Dark Mode System Review
Challenging a set design direction — with research, benchmarking, and a willingness to push back.
My role
Product Designer
Timeline
FY25
Scope
Design system · Dark mode · Accessibility
Outcome
Partial direction change · Org-wide design system rethink
Overview
One of my habits is constant benchmarking. I use Mobbin, study the products I use every day, and compare our work against what the industry is actually shipping.
So when Dell's dark mode concept looked off to me — color, hierarchy, readability, accessibility — I didn't just trust the feeling. I checked it against IBM Carbon, Azure, Figma, and ChatGPT to see if I was right. I was. I ran it by my manager and a few teammates, and they saw the same problems.
I raised it early, before these patterns could spread across products, and built a research-backed case: a formal heuristic evaluation, outside benchmarking, input from other Dell teams who'd already solved this, and a review from Dell's own dark mode specialist. It became the first formal challenge to our design system's direction — and it helped start a wider effort to modernize how Dell's products look and feel.
What I noticed
The dark mode was built on a saturated slate-blue background.
It looked branded — but underneath, it had real problems.
Image: Initial dark mode concept
The team was applying light-mode logic to a dark screen. The bright accent colors from light mode — the same reds, blues, and greens — were carried over without adjusting them. Bold outlines were drawn around everything to show grouping. White outlines were added to every status icon, because that color combination technically passed the contrast test.
That last one shows the core problem best. A dark background, a white outline, and a bright accent color — three layers of color competing where one would do. It passes a WCAG test. It still doesn't look right.
Observed issues
System status is hard to read: Status colors get lost against the saturated blue.
Doesn't match user expectations: Enterprise users expect dark mode to feel calm. Dell's felt loud.
Too much visual noise: High saturation works against a clean, minimal look.
Slower to scan: Heavy color and outlines make it harder to find things quickly.
Accessibility beyond the checkbox: Passing a contrast test isn't the same as being usable.
Harder on the eyes: A dominant, saturated background adds strain over long sessions.
At scale, these issues would pile up across products — making the system harder to use, harder to maintain, and harder to keep consistent.
How I checked
Outside benchmarking. I compared Dell's concept against IBM Carbon, Azure, Figma, and ChatGPT. Every one uses a neutral gray base, softer accent colors made for dark screens, and few outlines. They're calm and easy to scan. Dell's concept wasn't.
Inside validation. I talked to other Dell teams who had already shipped dark mode. Their approach was close to mine — neutral base, softer accents, minimal outlines. So the concept wasn't just out of step with the industry. It was out of step with other Dell teams too.
Expert input. I shared my work with Dell's internal dark mode specialist. Their view lined up with mine, and connected to design council and North Star efforts already underway.
A formal heuristic evaluation. On my own time, outside my normal project work, I wrote a full evaluation. I used Nielsen Norman's 10 usability heuristics as the frame, and referenced WCAG, research from PubMed, ACM, and ArXiv, and design systems from Apple HIG, IBM Carbon, and Google Material. I used ChatGPT to help structure the document, but kept it neutral on purpose, so the argument wouldn't be one-sided.
My Role
Found the problems myself, through benchmarking — not asked, not assigned.
Raised them early, before the patterns could spread across products.
Wrote a full heuristic evaluation, with citations across Nielsen Norman, WCAG, PubMed, ACM, ArXiv, Apple HIG, IBM Carbon, and Google Material.
Reviewed another Dell team's shipped dark mode to check my approach against theirs.
Shared my work with Dell's dark mode specialist, and connected it to existing design council and North Star efforts.
Brought the findings to my manager, who backed them and raised them to leadership.
Proposed a realistic fix that worked within the existing system.
Kept advocating through pushback — staying on the evidence, not the conflict.
What I proposed
I wasn't asking anyone to throw away their work. A full redesign wasn't realistic.
So I proposed a focused set of possible direction:
Switch the background from slate to neutral gray. Use the existing neutral palette as the base — just add two darker tones for depth. This fixes the core problem without rebuilding the color system.
Soften the accent colors for dark mode. The vivid light-mode reds, blues, and greens are too intense on dark. Lightening and desaturating them a little gives the right contrast without the noise. This is standard — Material, Apple HIG, and IBM Carbon all do it this way.
Remove the outlines from status icons. Dark background + white outline + vivid accent = three layers fighting for attention. Drop the outline, use a softer accent, and you get more clarity with fewer parts.
How it played out
The team pushed back — not on the evidence, but on the fact that the challenge came from outside their group. At the time, teams worked fairly independently, and a research-backed critique that cut across those lines was hard to place. I was eventually moved off the project.
But the argument outlasted my involvement. After I shared the evaluation, the team moved away from the original slate toward a darker, less saturated version. Not the full neutral gray I'd recommended — but a real step in the right direction.
And the work traveled upward. My manager raised the concerns to leadership using my evaluation as the foundation, and that fed into a broader agreement that the design system needed a rethink.
The team moved. Not all the way. But they moved. And the conversation kept going after I left it.
Impact
Influenced a real change to the dark mode direction — from saturated slate toward a calmer, darker base.
Raised the concern from a single feature up to a system-level conversation.
Helped prevent flawed patterns from scaling across products.
Fed into leadership alignment that the design system needed modernizing — helping kick off a wider effort.
What I learned
Passing an accessibility test and designing something usable are not the same thing. Compliance is the floor, not the goal. The real bar is what the best products in your category actually do — and the only way to know that is to keep looking. Not just when there's a problem, but as a habit.
Benchmarking all the time is what gave me the confidence to push back — and what made the case hard to dismiss, even when it was uncomfortable to hear.
The team moved. Not all the way. But they moved. And the conversation is still going.