What a dashboard is for
The purpose of a marketing dashboard is to let a person who was not involved in the data answer a question and act on it without asking anyone. That is a higher bar than "shows the numbers". It means the page has to know what the person is deciding — whether to move budget, whether to escalate a drop, whether the month is on track — and show exactly the evidence that decision needs and nothing that competes with it.
Monitoring and explaining are different jobs. A monitoring page is opened every day, carries the same charts and should be readable in the time it takes a coffee to cool. An explanatory report is built once for a review, carries narrative and annotations, and is read start to finish. Looker Studio builds both well, and the commonest failure is a page that tries to be both: a dense monitoring grid with paragraphs of commentary, or a narrative with so many charts the story is lost.
The question every page must answer comes from the KPIA metric tied directly to an objective. tree that the team agreed when it planned measurement. A page sits at one node of that tree: the headline metric, the inputs that move it and the breakdowns that show where it moved. The SEO KPIs and reporting guide shows the same discipline applied to one channel; this page is about the design of any decision dashboard, with Looker Studio as the medium.
One question per page: structure and hierarchy
Start with the question written as a sentence at the top of the page, in words a non-specialist understands: "Is paid search delivering leads at the agreed cost this month?" Everything below it answers that sentence. The structure that works for most monitoring pages reads top to bottom in order of importance:
- The answer. One or two scorecards: the headline metric, its target or comparison and the gap. A person who reads only this row leaves informed.
- The trend. A single time series of the headline metric over a period long enough to show whether this month is unusual.
- The drivers. The handful of inputs that move the headline — spend, clicks, conversion rate, average value — as small scorecards or a compact table, so a reader can see which one explains the gap.
- The breakdown. One dimension, chosen because it is the one the viewer can act on: campaign, market, device, landing page.
- The detail. A table for people who need rows, at the bottom, where it does not compete with the answer.
Hierarchy is enforced by size and position, not by colour. The headline is largest and top-left, because that is where reading starts; supporting charts are smaller; the detail table is full width but low on the page. Keep a consistent grid, align edges, and leave space between groups; crowding is the most common reason a page reads as noise. A second question gets a second page, linked from the first with Looker Studio's page navigation, rather than a second column on the same canvas.
The wireframe tool below lets you lay out these five zones for a page, drag the blocks to match your own question and export the sketch to share with whoever will build the report. Before the build starts, a dashboard specification records the audience, the question each page answers, the metrics with their definitions, and the sources, refresh, access and owner, so the sketch and the agreed requirements travel together.
Dashboard wireframe
Choosing chart types
The chart type should follow the comparison the viewer needs to make, not the chart's visual appeal. Looker Studio offers scorecards, time series, bar and column charts, tables with heat-map and bar options, pie and donut charts, geo charts, scatter charts, bullet charts and more; most monitoring pages need only the first four.
| The viewer needs to see… | Use | Avoid | Why |
|---|---|---|---|
| A single value against a target or prior period | Scorecard with comparison | Gauge | A scorecard states the gap in numbers; a gauge hides it in an angle |
| Change over time | Time series (line) | Column chart with many periods | Lines show shape; many columns become a wall |
| Rank among categories | Horizontal bar, sorted | Pie or donut | Length is read precisely; angle and area are not |
| Composition of a whole, few parts | Stacked bar or a table with share column | Pie with many slices | Beyond a few parts, slices are unreadable |
| Two measures across many items | Scatter | Dual-axis line | Dual axes invite false correlation |
| Many values the viewer will scan or look up | Table with conditional formatting | Dozens of scorecards | A table gives rows; scorecards give isolated numbers |
| Performance by place | Geo chart, only when geography is the decision | Geo chart as decoration | A map is slow to read when the question is not spatial |
| Progress toward a goal with ranges | Bullet chart | Multi-ring gauges | Bullet charts show target, actual and ranges in one line |
Two habits make chart choice easier. Sort every bar chart by the metric unless the dimension has a natural order such as months. And resist the dual-axis line: when a viewer needs two measures over time, two small single-axis charts stacked vertically are read correctly far more often. The data visualisation hub covers the perceptual reasons behind these rules in more depth.
Targets, comparisons and context
A number alone is trivia. A number against something becomes information: against a target, against last month, against the same period last year, against the rest of the account. Every headline scorecard needs one of these, and the choice should be deliberate. Targets are best when the team has a plan; prior period is the fallback for metrics without a plan; year-on-yearComparing a period with the same period a year earlier, aligned by weekday and by holidays that move, so that the seasonal pattern drops out of the change. is for seasonal businesses where last month is misleading.
Targets have to live somewhere Looker Studio can read. The practical pattern is a small table — a sheet or a warehouse table — keyed by month and channel with the agreed CPACost of each conversion., ROASRevenue generated per unit of ad spend. or lead volume, connected as a data source and blended into the headline chart so the gap is on the page. Looker Studio's scorecard comparison, its reference lines on time series and its conditional formatting in tables then show the gap without a calculation in the viewer's head. The calculated fields tutorial covers writing the gap and the status flag as fields rather than typing them in.
