Business Intelligence Faster, Simpler, Better
Delivering Immediate Insights
| "Above all, show the data." - Edward Tufte |
People Are Smart. |
Deliver Value Early And Often. |
All BI is local. |
Improving Dashboard Insight's "Best Dashboard of 2011"
Dashboard Insight announced its winner of the best dashboard of 2011 contest here .
This article describes some of its flaws, and adjustments that improve its clarity and communicative power.
The Winner
The winning dashboard was created using Tableau and published to Tableau Public, which makes it easy to access and download. The dashboard is shown here—clicking on it will launch it in Tableau Public.
(Tableau Public dashboards can be functionally embedded; this one's too wide to show it all, so its image is shown reduced to fit)
The Winner Improved
My redesigned dashboard is also published to Tableau Public:
Hopefully, you find mine more accurate and meaningful, a bit cleaner, perhaps a bit more spacious, even a bit more appealing, approachable, and engaging.
The Improvement Zones
Designing an effective dashboard requires a balanced holistic approach that combines the specifics of the atomic elements with the overall design relating the parts to the whole.
The following image outlines the areas in need of help.
The assessments of the shortcomings, the solutions addressing them, and the rationale behind the choice of solutions are below.
Zone redesign descriptions:
1 – Bar Charts
2 – Upper Section
3 – Month Selector
4 – Activity Types
5 – Last 12 Months
6 – Spend vs Budget
7 – Bullet Graphs
8 – Retirement Plan
Additional redesign considerations:
1 – Bar Charts
My first impression of the winning dashboard is that there's a visual mass clustered in the lower left section. Even before recognizing that there are three separate bar charts, the visual weight of the fat blackness draws attention.
Upon examination the charts are actually quite poor at their (presumably) intended purpose—comparing and contrasting monthly expenses by Activity Type in the left chart with payments from the two separate accounts in the left charts. Here's a close up look at them:
To illustrate the shortcomings of this implementation, answer these questions:
Which element appears in both the "Checking Account" (CA) and "Credit Card Account" (CCA) charts?
Shouldn't it be easy to tell?
Which "Monthly Transaction Summary" (MTS) element appears in neither of the CA or CCA charts?
It should take little more than a glance to find out.
Trick question: all the MTS elements appear in one of the other charts, but it's hard to tell.
Which is greater: House Expense in CA or Dine Out in CCA?
Hint: the longest one isn't it.
Which is greater: Interest in CCA or Interest in MTS?
Hint, the longest one isn't it.
If the values are the same, shouldn't the bars have the same dimensions?
The main problems with these charts include:
The charts' bars are differently dimensioned.
The bars in the charts are fat on the left and thinner, but variable, on the right.
This visually weights the presented quantities differently across tables, misrepresenting the values and making comparisons much more difficult than they should be.
The charts are scaled differently, so equal quantities' bars are different lengths.
The MTS and CA charts seem to be close in scale: the bars labeled $1,816.60 seem to be close in length.
But the CA chart and CCA chart use obviously different scales, leading to an over estimation of the magnitude of the CCA values vis-a-vis their counterparts (if any) in the CA chart.
For example: both the CA and CCA charts contain "Car Expense", but their magnitudes are very different, although in comparing them it is very difficult to make a an accurate assessment; without their labels there's no way to tell how the relate, and if we need the numbers there's little point in having the bars.
The CA and CCA charts' baselines are not vertically aligned.
This makes direct vertical comparison of the charts' quantities very difficult.
The elements in the MTS, CA, and CCA charts are not uniformly sorted.
At first blush it looks like the sorting is largest-smallest: the CA elements are in big-small order, as are the CCA elements, but in the MTS chart the Travel element is out of sequence.
But, if the placing of these charts near one another is intended to provide cross-chart comparison a common, consistent ordering would serve much better.
Data inconsistency – missing "Activity Type" values.
The "Activity Type" chart –4– in the Upper Section contains 12 values.
The "Monthly Transaction Summary" chart only contains 10 values.
Presumably, the 2 Activity Types missing from the MTS chart convey the values shown in the Upper Section, therefore their omission from the bar charts is a real problem.
