Two kinds of page, two sets of rules.
A presentation page makes one point. A dashboard page lets the reader find their own. Most bad report pages are one trying to be the other.
Decide which page you're building before you place a visual.
Cole Nussbaumer Knaflic splits data work into two kinds in Storytelling with Data. Exploratory analysis is how you find something: she compares it to opening a hundred oysters to find two pearls. Explanatory communication is showing the pearls. A presentation page is explanatory. A dashboard is a tool for the reader to do their own exploring.
Every later choice follows from this one: the title, how many visuals, whether there are slicers, where colour goes. Ask these questions first:
| Question | If yes, build a |
|---|---|
| Do you already know the one thing the reader should take away? | Presentation page |
| Will someone present it, or will it be read once to make a decision? | Presentation page |
| Will readers come back every week or month to check where things stand? | Dashboard page |
| Will they filter, compare groups and look up their own part of the business? | Dashboard page |
| Would your headline stop being true if someone changed a slicer? | Presentation page (a fixed view), or a dashboard with a dynamic title |
A presentation page argues one point. A dashboard page answers many questions, none of them chosen in advance.
One report can hold both. Put the presentation pages first, in the order of the argument, and the dashboard pages behind them for readers who want to check the numbers or find their own department.
One point, one main visual, and a title that says it.
This is the shape consulting firms such as McKinsey, BCG and Bain use for every slide, and the one Gene Zelazny teaches in Say It with Charts: decide the message first, then pick the one chart that proves it.
- 1. Action title. The finding, its number and its comparison, in one sentence at the top left.
- 2. One visual, large. Coloured to match the title: Production in the accent, Sales in a lighter shade, the rest grey.
- 3. Callout. The "so what": what the chart means for the decision.
What's missing is part of the design:
- No slicers. The view is fixed, so the title stays true. A filter would let the reader change the evidence under the claim.
- No card row. If one number matters, it goes in the title or the callout, or as big text (see one number as text).
- No second visual unless it proves the same claim another way. A chart that makes a different point needs its own page.
A good test: read only the title and the callout aloud. If they make sense on their own, the page works.
Built for exploring: cards first, a grid, filters on the edge.
A dashboard has no single message. Readers come with their own questions: how many left last year, which department, which hiring source. The layout's job is to make every answer easy to find in the same place every time.
- 1. Card row, top left. The three-second answer, each number with a comparison.
- 2. Four to six visuals on a grid. Equal sizes and equal gaps, read left to right, top to bottom, in a Z.
- 3. Slicers grouped on one edge. Same place on every page, synced, never in the top-left corner.
- 4. Descriptive or question titles. They stay true whatever the reader filters.
No callouts. A typed sentence such as "Production drives our turnover" is true for one view only. The first time a reader filters to Sales, it's wrong on screen. If a dashboard needs words, build them from a measure so they change with the filters, as in dynamic titles.
Colour works differently here too. With no single point to make, bars stay neutral and colour flags exceptions only: the department above the company line, the card that is off its comparison.
A presentation page built like a dashboard makes no point at all.
The most common mistake is to put everything you found on the page you present. Every visual is accurate, and the reader still can't say what the page is about.
- The title names a topic. No finding is stated anywhere on the page.
- Six charts at the same size and weight: nothing says which one matters.
- Slicers invite the audience to change the view while you're talking.
- The finding that was hidden in the top-left chart is now the title.
- The other five charts aren't deleted. They move to a dashboard page behind this one.
If the title needs "and", it's two pages.
Readers hold only a few things in mind at once. A page that makes two claims splits their attention, and each chart has to compete for the title. Split it, and each page gets a full-width visual and its own title.
- The title joins two unrelated findings with "and".
- Neither chart is coloured to match the whole title, because no chart can be.
- Each title states one finding with its number.
- Read the two titles in order and you have the story, which is the consulting "storyline" test.
A dashboard without a grid makes the reader hunt.
Dashboards tend to grow one visual at a time, each dropped wherever there was room. The result has no starting point, filters in three places, and sizes that say nothing about importance.
- Three slicers in three places, one of them in the top-left corner.
- No cards, so there is no three-second answer.
- The biggest visual is a donut of employment status, the least useful chart on the page.
- Cards, then a two-by-two grid with equal 16-pixel gaps, then filters on the right.
