TL;DR: A dashboard is modern when it ranks its data, keeps density legible, uses colour for meaning only, offers a real dark mode, survives a narrow screen and clears WCAG contrast. Open all five below and check: Plausible shows the cleanest KPI strip, Umami colours its deltas by meaning, Grafana handles density better than anything else, GoatCounter proves you can ship a dashboard with no metric cards at all, and Matomo demonstrates what equal-weight widgets cost. None of them need a login; every link was re-checked on 8 October 2026.
Every article about modern dashboard design shows you screenshots of Linear, Stripe and Vercel. Here's the problem with that: you can't open any of them. They're behind a login, so you cannot check whether the writer described them accurately, and neither could the writer's editor.
So this page does the opposite. Every dashboard below is published to anonymous visitors on purpose. Click the link, look at the thing, and disagree if the description is wrong. That's a lower bar for the writer and a much higher one for the claims.
Modern dashboard design in 2026 means arranging live data so one person sees the most important number first, then everything else in descending order of how often they act on it. It's a ranking problem solved with layout and type scale, not a coat of dark paint. The five below solve it in public, so you can check the work instead of trusting it.
If you want the underlying spec rather than the examples, the dashboard design patterns guide covers the 256px sidebar, the metric strip and the 12-column grid in detail. This page is the field trip.
What Are Six Checks for a Modern Dashboard?
A dashboard is modern when it ranks information, not when it is dark. Six checks, each one verifiable on the demos below:
- Data hierarchy. One number is visibly the number the page is about. Grafana's Kubernetes view sets "Running pods" and "Running containers" as oversized stat panels above a quieter grid; Matomo's dashboard FAQ lets you pick one to four columns and nothing in it ranks one widget over another.
- Density that stays legible. Grafana's dashboard JSON packs ten panels onto one screen and it still scans, because panel size tracks importance. ThoughtSpot's dashboard guide recommends five to seven primary KPIs; Plausible ships six, Umami five. The data density patterns article covers where the ceiling sits.
- Semantic colour. Colour means status, never decoration. Umami renders a rising bounce rate red because rising is bad. WCAG 1.4.1 adds the second half: colour must not be the only cue, which is why Umami pairs its colour with an arrow and Grafana's green tiles carry a label.
- Dark mode as an option, not a brand. Plausible follows the system theme by default and lets logged-in users pick system, dark or light. Grafana's default is dark. Matomo added a dark theme in 5.10 with light as the default and a "match browser" option. The dark-mode dashboard patterns explain what breaks when a theme is inverted rather than designed.
- Responsive layout. None of the five documents its mobile behaviour in the pages opened for this article, so judge it by narrowing the window on each demo. Grafana's docs do expose the mechanism: an auto grid layout with a minimum column width and a cap of ten columns. The responsive breakpoints strategy covers how to set those thresholds.
- An accessibility floor. WCAG 1.4.3 asks for 4.5:1 contrast on normal text and 3:1 on large text. The risky spots on every dashboard here are the small grey deltas on white (Plausible's "0%" bounce-rate change) and secondary labels on near-black (Grafana). Bake the ratio into tokens, as the design-system contrast guide shows.
The five examples against those checks
| Hierarchy | Density | Theme default | Theme options | Public access | |
|---|---|---|---|---|---|
| Plausible | six equal KPI cards; click one to drive the chart | low: one chart, two panels | follows the system theme | system / dark / light (logged-in users) | plausible.io's own live stats |
| Umami | five KPI cards with colour-coded deltas | medium: cards, one chart, panels | light in the demo (not documented in the sources opened) | not documented in the sources opened | public share URL, no login |
| Grafana | two oversized stat panels plus four status tiles | high: ten panels on one grid | dark | dark / light | anonymous access on Grafana Play |
| Matomo | none; a widget grid | high: one to four columns of widgets | light | light / dark / match browser (5.10+) | public demo site |
| GoatCounter | one row per page, each with a sparkline | medium: a single list | light in the demo (not documented in the sources opened) | not documented in the sources opened | public view, refreshed hourly |
Theme facts come from each product's documentation; layout facts come from the demos and their screenshots below. Where a vendor does not document something, the table says so rather than guessing.
How Does Plausible Do the KPI Strip With Restraint?
Plausible publishes its own live traffic, which is a confident thing to do. The layout is six metrics in a row, one chart, then two list panels side by side. That's the entire page. The documentation names the six: unique visitors, total visits, total pageviews, views per visit, bounce rate and visit duration, and clicking any of them swaps it into the graph.
What makes it work is what's missing. No sidebar. No filters cluttering the top. The date range and the filter control sit right-aligned on one line and stay out of the way until you want them. When your product does one thing, a sidebar is furniture nobody asked for.
The one to argue with: six metrics is one too many. Unique visitors and total visits sit next to each other and most people cannot tell you the difference without hovering the tooltip. Two adjacent numbers that need a tooltip to distinguish are one number and a footnote.
How Does Umami Make Colour Mean Something?
Umami runs five cards instead of six, and every card carries a percentage change in a coloured pill with an arrow. Plausible shows deltas too, but as small green figures beside the number; Umami gives them a tint and a size that you read without looking for them. A number alone tells you the state. A number with a delta tells you whether to care.
Note the colour discipline: the changes render red when they're bad, including the bounce rate that went up eight percent. Rising bounce rate is bad, so red is correct there even though the arrow points up. That's a small thing that a surprising number of dashboards get backwards by tying the colour to the arrow direction instead of to the meaning.
The sidebar is grouped into Traffic, Behavior and Growth, the same three groups Umami's share-URL settings let an owner expose. Seven, four and three items respectively, and the seven-item group is about the ceiling before people stop reading labels and start hunting by position. This demo is the public share link from Umami's own README, so the view is the one the vendor chose to show.