The improved bar charts design:
With this design, all of the questions above can be answered quickly and easily, and it's far easier to spot some interesting things like:
"Interest" has no values presented in any of the charts.
This is illuminating, and important for cognitive and technical reasons.
The element's presence is a signal that somewhere in the data there is data for "Interest", even if there is none for the month being presented.
There are 12 Activity Types in the data; this design ensures that they are each presented no matter what month is being presented, therefore ensuring that the chart's bars will be the same width in all presentations, avoiding the problem for some-fat, some-thin bars in the original dashboard.
"Repayment" shows Checking Payment made even though it's not listed as an expense.
This is most likely to be a data problem: any time a payment is made there should be an expense it's recorded as payment towards.
Well-designed dashboards make it easy to spot anomalies such as this.
The original design made it extremely unlikely that this scenario would have been discovered, not many people have the time, the inclination, or the training to back-chain the data relationships through disjoint, disconnected, visually dissimilar tables like those in the original dashboard.
2 – Upper Section
The main problems with the Upper Section include:
Implied connections that do not exist. — there are two levels of implied connection: structural and decorative.
Proximity
The "Click on Month" component abuts the "Activity Type" component.
This close alignment implies a relationship between the components' members that isn't there.
Row Shading
The row shading in the Upper Section implies a left-right linkage from the Month selector across into the quantitative section where there is none.
In this rendition, the visual linkage between the similarly colored rows in the various components is compelling.
Even when one realizes that there's no semantic association between the rows in the "Click on Month" list of Month/Year labels and the Activity Types it's difficult to disregard the visual attachment.
Organization
The large scale organization has the "Last 12 Months" line graphs immediately following the list of Activity Types. This is at odds with the dashboard's main purpose: to provide a detailed look at monthly budget and spending for individual Activity Types.
It makes sense to move the monthly details to first place in out left-to-right reading order, where they are immediately linked to their Activity Types and elevated to top priority in our cognitive assessment.
It makes sense to move the monthly details to first place in out left-to-right reading order, where they are immediately linked to their Activity Types and elevated to top priority in our cognitive assessment.
3 – Month Selector
Original.
Redesigned.
The original month selector is too visually similar to the adjacent data-presenting elements, making it difficult to distinguish between them.
The selector is now visually distinct, clearly segregating it from the other elements.
- The rows are the same size, and in number.
- The row shading, while not identical, is close enough to visually associate the individual months with the Activity Types immediately to the right.
- The selector is border is the same as the data-presenting elements.
- The selector abuts the Acvitity Type element.
- The row shading is removed—only the selected month is highlighted.
- The border is removed.
- A distinct space between the Month Selector and the Activity Type element has been added.
- The months' presentation has been changed to nested year/month.
4 – Activity Types
Original.
Redesigned.
In and of itself, there's not much wrong with this. The only quibble is that single row shading is generally unhelpful and can be a hindrance.
The drawback is that, in the original dashboard, the Activity Type chart is stand-alone, disconnected from the other elements that it's related to.
In the dashboard redesign the Activity Types are integrated into the Spend vs Budget chart, which makes it a great deal easier to grasp the details and relationships between the data.
- The row banding is changed to 2 rows per band.
- See Spend vs Budget for additional information.
5 – Last 12 Months
Original.
12 Months of what?
Redesigned.
The big problem here is that there's no indication of what's being measured over the last 12 months.
Otherwise, sparklines are a good choice for showing trends over time in a constrained space. Although these aren't strictly sparklines,.they're reasonably effective.
However, they do have some problems, including:
This design clearly highlights those months where Sheila was "in the hole", when her spending exceeded her budget.
It's cleaner, tidier, and tighter, allowing the exceptional months to stand out.
- The lines are fatter than they need to be. We can discriminate finer lines than these.
- There's a cognitive conflict between the coloring and vertical dimensionality. Our cultural metaphors are: red is bad; and up is positive in number lines. In these pseudo-sparklines, the bulk red represents the negative monthly variance (spending exceeds budget).
- The lines are thinner—there's less low-information-value ink to distract from the real signals.
- The range of size variation has been narrowed. The original chart lines were too fat, adding bulk and detracting from clarity.
- The values have been inverted—the negative values are now below the zero line, which is where we expect them to be.
