Home Assistant
OPEN SOURCE
WEB
PROJECT PITCH
Design and evolve Home Assistant’s core experience so that a powerful, privacy‑first smart home platform feels approachable to both new and advanced users.
COMPANY
Open Home Foundation
ROLE & TEAM composition
Senior Product Designer
3x PM, 3x Product Designers, 4x Front-end
PERIOD
November 2024 - Currently
TOOLS
Lyssna
Google Forms
Code Editors *
Figma Make *
Figma
Protopie
STUDY INDEX
RESPONSIBILITIES
DELIVERABLES
A new token‑based UI kit (base + semantic) aligned with Web Awesome and MD2 evolution
Starting a new Figma UI kit and coordinated migration from the legacy one, accommodating for the newest Figma features for developer hand-off
Research artifacts (surveys, synthesized findings) that informed IA and automation improvements
Production-ready Figma designs linked directly to GitHub PRs
Contributing small PRs addressing UI-related issues
HOW I USED AI ✨
Rapid prototyping: Generated interactive code prototypes with tools like Google Antigravity and Cursor to validate new IA and automation concepts within hours instead of days
Design exploration: Used generative tools to produce draft layouts and flows from structured specs, then refined them to match Home Assistant’s design language
Data Analysis: Synthesized large volumes of user feedback to identify patterns and inform feature improvements
Technical Fluency: Analyzed project repositories to understand UI logic and constraints, ensuring better integration between design and code
STATUS
ONGOING
HOME ASSISTANT DEMO
A note on metrics and privacy …
Because Home Assistant is a privacy-first platform, we do not track user behavior or collect invasive telemetry data. This makes standard quantitative metrics unavailable. Instead, I validated my designs through qualitative research—relying on interviews, targeted surveys and opt-in community feedback to drive the discovery and delivery phases of the design process.
Empathize
To understand how people really use Home Assistant, the focus here was on listening to different kinds of users. Their stories, questions, and frustrations helped reveal where the experience broke down and what problems were most important to solve first.
Define
While I continued to identify and ship quick wins, like dashboard tweaks and general usability fixes, the Empathize phase revealed deeper structural issues. While the wider team tackled other roadmap goals, I joined forces with a few colleagues and narrowed our focus to two critical areas and defined their problem statements:
Home Assistants' information architecture 🗃️
PROJECT PROBLEM STATEMENT
Surface and clean up the metadata of users' homes, overhauling when and how data is displayed.
Our goal was to ensure the interface provides the full context needed to make decisions, making the system speak "human," not just "machine."
To get user-validated insights, I've conducted a user survey published on Reddit. Among other questions that provided valuable insights on how to fix this problem, the question:
"If you rename devices in Home Assistant, what are your reasons for doing so?" gave the most insight on the scale and nature of the problem.
"At least 90% of my naming convention has been so I can tell where the device is that I’m working with on a dashboard, automation, blueprint, etc."
Survey respondent #1
"My devices have their room as first part of their name because HA sadly doesn't have room 'namespaces'"
Survey respondent #2
"For example i got multiple devices named mainlight. But cause i know the room, i know witch is witch"
Survey respondent #3
Most users rename their devices regularly, with the largest group (33%) doing so 'Sometimes' and a combined 60% stating they do so either 'Sometimes' or 'Often' (view source).
PAIN
Users are forced to rely on complex, manual naming conventions (hacks) just to locate their devices, because the system fails to group them logically or display the information that is already there.
When building a smart home, users dutifully input context, like floors, areas, and names, during setup. However, Home Assistant fails to surface this rich metadata in the interface later. This 'data silence' forces users to invent and maintain complex naming strategies and manual hacks just to distinguish their devices and restore the context the system is hiding.
CAUSE
Home Assistant does not show the necessariy context (meta data) for a given item to understand it's connections or relation to the physical device in the users home
The system suffers from a data disconnect: rich device metadata exists in the 'Settings' page but is not utilized in the actual user experience. When users perform high-value tasks, such as viewing dashboards or building automations, the system fails to leverage the data it already possesses, presenting a flat, context-free list instead of the rich information available in the admin view.
IMPACT
After rolling out contextual info in over 7 key areas, users reported a better understanding of devices and their relations to other parts in their homes. This cut confusion, freed the team for evolve and work on the dashboarding experience, and continues improving the daily experience of using Home Assistant.
Automation engine 🤖
PROJECT PROBLEM STATEMENT
Align the automation engine’s learning curve with the evolving needs of our users.
Based on in-depth analysis of one-on-one interviews with users who span different types of knowledge and backgrounds, the results of those interviews provided us with a list of opportunities aimed at solving the problem statement. One of which was to transform the way users perceive and use Triggers and Conditions.
PAIN
Trigger and condition system is overly technical and entity-centric, making it difficult for users to search for real-world concepts or automate based on broad scopes like areas, floors, and labels.
Furthermore, creating common automations is currently non-intuitive because it requires users to possess deep knowledge of internal state attributes and data flows rather than simple, device-centric behaviors.
CAUSE
The nauture of using a device entity as a trigger or condition made the entry point technical and not aligned with a more natural way of thinking of a chain of events that the user wanted to achieve.
Research shown that users also tend to think of an automation in a more scenario and natural way, and less procedural and technical.