How Does Grafana Keep Dense Data Readable?
This is the densest layout of the five and still the easiest to scan, which is the whole trick. Two numbers are enormous: 92 running pods, 165 running containers. Four status tiles sit above them in solid green, labelled Kubelet, Kube State Metrics, cAdvisor and Node Exporter. Everything else is smaller and quieter.
Look at what the size difference is doing. Nobody has to decide what matters on this screen, because the type scale already decided. Most custom dashboards set every metric at the same size and then wonder why the page feels flat.
The honest detail worth pointing at: there's a panel reading "No data". It hasn't been hidden, and it hasn't been dressed up. Grafana's standard options even let a panel author choose what an empty value displays, with a hyphen as the default, so emptiness is a designed state rather than an accident. A monitoring tool that quietly hides an empty panel is lying to you about coverage.
Also worth stealing: the filter bar is a distinct strip between the breadcrumb and the panels, so the data source, cluster and instance selectors never get confused with the data itself.
What Does Matomo's Widget Grid Cost You?
Matomo is the counter-example, and that is a description rather than an insult. It's a genuinely powerful analytics product with a deep left navigation. But the dashboard itself is a widget grid, and a widget grid gives every box the same weight by construction. Matomo's own FAQ describes exactly that model: pick a one-to-four-column layout, add widgets from a menu, drag them by the crosshair to reorder.
Scan it and try to answer "how is the site doing?" in two seconds. You can't. A real-time visit log, a time-series chart, a world map and a channel table are all sitting at the same size with the same card treatment. The one headline figure, 13,250 visits with a green +8.8%, lives inside a "Visits Overview" widget two rows down. The information is all there. The ranking isn't.
There's a second problem the screenshot catches: one widget reads "No rows match the criteria" and takes up as much room as the chart next to it. Empty widgets that hold full-size slots are the tax you pay for a user-arrangeable grid.
A widget grid gives every box the same weight by construction, which means the layout expresses no opinion at all.
If you're building an arrangeable dashboard anyway, at least ship an opinionated default arrangement. Most people never rearrange anything.
Why Does GoatCounter Skip Metric Cards Entirely?
Here's the one that breaks the rules and gets away with it. No sidebar, no KPI cards, no hero chart. Just a date range, a filter box, and a list of pages where every row carries its own sparkline. Breakdowns for referrers, browsers, systems, locations and screen sizes sit further down as collapsible sections.
The layout is a single strong idea: the unit of interest is a page, so make the page the row and give each row its own trend line. Once you accept that, a summary chart at the top would be redundant.
Would this work for a product with five personas and a billing section? No. But it's a useful corrective, because the sidebar-plus-KPI-strip pattern gets applied reflexively to dashboards that have exactly one thing to show. If your dashboard answers one question, the honest layout might be a list.
Note the small honesty at the top right: "525 out of 871 visits shown" in the screenshot, a different pair of numbers whenever you open it. It tells you the view is truncated instead of quietly rounding. The header adds another: "The public view is updated once an hour", so nobody mistakes it for real time.
What Do the Five Dashboards Have in Common?
Strip away the styling and the same skeleton shows up in four of the five, GoatCounter excepted:
| Sidebar | KPI cards | Hero chart | Default theme | |
|---|---|---|---|---|
| Plausible | no | 6, small deltas | yes | follows system |
| Umami | yes, grouped | 5, coloured deltas | yes | light in the demo |
| Grafana | yes | 2 big + 4 tiles | no, panel grid | dark |
| Matomo | yes, deep | none | inside a widget | light |
| GoatCounter | no | none | per-row sparklines | light in the demo |
Two patterns fall out of that table. Products with one job skip the sidebar. Products people stare at for hours default to dark.
And the one thing every single one of them does, including Matomo: numbers are right-aligned or set in a tabular figure so the digits line up. It's the least glamorous item on this page and the one most custom dashboards miss.
What Looks Modern Versus What Is Modern?
Notice what the five share visually, because it's the opposite of what a trends listicle tells you. None of them lean on glassmorphism, heavy drop shadows, or gradient-filled KPI cards. Those read as dated within a year, and they fight the dense tables people actually work in all day.
The modern look in 2026 is almost invisible. A neutral surface. One accent colour for anything interactive. Semantic colour held back strictly for status: green for healthy, amber for warning, red for a breach. Tabular figures so digits line up. Type scale, not boxes, doing the hierarchy work. Grafana and Plausible both look plain on purpose, and both are years old, which tells you the modern part is restraint that holds up over time, not a visual season you have to chase.
Here's a quick test. Would this dashboard still look current if you stripped every colour except one? If yes, the structure is carrying it and it'll age well. If the whole thing collapses into grey mush, the style was doing work the layout should have done, and it'll look dated by the next redesign.
What actually dates a dashboard is rarely the palette. It's equal-weight widgets, a metric strip nobody curated, and spinners sitting where skeleton screens belong. Fix those three and a five-year-old dashboard still reads as modern. Chase the visual trend instead, and you're repainting the thing every single year.
Where Should You Go Next?
Run the five-minute judging method from the patterns guide linked above on each of these before you copy anything, then build the layout with the dashboard grid generator, pick each panel's visualisation with the chart-type guide, and check your colours against the WCAG contrast checker. Grafana's dark theme in particular is easy to imitate badly, and low-contrast grey on near-black is the usual way it goes wrong; the dark-mode patterns article linked in the six checks above walks through the contrast and elevation rules that keep a dark dashboard legible.
These are live third-party dashboards. They change, sometimes daily, so if a screenshot here no longer matches what you see, the link is the source of truth and the link deserves more trust than the screenshot.