Context also means stating the scope. A scorecard labelled "Leads" is ambiguous; one labelled "Qualified leads (CRM, month to date)" is not. Where a figure is partial — month to date, a sample, a platform's own attribution — the label says so. A comparison that is not like for like, such as a month to date against a full previous month, is worse than no comparison; use the same-period comparison that Looker Studio provides.
A paid search page that answers its question
A home insurance brand's paid search page opens with the sentence "Is paid search delivering quotes at the agreed cost this month?" The headline scorecards show 1,840 quotes month to date against a pro-rated target of 1,900, and a cost per quote of £52 against a target of £45, with the gap in red. The trend chart shows cost per quote rising since the ninth of the month. The drivers row shows spend flat, clicks up 6% and quote rate down from 4.1% to 3.4%. The breakdown bar chart, sorted by cost per quote, puts a new generic campaign at the top at £71. A reader who has never seen the account can say, in under a minute, that the gap is a conversion-rate problem concentrated in one new campaign, and the detail table at the bottom names its landing page.
Filters, date controls and defaults
Controls let one page serve several viewers, and defaults decide what the page says to the viewer who touches nothing, which is most of them. The date range control should default to the period the page's question is about — month to date for a pacing page, the last complete week for a weekly review — with the comparison period set to match. A page that opens on an arbitrary recent window shows nothing anyone asked for.
Filter controls should be few, placed in a consistent position at the top of the page, and limited to dimensions the viewer actually needs to cut by: market, channel, campaign group. Each control is itself a query, so a row of six drop-downs slows the page and intimidates the viewer. Report-level controls apply to every page and suit the dimension everyone filters by; page-level controls suit the dimension specific to that page's question. Where a control must apply to some charts and not others, Looker Studio's groups let you scope it, and the exceptions should be labelled, because a chart that ignores the filter looks like a bug.
Defaults extend to the state of the charts: sort order, the number of rows a table shows, which optional metrics are visible. Set them for the common case and leave the toggles for the exceptions. Interaction features such as cross-filtering — clicking a bar to filter the page — are useful on an exploration page and confusing on a monitoring page where the viewer expects stability; enable them deliberately rather than by default.
Annotations and definitions on the page
Numbers travel without their definitions, and a dashboard that leaves definitions in a separate document guarantees the argument will happen in the meeting. Put the short definition where the number is. A scorecard label can carry the scope; a text box under the headline row can state in one line what counts as a lead and which attribution the revenue uses; a tooltip or a small "about this page" panel can link to the full metrics layer entry for anyone who wants the long version. Definitions on the page work best when the team has learned to look for them, which the guide to building data literacy in a marketing team covers through sessions built on the team's own dashboards.
Annotations do a different job: they explain a visible change. A line that dropped on a known date — tracking broke, a campaign paused, a site launched — should carry a note saying so, or every viewer will rediscover the drop and ask. Looker Studio has no native annotation layer on time series, so the practical options are a text box positioned beside the chart, a reference line with a label, or an events table blended in from a sheet that the team keeps. The sheet approach scales best, because the same events appear on every page that shows the period. The measurement change log tutorial sets that sheet up as one log for the whole measurement stack, with the fields a dashboard needs to filter it.
Definitions and annotations also need owners. The dashboard governance guide covers who maintains them, how changes are recorded and how a report is retired; here the point is that a page without a stated owner and a last-reviewed date is a page nobody will trust for long. Before a new or rebuilt dashboard is shared, the dashboard launch QA checklist checks the labels, defaults and notes page alongside the sources and totals.
Accessibility and colour
A dashboard is read on laptops, projectors and phones, by people with varied colour perception, often in a hurry. Design for the worst of those conditions. Text must be large enough to read from the back of a meeting room; a scorecard whose value is legible and whose label is not has failed half its job. Contrast between text and background must be high; the subtle greys that look refined on a designer's screen vanish under projector light.
Colour should carry meaning sparingly and never alone. One accent for the headline metric, one for its comparison, and a neutral for everything else is enough; a page where every series has its own saturated colour has no hierarchy. Status — on target, at risk, off target — should be shown with a label or an icon as well as a colour, because red and green are the pair most often confused. Where series must be distinguished, vary lightness and use direct labels rather than relying on a legend that forces the eye to travel.
Tables are the accessibility fallback for every chart: a viewer using a screen reader, or someone who needs the exact values, should be able to reach a table that carries the same data. Keep the reading order sensible — Looker Studio renders elements in the order they were added — and give each chart a title that says what it shows, not just the metric name. Finally, test the page on a phone; Looker Studio's responsive options are limited, and a page that only works at full width will be read at full width by nobody on the way to a meeting.
Common mistakes
- A page with no question. Charts chosen because the data was available, not because a decision needs them.
- Headline metrics without a target or comparison. Numbers nobody can judge.
- The dual-axis line. Two scales, one chart, and a correlation that is not there.
- Pie charts for rank. Slices cannot be compared by eye; sorted bars can.
- Defaults that show the wrong period. The viewer who touches nothing sees a window nobody asked about.
- Six filter controls in a row. Slow to load, intimidating to use, and most of them never touched.
- Definitions in another document. The argument about what counts as a lead happens in the meeting instead of on the page.
- Colour as the only status signal. Red and green alone exclude a share of every audience.
The Looker Studio hub links the rest of the series; the dashboard governance guide covers keeping a well-designed page trustworthy after launch.
