Good Data, Better Presentation: telling stories with your data

A meaningful dashboard clearly depicts volume, trends, and relationships, provoking action as a result. This is, after all, the very point of web analytics. However, by taking a more narrative approach to their creation—with characters and relationships, context and conclusions—dashboards can be made richer, more communicative, and more actionable than they commonly are.

[This was originally written for a provider of web analytics services. The organization refocussed on so-called big data and associated business intelligence, and it was never published. Some of the references are specific to SaaS applications and subscription-based digital products, but many of the basic guidelines about data presentation remain as relevant as they were at the time of writing.]

Decompartmentalize your data

Imagine for one moment that your dashboard displays a series of spikes in site traffic – is there data on particular promotional campaigns that correlates with them? On a typical web marketing dashboard, Traffic and Marketing might be presented as separate tabs, meaning the user would have to flip back and forth to make any connections. Or a traffic report may be repeated under the Marketing tab in an attempt to make the connection. (And contrary to popular opinion, repetition can quite often cloud rather than clarify; more on that later.)

This approach, as with most dashboards, is purely quantitative – literally, quantities of numbers collected together. Dashboard tabs are organized in a compartmentalized manner, with content arranged by category (everything pertaining to Users arranged under a single tab, for example). Though far from ideal, this approach tends to be the norm amongst business users, since grouping like-with-like is the most natural way to organize things. However, it often means that the dashboard becomes a collection of loosely related reports arranged merely for convenience rather than conclusion. Trends and relationships are not as clear as they could be: connections between reports may be overlooked, or too much interpretation may be required by the reader in order to make those links for themself.

A more qualitative approach is to tell a story around those numbers, perhaps segmenting them in different ways or pairing them with data from other categories (under a single tab) that, together, paint a fuller picture. Tabs can be organized more thematically, and there’s potential for less repetition too.

Take, for example, an operational dashboard that presents the total number of support tickets created, segmented four alternate ways (by submission method, by time of day, and so forth). Those splits could instead be deployed across separate tabs to build a more rounded story. A “Ticket Volume” tab, for example, could include “Tickets Opened” paired with “Tickets Resolved” to provide the same information—and therefore the difference between the two—without the resultant break in thought when split across the more compartmentalized “Opened” and “Resolved” tabs. If, in this case, ticket creation and resolution are diverging and the backlog growing, isn’t it easier to assimilate this information if presented within a single view?

Good dashboards require authorship

Taking this fuller, narrative approach to dashboard creation necessarily requires a degree of editorial oversight – and clear objectives at the outset. As the dashboard’s creator, you’ve made a qualitative assessment and analysis of the data, providing a conclusion or framework for how to view the information. Further interpretation therefore shouldn’t be left to the viewer; they need to be guided, though without leading or bias. Descriptive titles are one way to do this; many dashboards update automatically however, so titles should be monitored to maintain accuracy and appropriateness.

The 5 C’s of good dashboard design – and some avoidable errors too

A dashboard is, by nature, a synthesis of visual presentation, appropriate structure, and relevant content. By following a few simple best practices—which, conveniently enough, all begin with C—good, communicative dashboards can quickly be mastered.

Present conclusions and avoid conjecture

First, and as mentioned above, start at the end: know what you want to say before starting. Good data presentation requires clear communicative intent, and working backwards from an endpoint like this helps to ensure that only the most relevant, useful information is included. It also helps minimize any interpretation required by the viewer. Remember, too, that this is often an iterative process: create a dashboard, analyze the results, refine or add to the reports (perhaps more than once), presenting gradually more defined conclusions.

Provide sufficient context

Compartmentalizing data starves it of context and obscures relationships. If last week’s sales were $4 million, is that good or bad? That depends on what sales were the week before, and what they were predicted to be today. Without providing sufficient context, it’s hard to know how that impressive-sounding metric fits into a trend, or how it relates to the other dashboard components around it.

If, however, the page title stated that “Sales continue to trend upwards”, this would cue the reader as to how to interpret the data presented to them. Additional context (in the form of trend data, supporting information, benchmarks, and key segmentation data) would provide greater meaning and only serve to enhance this. And, in the case of a percentage figure, the raw numbers underlying it would help clarify the sampling size – 62% of 10 is a lot less representative than, say, 62% of 8,000.

