DashboardAll dashboard templates / Learn
Learn23 Jul 2026 · 8 min read
Which chart types belong on a dashboard?
Eight chart types cover nearly every panel a dashboard needs: line, bar, area, donut, funnel, scatter, sparkline, and the plain table. Each answers one shape of question, and each fails in one predictable way. Pick the chart by the question, never by the space left over.
Charts are the middle layer of a dashboard. KPI cards give the headline numbers, tables hold the individual records, and charts sit between them, explaining why the cards moved and which records deserve a look. A dashboard with the wrong charts reads like a language translated word by word: every element is present and the meaning is gone.
| Chart type | Question it answers | Use it when | Failure mode |
|---|---|---|---|
| Line | How is the metric trending over time? | Continuous time series, 1 to 3 series per panel | Ten tangled series; dual axes implying fake causation |
| Bar / column | Which category is largest, and by how much? | Comparing discrete categories or stages | Used for time series where a line reads better |
| Stacked area | How does the total grow, and how do its parts shift? | Composition over time with few parts | Parts stacked on moving baselines become unreadable |
| Donut | How does one whole split into a few parts? | 2 to 5 parts, one moment in time | Angry pie with 12 slices; comparing two donuts |
| Funnel | How many records sit at each stage right now? | Stage counts as a snapshot, labeled as one | Read as stage-to-stage conversion, which it is not |
| Scatter | Are two variables related, and where are the outliers? | One dot per record, two numeric axes | Overplotting; trend lines fitted to noise |
| Sparkline | Which direction is this number drifting? | Beside a KPI card, as context for the value | Stripped of axes until the scale lies |
| Table | Which specific records need attention? | Lookups, ranked lists, anything a user opens | Forced to do a chart's job with 200 unsorted rows |
When does a line chart earn its place?
A line chart earns its place whenever the question is about direction over time. Revenue by month, signups by week, latency by hour: if the x-axis is continuous time and the reader wants trend, the line is the default. It earns the spot because the slope of a line encodes rate of change, and rate of change is what trend questions ask about.
Keep one to three series per panel. The ARR run rate panel in Northstar, a SaaS metrics dashboard, works because the reader is tracking one number's trajectory. Past three series, split the panel into two or use small multiples, because untangling a knot of colored lines costs more attention than the chart returns.
When do bars beat lines?
Bars beat lines whenever the categories are discrete and the question is comparison, not trend. Opportunities by stage, CAC by channel, outflow by expense category: the reader is ranking groups, and bar length is the most accurate encoding the eye has for ranking. Horizontal bars also solve the long-label problem, which is why stage names and channel names so often sit on the y-axis.
Real builds lean on this constantly. Relay, the free CRM dashboard template, uses bars for Opportunities by Stage and Opportunities by user, because both questions are rankings. Vault, the free finance dashboard, compares income against expenses as paired columns, because the reader is judging one number against its counterpart each month.
When is a donut acceptable?
A donut is acceptable for one job: showing how a single whole splits into two to five parts at a single moment. Traffic by device, MRR by plan, spend by top-level budget line. The reader wants the rough proportion, not precision, and the chart should admit that by staying small and carrying its values as labels.
The donut's failure mode is ambition. Twelve slices become a color wheel. Two donuts side by side ask the reader to compare angles across different circles, which the eye does badly; bars do that job in half the ink. If a template's centerpiece is a ring of donuts, its author chose decoration over answers.
When does a funnel tell the truth?
A funnel tells the truth about exactly one thing: how many records sit at each stage right now. It lies, by implication, about conversion between stages, because the 61 deals in proposal today did not descend from the 128 in qualification today. Those are different deals from different months. Stage-to-stage conversion must be measured on a cohort: take the deals created in one period and follow them forward.
This is the most common chart crime on CRM dashboards, and it is worth one more sentence because the picture is so persuasive. A snapshot funnel presented as conversion analytics flatters whoever filled the top recently and hides the stage where deals actually die. The full treatment is in our CRM dashboard examination; the short rule is that funnels are stage counts, and conversion belongs in a cohort table.
When is a table the best chart on the page?
A table is the best chart whenever the reader's next move is to open a specific record. Charts compress; tables retrieve. Recent transactions, expansion-ready accounts, upcoming bills, top rising queries: these panels exist so the reader can find a row and act on it, and no geometry beats a sorted, filterable row for that.
Tables also carry precision that charts cannot. A line chart says revenue dipped in May; the table says it dipped to $41,208 against a plan of $47,000. Serious dashboards pair the two: chart for the pattern, table for the exact values and the drill-down. Northstar's expansion-ready accounts panel and Beacon's keyword positions table in the Beacon marketing dashboard both work this way.
Which charts should a dashboard avoid?
Three chart families fail the dashboard test consistently. Gauges invent thresholds and spend a panel's worth of space on one number a KPI card shows better. Radar charts compare shapes nobody can rank, and the axis order changes the answer. Anything 3D distorts the values it claims to display, because the depth cue is decoration the data did not earn.
The broader rule: every panel should survive the question "what does the reader do when this moves?" A gauge that drifts into a red zone triggers nothing but unease. A bar chart of CAC by channel, like the one in Northstar's acquisition section, triggers a budget conversation. Choose the second kind.
Where can you see these chart types working?
Every template in the catalogue of 65 dashboard templates previews live, so you can inspect real chart choices per domain instead of trusting a taxonomy. Start with Relay for bars and funnels, Northstar for trend lines and cohort tables, and Vault for paired columns and record tables. The free files are on the free templates page; the rest are $6.99 each.
Browse all dashboard templates → · Free templates · Notes on choosing and using dashboard templates