6 – Spend vs Budget
Original.
Redesigned.
Aside from being disconnected from the Activity Types the values related to, it's not bad, but there are a couple of small improvements that can be made.
The redesign integrates and extends the Budget, Spent, and Variance with their Activity Types into an holistic display that makes it much easier to make the connections between them.
- Single-row banding isn't very useful.
- Using $s everywhere adds visual clutter when there's no need: everything's money, so the $s add no semantic value.
- Using "Left" to label the result of subtracting Spent from Budget doesn't make sense when Spent is greater: there is nothing "Left".
-
The Activity Types are added to the chart.
(done in coordination with reconfiguring the Upper Section) - Totals are added for Budget, Spent, and Variance. The Variance total is emphasized and color-coded red when negative.
-
The order of Budget and Spent is changed, with Budget coming "before" spent, which is more in line with common practice, e.g.
arithmetically:
Budget - Spent = Variance -
The row banding is changed to 2 rows per band.
(done in coordination with the row banding in the bar charts, simplifying cross-chart referencing) - The $s are removed.
7 – Bullet Graphs
One of Dashboard Insight's suggested dashboard improvements is to provide additional state markers to identify those Bullet Graphs where the spend exceeds the budget.
This would be a mistake.
Bullet Graphs were invented to convey this information quickly and accurately; adding another element to the dashboard to accomplish the same thing is redundant, sloppy, and imposes an unnecessary cognitive burden on the dashboard reader.
The proper solution is to adjust the Bullet Graphs to make them easier to read—this is fairly easily done.
Original.
Redesigned.
The label "Month to Date" is misleading when the data being displayed is for every month except the current one.
The Bullet Graphs are cluttered, making it difficult to evaluate the characteristics of the individual graphs.
The redesign simplifies and declutters the presentation, making it easier to grasp the essential relationships.
- The chart label is changed to "Budget vs Spend" to match the actual data being presented.
- The central bars, representing the amount spent, are narrower, making them easier to pick out.
8 – Retirement Plan
There is no horizontal axis on the chart, making it impossible to determine how the values are being grouped.
So I added it.
The redesigned chart is smaller than the original. It still conveys the essential characteristics of the data and is proportional to the relevance of its place in the overall financial framework.
Additions – elements added to redesign.
The original dashboard provides the details about the types of monthly expenses, including the evaluation of spending vs budget for each type.
But it's missing monthly totals for the budget, spending, and net income (budget - spending). So I added them:
Descriptions:
The monthly totals for Budget, Spend, and Variance are provided, in the customary locations.
The Variance total is larger than the others to make it more noticable. It's also color coded: red when negative, black otherwise.
The absence of dollar signs –$– is intentional, employing them would add little benefit and add non-trivial cognitive load
The monthly Variance total is a critically important piece of information, particularly when expenditures exceed budget, i.e. Sheila is in the hole for the month.
Placing it in the top-right corner makes it easy to assess the month's spend vs budget relationship at a glance.
Harmony – making the parts work together.
The original design's Upper and Lower Sections cover the same data, but are not well coordinated, making it much harder than it should be to track elements acrosss the sections, e.g.:
- It should be easy to locate the car expenses in the Lower Section once it's identified in the Upper Section—but not only is it not easy, the existing design actively impedes the effort.
- It should be easy to spot the Activity Types that are in the Upper Section but missing in the Lower Section—"Misc" and "Repayment"
The redesign remedies this by making these adjustments to the upper and Lower Sections' Activity Type charts:
- Left-aligning the Lower Section's chart with it's Upper Section mate.
- Ensuring that both charts contain the full set of Actitvity Types and that they're sorted the same way—alphabetically by name.
- Applying the same row shading to the charts – using two-row subtle band shading makes is easy to locate individual Activity Types in either chart, and then visually transition to the same (relative) location in the other chart.
Once again, here's the redesigned version:
Conclusion.
The final redesign is better than the original, but it's not perfect. Dashboards almost never, if ever, are.
There are improvements that could still be made. Some of these are technical and could be implemented straight away.
Some require engagment with the dashboard's reader—Sheila, to ensure that she has the information she needs, and that she can read and absorb it with minimal effort.