It should be said that, though KPIs are inherent to the dashboard concept, they actually run counter to this notion of providing fuller context, since they provide only snapshots – the name itself suggests how lightweight they are (mere excerpts, rather than full paragraphs). KPIs are probably too ubiquitous to do away with altogether, but their ‘quickie’ nature should be borne in mind. In addition, KPI reports are often repeated or subsetted under other tabs, which can be confusing. Many readers may feel it necessary to compare the repeated elements to make sure there isn’t any difference, however slight, and suddenly it’s not so ‘quickie’ anymore.

Communicate with clarity

Will your audience be able to understand your dashboard if you’re not around to talk them through it? It’s a good idea to create your dashboard with this possibility in mind, since it forces you to communicate with absolute clarity. Some key considerations follow.

  • Include the fewest metrics needed: an online community, for example, should be measured by engagement, traffic, responsiveness, and self-direction – no more, no less.
  • Use consistent alignments and similar-sized components – users then don’t have to reorient themselves with each new chart or dashboard. (This doesn’t necessarily mean applying a straitjacket layout either, but the overall layout should impart a clear sense of organization.)
  • Don’t be tempted to use different chart styles just because you can; too much variety becomes disruptive to the process of understanding. If a page of bar charts is the most effective solution, then so be it.
  • Content should be spaced generously enough to help delineate chunks of information. In addition, this makes entry and exit points more findable.
  • Avoid repetition: it obscures patterns and commonalities between reports. (A metric may be repeated if split differently, to provide a different view of the same data or better contextualize an accompanying report.)

Remember that striking the right balance between content and clutter is the challenge in creating cogent dashboards, hence the iterative development cycle previously mentioned.

Engage your audience in a conversation

Imagining that your dashboard doesn’t have a narrator is one way to stay focussed. Another is to take a conversational approach: provide an engaging thread or train of thought, draw from a range of sources, and work towards an endpoint.

  • Use natural language: deploy clear, explanatory tab names, avoid abbreviations and acronyms (brevity doesn’t necessarily aid communication), and avoid excessive punctuation and/or capitalization.
  • Provide clear, informative annotations that assist in the communication process – they should be pertinent but without being repetitive or redundant. (One exception to this might be on dashboards that are only viewed infrequently, such as a quarterly executive-level summary.)
  • Remember that a conversational approach doesn’t have to mean dumbed down, just accessible – and backed up by solid, quantitative data.

Don’t ignore convention

Enable quick, unambiguous comprehension of your data by observing conventions.

  • Use the correct type of visual aid: line graphs to show trends over time; bar charts to compare proportions or volume; and tables to display large amounts of complex data. Other, less common types of visual aid can be harder to interpret – and this includes pie charts, which are only worthwhile in very simple situations where there are perhaps two or three segments of clearly differing proportions.
  • If charts or graphs are positioned adjacent to one another, it should only be for very specific comparative reasons.
  • Data should be more prominent than the axes and legends.
  • Use the same units along the same axes across a series of charts or graphs; try to maintain the same scale too.
  • Always start graph axes at zero.
  • Always show time along the horizontal axis of a graph.
  • Remember some absolute basics of color association: red may be attention-grabbing but it generally means something negative.

Common mistakes – that are entirely avoidable

  • Not maintaining a clear hierarchy and organized structure to your dashboard. Keep common sizes and alignments throughout, and ensure that any emphasis (callout or highlight) provides meaning rather than just decoration.
  • Cramming in too much information. There’s a limit to how much information most people can absorb in one go. Use separate tabs to break up the dashboard into more manageable chunks (think of them as chapters in the story). Don’t go crazy though: twelve short dashboards are probably just as difficult to digest as four long ones.
  • Not making relationships clear. If particular metrics are related (or the same metric appears more than once), it should be made obvious, by means of color association, alignment, or proximity.
  • Using inconsistent ranges on related charts. Use matching scales, or you’ll always be comparing apples to oranges
  • Including redundant annotations or obfuscatory titles. Your charts and graphs should carry the message; any supporting text is just there as a guide.
  • Including too much detail – a dashboard is only supposed to be a summary, after all.
  • Assuming that your audience will “get it”. Instead, put yourself in their shoes and ask yourself what the message of the dashboard is – and how hard is it to grasp the information.


This article provides some general guidelines on how to approach data presentation, so that even if you can’t (or don’t want) to make the so-called narrative leap, you can at least understand some key aspects of better storytelling and more impactful presentation. Generally speaking, the clearer and more organized a dashboard appears to be, the more engaging and communicative it’s likely to be. And on a final note, there’s one last C: a compelling dashboard should:

  • convey meaning;
  • increase understanding;
  • provoke thought; and ultimately
  • spur action.

Care to comment?