RELIEF
Create a set of triggers and conditions that are modeled on a scenario or intent way of thinking
This significantly lowered the barrier of entry for non-technical users.
IMPACT
After a 3-month rollout of UI improvements, including a dynamic sidebar that split automations into view/config modes and mobile-optimized patterns, usability scores improved significantly. Users now build and manage automations more intuitively with less cognitive load.
Ideate
With problems prioritized, I sketched out flows, automation patterns, and component updates using tools like Figma designs, Figma Make, and code experiments with agentic code editors - aimed at power users and newcomers alike, while working within our monthly release schedule. Narrowed ideas with PMs and engineers based on impact and fit. Documented what we'd tackle now, defer, or skip to stay focused.
Deliver
With concepts locked in, delivery meant shipping small, steady improvements to general usability, automations, and contextual info, also working tight with maintainers to avoid breaking setups or missing monthly releases. In open-source, this wasn't just handoff; I focused on clear documentation, quality checks, and guiding volunteer contributors to match design standards. This got real value to users faster while ensuring consistency.
dELIVER: TOKENS
Lead and introduced a two‑layer token system (base + semantic, for light and dark themes) aligned with the Web Awesome library, enabling theming and reducing hard‑coded values
IMPACT
Significantly reduced single-time usage of hard-coded values (DRY)
Made the work of developers simpler and reduced inconsistencies in the codebase
Laid groundwork for reusing tokens for other components
Introduced a systematized way of theming
DELIVER: UI KIT
Created and implemented a new UI kit based on new tokens and components
IMPACT
Introduction of new components improved the user-experience on desktop and mobile and helped in displaying information dense screens (i.e.: automation editor)
Using the "slot" method in Figma components removed the need to detach components from the library, making the design files more future-proof, and reduced maintenance for designers
DELIVER: CONTEXTUAL INFORMATION
Information architecture workstream
I designed and helped deliver contextual information across several key areas of the Home Assistant UI, providing the right details at the right moment to cut confusion.
DESIGN GOALS
1
Display other relationships to the related item after clicking on the breadcrumb
2
Display entity relationship to parent device, area and/or floor by the means of a breadcrumb component
In the Home Assistant UI a user can select to get the data or point to different types of so-called "targets". In versions preceding my work, these UI pattern had some major downfalls in terms of accessibility, readability and clarity of information.
DESIGN GOALS
1
Display item relationship to parent device, area and/or floor
Progressive disclosure of information (by displaying of number of targeted children entities)
More readable and accessible list of selected items
Mobile-friendly and responsive
Grouping by type
DELIVER: AUTOMATION EDITOR
Automation editor workstream
As noted in the define phase, the automation editor redesign also overhauled its visual layout - starting with a simplified list view for triggers, conditions, and actions (1).
Each row's configuration was then moved to a dynamic sidebar (2) that appears on demand, cutting information overload and cognitive load. Mobile-specific patterns, like a resizable bottom-sheet were also introduced to compensate for the lack of real estate on mobile to display all of the information.
DESIGN GOALS
1
Improve readability by visually differentiating key values of a row item in a chip-style component
Reduce the "matrioszka effect" of nesting elements in elements by introducing simplified row indentations (inspired by code and text editors)
2
Reduce cognitive load by moving the configuration to a dynamic sidebar
The second part of the work done by me has been devoted to creating a new type of triggers and conditions based on user intent, and not by a technical property of an entity.
To enable that the dialog for adding triggers and conditions has been completely re-worked from the ground up, to enable the user to browse by a tree view of the floor-area-device-entity hierarchy and show available triggers or conditions based on their currently selected child element.
DESIGN GOALS
1
Provide a "target first" approach of building automations rather than by intent
Show relevant triggers, conditions, or actions that the user can add for any given selected floor, area, device or entity and/or child elements
Show a structured and hierarchical view of the whole home and its assigned devices to promote discoverability, or identify discrepancies in the physical-to-digital model of the users home
2
Provide a "intent" first approach of building automations that current Home Assistant users are accustomed to, but expand it with simplified triggers and conditions
Challenges & Solutions
Transitioning into a large, legacy open‑source ecosystem meant designing for a global community rather than around it. The main challenge was modernizing the experience without blocking contributors or breaking existing setups.
How I've handled it
QUALITY CONTROL AT SCALE
With hundreds of unique contributors, maintaining UI consistency was difficult.
I shifted from being a sole designer to a 'Design Reviewer,' actively testing Pull Requests (PRs) in VSCode and GitHub. I provided detailed design feedback to contributors, ensuring their code met our quality standards before merging.
How I've handled it
TECHNICAL DEBT
The platform relied on an aging Material Design 2 implementation with accumulating technical debt.
I led the systematic evolution of the UI kit, introducing semantic tokens and upgrading components to improve accessibility (a11y) and minimize breaking changes in theming.
How I've handled it
THE COMPLEXITY CHALLANGE
Home Assistant is powerful but historically intimidating for new users.
To solve this, I launched user surveys to identify pain points for specific product areas and used AI tools (like Cursor and Figma Make) to rapidly prototype simplified flows. This allowed us to validate easier experiences without sacrificing the advanced control power users expect.
Key outcomes and impact of the team
Up next
marcinbauer.com © 2026
Last updated: 20.11.2025