your interface" (Frost, Interface Inventory, 2013). You go through the product and take a screenshot of each unique version of each part, grouped by category: all the buttons together, all the text fields together, all the colours together.
The point is the grouping. On a single screen, a button looks fine. Next to its ten cousins, the drift is obvious. Frost's phrase is that the inventory shows an interface "as a system of components rather than a series of discrete pages".
2. Why it persuades
Frost is blunt: "You don't need to be a designer to recognize that having 37 unique button styles probably isn't a good idea" (Frost, Atomic Design, chapter 4). An inventory can be shown to anyone, a developer, a product manager, a director, and they get it in seconds. No slide about "consistency" does that.
It also keeps the conversation honest. You aren't saying the product feels messy; you're showing fourteen blues.
3. How to run one
Frost's process has five steps, from chapter 4 of Atomic Design:
- Round up the troops. Invite people from every discipline: design, development, product, content, QA. Frost says it's essential "to get all members of the team to experience the pain of an inconsistent UI".
- Prepare for screenshotting. Agree on one shared tool (Frost suggests Google Slides; a Figma or FigJam file works the same) and one page or slide per category.
- Screenshot. Each person takes some categories. Capture one instance of each unique version, not every occurrence. Frost suggests between 30 and 90 minutes for the first pass.
- Present findings. Each person shows their category to the group, for five or ten minutes.
- Regroup and agree next steps. Decide what to keep, what to merge, and who owns the follow-up.
His suggested categories include global elements, navigation, image types, icons, forms, buttons, headings, blocks, lists, interactive components, media, third-party components, advertising, messaging, colours and animation, and he adds that they "aren't set in stone". Pick the ones your products have.
4. Tidewater's inventory
Tidewater ran the inventory across both products in one afternoon, with two designers, two developers and the product manager of the console. Here's the summary sheet they ended with:
| Category | Booking app | Staff console | Different versions | Harbour keeps | Note |
|---|---|---|---|---|---|
| Buttons | 9 | 8 | 11 | 4: primary, secondary, danger, ghost | Two fail text contrast |
| Blues in use | 8 | 9 | 14 | 3: blue.400, blue.600, blue.700 | Most differ by a few hex digits |
| Spacing values | 15 | 19 | 23 | 8: space.1 to space.8 | Includes 13, 15 and 18 px |
| Text fields | 4 | 5 | 7 | 1, with states | Three ways to show an error |
| Date pickers | 2 | 1 | 3 | 1 | One is a third-party widget |
| Delay and error messages | 3 | 4 | 6 | 1 pattern, 3 levels | "Delayed" is red in one, orange in the other |
| Icon sets | 1 | 2 | 2 | 1 | Two arrow styles side by side |
Three things to notice:
- The numbers are per unique version, not per screen. "Buttons: 11" means eleven different-looking buttons, not eleven screens.
- The "Harbour keeps" column is the first draft of the system. Fourteen blues become three primitives; twenty-three spacing values become an eight-step scale; eleven buttons become four variants of one component. That's Steps 2 and 3 of this course, in one column.
- The notes are the arguments. "Two fail text contrast" and "'Delayed' is red in one, orange in the other" are the lines that get a system funded, because they show risk, not taste.
5. From inventory to backlog
After the session, turn each row into work:
- Colours, type, spacing, radii → the first tokens. Merge near-duplicates; give each kept value a name.
- Parts that appear in both products → the first components, most used first.
- Parts that appear in one product only → leave them in that product for now. They may become recipes later (see Anatomy of a design system).
- Third-party widgets → decide: theme them with your tokens, replace them, or document them as exceptions.
Keep the screenshots. When someone asks in six months why the system has four button variants and not eleven, the inventory is the answer.
6. Common mistakes
- Screenshotting every occurrence instead of every unique version, and drowning in hundreds of images.
- Doing it alone. The inventory persuades because people see it happen.
- Turning the presentation into blame. The drift came from reasonable choices made apart; say so.
- Stopping at the screenshots. The value is in the "keep" column and the backlog.
- Inventorying only visuals. Error messages, empty states and wording drift too.
7. Vocabulary
| Term | Meaning |
|---|---|
| Interface inventory | A collection of every unique version of each interface part, grouped by category |
| Category | A family of parts: buttons, forms, colours, messages |
| Unique version | A variant that looks or behaves differently, captured once |
| Near-duplicate | Two values so close nobody chose the difference, like two blues a few hex digits apart |
| Backlog | The ordered list of work that comes out of the inventory |
8. Try it (no tools needed)
- Pick one category, buttons, and one product.
- Go through ten screens. Each time you see a button that looks different from the ones you've seen, sketch it or screenshot it.
- Count the unique versions. Then decide how many the product actually needs.
Reflection questions: Which difference surprised you most? Who on your team would be most convinced by seeing them side by side?
9. If you already practise
- Run a smaller inventory every year, or before a big release. It's the best measure of drift you have.
- Inventory inside the system too: components that overlap, tokens nobody uses.
- Add a column for accessibility (contrast, focus, label) to each row. It turns an inventory into an audit.
- Practise the whole method in the lab Run an interface inventory.
Checklist
Learn more
- Brad Frost: Interface Inventory (2013) · Atomic Design, chapter 4, The Atomic Workflow
- W3C: Understanding 1.4.3, Contrast (minimum), for the accessibility column
Books, courses and videos to go deeper
- Atomic DesignBrad Frost2016, free online · chapter 4 (Amazon)
- Design SystemsAlla KholmatovaBook · Smashing Media, 2017 (Amazon)
Book links go to Amazon. As an Amazon Associate I earn from qualifying purchases, at no extra cost to you.
Related
Run an interface inventory · What a design system solves · Anatomy of a design system · Token tiers · Heuristic evaluation · Audit the images in your product