This guide covers nine core practices for designing dashboards in Omni, drawn from patterns we see across customers. Treat it as a starting place — not a checklist. Each section includes Omni-specific guidance and links to the relevant docs.

Before you build

1. Know your audience

Every good dashboard starts with a person, not a dataset. Who’s actually going to use this? A sales rep glancing at pipeline between calls has different needs than a finance analyst running a quarterly review. Build for one of them, not both at once. Think about data fluency too. If your audience isn’t used to slicing and dicing, lean toward clear labels, obvious filters, and a single clear takeaway per tile. If they’re power users, give them room to explore — drill paths, controls, links to the underlying workbook. When in doubt, start simple and add complexity based on real feedback.

2. Start with a question, not a dataset

A lot of dashboards get built by staring at available data and asking “what could I show?” Flip it. Write down the two or three questions your audience is actually trying to answer, then build only what helps them answer those questions. Specific questions produce better dashboards. “How is revenue trending?” gives you a line chart and not much else. “Which sales reps are at risk of missing quota this quarter, and in which product lines?” gives you something worth building.

Informed design

3. Match the chart to the question

The chart type is part of the answer. A bar chart that should be a line chart doesn’t just look wrong — it actively obscures what you’re trying to say. Omni picks a reasonable default when you run a query, but it’s worth knowing the common patterns: Omni’s chart picker previews your data in each format as you switch types — try bar vs. line vs. area before committing.

4. Put the most important thing top-left

Eyes land in the upper-left and scan right then down — same as reading. Whatever your dashboard is really about should be there. Not a logo, not a date filter, not a decorative header. The number or chart that answers the main question. A few principles to keep in mind:

• Design for scanning patterns. People typically scan in an “F” or “Z” pattern. Place high-level KPIs along the top and your most important trend chart in the top-left quadrant.
• Establish visual hierarchy. Use size and font weight to signal importance. The most critical number should be the largest; supporting text (like “vs. Last Year”) should be smaller and lighter.
• Group related tiles. Related tiles should feel related — keep them physically close, use consistent sizing, and let whitespace do the separating. You rarely need borders or shading; spacing alone usually works better.

5. Limit colors and charts, then theme consistently

Three to five charts per dashboard is a good starting point. If you’re cramming in eight charts because they’re all “relevant,” you probably need two dashboards, not one big one. Same goes for color: every extra color is cognitive work for the viewer. Use color to encode meaning, not to make things look busy. A couple of principles that tend to hold up:

• Color is a functional alert, not decoration. Reserve high-contrast colors (red/green, bright orange) for performance or status. Don’t assign every category its own color — use shades of a single color or neutral tones to keep focus on actionable data.
• The 90/10 rule. Use a neutral palette (grays, light blues, muted tones) for 90% of dashboard elements. Reserve a single accent color — usually your brand’s primary hex — for the remaining 10% to draw the eye to what matters.

Theming your dashboards

Once the layout is right, a well-applied theme is what takes a dashboard from “functional” to something people actually want to look at. Omni’s document themes give you control over color palettes, fonts, tile backgrounds, border styles, spacing, and more — applied across an entire workbook at once. Manage themes from Settings > Themes > Document themes, where you can create multiple themes and set a default that applies instance-wide, or apply a theme per-document for one-off needs. Set an org-wide default so every new dashboard starts on-brand.

6. Emphasize what matters most

Axis labels, chart titles, legends — these only earn their space if they’re adding information the viewer doesn’t already have. If a label restates something obvious from the chart, remove it. If a legend has only one item, remove it. Less chrome means more signal. On any given chart, the axes carry the most visual weight. Put your most important metric there. Color, size, and shape are for secondary dimensions — context without competition. Whitespace matters too. Don’t be afraid of empty space, and use tabs to separate themes of questions. Whitespace acts as a visual buffer, letting the viewer distinguish between groups of information without getting overwhelmed.