- The donut is gone: its numbers are in the cards. The reasons and sources charts took its space.
- Sorted bars, plain titles, neutral colour with one exception highlighted.
The same design choice, made two ways.
| Choice | Presentation page | Dashboard page |
|---|---|---|
| Job | Argue one point | Let readers monitor and explore |
| Title | Action title: finding, number, comparison | Descriptive or question title; a finding only if it's built from a measure |
| Visuals | 1-2, one of them large | 4-6 on a grid (up to about 10 on a dense page) |
| Callouts | 1-2, expected | None: typed text goes stale when filtered |
| Cards | Rarely; one or two numbers as big text | A row of 3-5, top left, each with a comparison |
| Slicers | None: the view is fixed | Up to 3 on the canvas, grouped on one edge, synced across pages |
| Colour | Grey plus one accent on the point | Neutral, with colour only for exceptions |
| Reading time | 3 seconds for the title, 30 for the proof | 3-30-300: cards, then charts, then detail |
| Reader | A decision maker who needs the answer, often in a meeting | An analyst or manager who checks and drills, often every week |
| Interaction | Little or none; bookmarks to reveal steps | Cross-filtering, tooltips, drillthrough |
Set up the canvas once, then snap everything to it.
Size the page
- Click an empty part of the canvas, open Format page > Canvas settings, and set Type to 16:9 (1280 × 720 pixels).
- Set Canvas background to a light grey such as
#F3F2F1so white visuals stand out as tiles.
Plan a grid with equal gaps
- Use a 24-pixel margin and a 16-pixel gap everywhere. On a 1280 × 720 dashboard with a 216-pixel slicer rail, that leaves 992 pixels: four cards of 236, or two charts of 488.
- Type exact sizes in Format visual > General > Properties > Size and Position instead of dragging.
- Turn on View > Gridlines and View > Snap to grid so visuals land on the same lines.
Line visuals up
- Ctrl-click the visuals in a row, then use Format > Align > Align top.
- Use Format > Align > Distribute horizontally to make the gaps equal.
- When the layout is done, turn on View > Lock objects so nothing moves by accident.
Group and sync slicers (dashboards)
- Put the slicers in one rail on the right edge, or one thin band under the title.
- Open View > Sync slicers. Tick the pages each slicer should filter, and the pages it should appear on, so a choice carries across the report.
- Use dropdown slicers for long lists and keep the rest in the Filters pane.
Keep the page in order with the Selection pane
- Open View > Selection. Rename each visual ("Card: exits 2015", "Slicer: department") so the list reads like the page.
- Select the slicers, right-click and choose Group, so the rail moves as one.
- Set the Tab order to match the reading order: title, cards, charts, slicers. Keyboard and screen-reader users follow it.
Lock a presentation page to one view
- Leave slicers off the page. Put any fixed filter in the Filters pane at page level, and click the eye icon to hide it from readers.
- To walk an audience through steps, save each step as a bookmark (View > Bookmarks > Add) and use the pane's View button to click through them, instead of adding visuals.
Pick the page type before you open the answer.
On Monday you'll ask leadership for a retention budget for Production. Which page type, and what goes on it?
A presentation page. Action title: "Production loses 40% of its staff, three times the rate in Sales". One bar chart with Production in the accent colour. A callout with the so-what: "83 of the 102 people who left worked in Production". No slicers.
A dashboard has a text box that says "Production drives our turnover". What's the risk?
A reader filters to Sales or to one performance score, and the sentence stays on screen while the charts say something else. Remove it, or replace it with text built from a measure that recalculates with the filters.
A page title reads "Production turnover is high and exits rose in 2015". What would you do?
Split it into two pages, one message each: "Production loses 40% of its staff, three times the rate in Sales" and "Exits nearly doubled in 2015, to 26 from 14". Each gets one visual coloured to match its title.
A dashboard's three slicers sit in the top-left corner. Where should they go?
In a rail on the right edge (or a thin band under the title), grouped and synced. The top left is where readers start, so it belongs to the card row: the three-second answer, not three boxes set to "All".
A colleague adds six visuals and two slicers to a presentation page "so leaders can explore". What would you suggest?
Keep the presentation page to its one point and one or two visuals. Move the extra visuals and the slicers to a dashboard page placed after it, so leaders who want to explore can, without burying the message.