Interactivity

7. Add interactivity to encourage exploration

A dashboard that lets viewers answer their own follow-up questions is one they’ll keep coming back to. The goal isn’t to anticipate every question — it’s to give people the right handles to pull so they can get there themselves. That said, interactivity that isn’t discoverable may as well not exist. If you add a control, make it obvious. Label filters with action-oriented text. Proximity is key — if a control only impacts one chart, place it where that’s clear. For complex controls, drop a short Markdown tile nearby that explains what they do.

Dashboard controls

Beyond standard filters, Omni’s dashboard controls give viewers structured ways to interact with data without leaving the dashboard. Controls are added in edit mode via Add Control and can be connected to one or many tiles at once.

• Field switcher — Swap one field for another. Let viewers toggle between Revenue, Orders, and Margin on the same chart.
• Multi-field picker — Select zero or more fields to show in a tile. Especially useful for table visualizations.
• Time frame switcher — Change date granularity (day / week / month / quarter) without editing the underlying query.
• Period over period — Compare a metric across two time periods, like this month vs. last month.
• Parent control — Coordinate multiple controls at once. Useful when tiles share different date fields that should move together.
• Dynamic text tiles — Reference the current control or filter value in a Markdown tile using {{controls.<id>.value}} syntax.

8. Build for self-service, not just self-sufficiency

There’s a difference between a dashboard that answers questions and one that helps people answer their own. The first requires someone on the data team to anticipate everything. The second gives viewers room to dig in without breaking anything. The goal isn’t to expose everything — it’s to expose the right things. Well-scoped topics, default filters, and clear drill paths cover 80% of questions. For the other 20%, the AI Dashboard Assistant means viewers can just ask in plain English, right on the dashboard, without pinging the data team.

Refining your dashboard

9. Complete the story with AI, summaries, and tooltips

A well-built dashboard still leaves interpretation work on the table. What does that dip in March mean? Is that trend significant? Tooltips, labels, and written context help — but Omni’s AI features can take this further, doing the interpretation work automatically so viewers spend less time puzzling over charts and more time acting on them.

Tooltips and labels

Tooltips show up at exactly the moment someone wants more context — make that moment count. Lead with the thing they’re actually asking about (the metric, the dimension), not field names that mean nothing outside the data model. And if you can answer the question with a mark label instead of a hover, do that — it’s one less interaction required. A good rule of thumb: if a viewer can understand the key insight without touching the chart at all, the chart is working. Interactivity is a bonus, not a crutch.

AI Summary tiles

Drop an AI Summary tile onto a dashboard and Blobby will write a plain-language interpretation of what the data shows — automatically. It updates as filters change, so an executive filtering to their region gets a summary specific to their view, not a generic one written when the dashboard was built.

AI Dashboard Assistant

The AI Dashboard Assistant puts Blobby directly inside every published dashboard. Viewers can ask follow-up questions in plain English, request a different cut of the data, or ask what a number means — without opening a workbook or pinging the data team. It’s available via the icon in the bottom-right corner once enabled.

Test for usability, then share

Before publishing, walk through your dashboard as if you’ve never seen it. Can you answer the core question within 30 seconds? Do the filters and controls work as expected? Are there any blank states that might confuse a viewer? Then test with a real member of your audience — watch where they click first, what confuses them, and what questions they ask that the dashboard doesn’t yet answer.

Getting it in front of people

A dashboard nobody sees is a dashboard that doesn’t exist. Omni gives you several ways to close that last mile — from scheduled pushes to embedded experiences to on-demand answers in Slack.

• Scheduled deliveries — Email, Slack, or Google Sheets on a custom cadence.
• Alerts — Notify stakeholders when a KPI crosses a threshold.
• Embedded dashboards — Signed iframe in any product or internal tool.
• PDF / CSV export — One-click export for stakeholders who prefer offline.
• Omni Slack Agent — Answer data questions without opening Omni.

A note on performance

Omni is designed to be fast out of the box. Because queries are defined and aggregated in the shared model and pushed down to your warehouse, most dashboards load quickly without optimization work. If you do hit slow load times, Omni’s Dashboard performance profiler helps diagnose the cause. Open the dashboard, click View > Performance profile, and the profiler opens as a side panel showing query timing, caching behavior, and result sizes for each tile — with specific guidance on where to focus.