Visual Analytics in Incorta
This article looks at visual analytics in Incorta, different insight and graph options available, and what each are best fit to represent.
We recommend that you be familiar with these Incorta concepts before exploring this topic further.
These concepts apply to version 4.6 and above of Incorta for customers implementing Incorta on premises or in their own cloud and versions 21.3.1 and above for those using the Incorta Cloud.
Incorta provides numerous functionalities and capabilities for data visualization that allow you to consolidate all your key information in one page. As an end-to-end analytics platform, Incorta ensures your data is current and allows you to establish a single source of truth for all your users to look at. You can build multiple insights within Incorta, representing your data in different ways, and combine them within a dashboard, ensuring that interactivity is in place. Simplicity is key, so focus each dashboard on answering specific questions and refrain from adding unnecessary and confusing insight to the final view.
As you look at the data you plan to include in your dashboard, separate the different data elements and review the characteristics of each one. These will fall into two categories, measures and dimensions, which you will work with differently.
Measures are numerical data that are calculated or aggregated (totals, averages, etc...) or non-numerical data that you are counting, while dimensions are categorical information within your data. Data type (integer, long, string, date, etc...) does not always determine whether you are dealing with a measure or a dimension, as numerical information that does not get aggregated can be considered dimensional (phone number, numerical ID...).
Measures will always have an aggregated element added to them, and should be formatted based on what the data represents (total sales would be represented with the currency symbol appended and in thousands, while ratios would be represented as a percentage...).
Dimensions can be nominal, ordinal, interval or ratio with each represented differently.
- Nominal dimensions are simple discrete values within the data with no order in place. Examples: gender, customer segment, country
- Ordinal dimensions have a default order in place. Examples: sales stage, education level
- Interval dimensions represent dimensions for which each category represents an equally spaced range of values. Examples: credit scores, years and time
- Ratio dimensions are similar to interval dimension with the addition of a clear definition of "0". Examples: weight, length
Choosing the wrong charting type can lead to a misrepresentation of your data, so let's go through examples of the most common types of graphs and what information they represent the best.
Table Charts show values in rows and columns, where the values are aggregated at a certain level (Aggregated Tables, Pivot Tables) or listed at the granularity level of the raw data (Listing Tables). They give the possibility of representing many related and unrelated measures seen at the level of multiple selected categories. Tables charts should, in theory, be used as a last visualization option as they do not add any value in terms of quickly interpreting data. That said, having the raw data available in a table as a reference, perhaps on another tab, can be useful, especially for seeing the effects of filtering.
When working with table charts, consider how your table will be used and determine your audience type. It is recommended that numbers be aligned to the right, as this makes it easier to compare. Text can be left aligned, but you may prefer to center it for readability. Use text color or or background conditional formatting to draw the user to specific values in your table.
But most importantly, consider another type of graph that offers better interactivity prior to settling on a table if your audience needs to review a lot of data. Scrolling through hundreds, thousands or millions of rows will not be a good experience for your end users.
KPI Charts allow the representation of high level metrics with little to no relation with one another. These are, by definition, the Key Performance Indicators that most people starting an analysis will want to look at, and they direct how your users interact with the dashboard and what they focus on.
When working with KPI charts, place them at the top of your dashboard as that is where the user's attention will go first. This placement allows the user to take note of the critical numbers immediately and decide what they want to drill down to next. The numbers should be labeled correctly and formatted in a manner that makes them easy to interpret.
Pie and Doughnut Charts are some of the most used and hated types of charts. They are used to show parts of a whole. It represents numbers as percentages and the sum total of all segments equals 100%.
When working with these charts, make sure the sum of your segments equals 100%, as otherwise the pie chart will be rendered erroneous. You should keep the graph clear and consistent and limit the comparison to a few categories to get your point across; too many small chunks make the chart hard to read. If the different parts of the chart are roughly of the same size, consider using a bar chart or histogram instead.
Refrain from using 3D imaging or finding ways to tilt the pie chart. This often makes your data unreadable as the viewer tries to compare angles.
Column Charts and Bar Charts allow the comparison of measures across multiple categories. They can be used to compare multiple measures that are at the same scale, though it is advisable to limit the number to 3. They are typically used to avoid clutter when a data label is long or there are more than 10 items to compare. They are easy to understand and interpret.
Categories can be kept in their original order for ordinal and ratio dimensions, or sorted in an increasing or decreasing order. Column/Bar Charts are ideal for showing the top or bottom N values to emphasize the largest or smallest values
When working with these charts, start the y-axis at zero. Our vision is sensitive to the size of bars on a graph; if they are are cut off, the viewer may draw the wrong conclusions. Also, make sure to specify the correct names for the axis labels as this allows viewers to quickly understand the content of the graph. When possible, place values next to the bars as this helps to keep the graph clear. You should avoid using too many colors. Use a single color instead or vary the tone of the same color ; you can highlight a particular bar with a different color if that bar's value is focus of the message you want to get across.
Another variation of this chart in Incorta is the "Radial Bar Chart".
Line Charts are mostly used when representing time series or data that changes over time. They can visually highlight a temporal trend and can compare the performance of multiple dimensions over time revealing changes. They can also be used to show relationships within a continuous data set.
When working with line charts, clearly label your axes so that the viewer knows what they are evaluating. Remove distracting colors, and overloaded captions that can prevent the viewer from quickly seeing the overall trend. You should zoom in on the y-axis if your dataset starts above zero; you can even consider changing the scale of the Y axis as it makes it easier to interpret the graph sometimes. When comparing multiple measures/dimensions, limit the graph to 5-7 lines to make it easier to read.
Other variations of this chart in Incorta are the "Line Time Series" and the "Time Series".
Area Charts are similar to line charts, with some subtle differences. Both can show changes over time, general trends, and the continuity of a data set. However, the space between the axis and the curve is filled, indicating a volume. These charts are used to illustrate the big picture. When looking at population for example, line charts are useful for showing net changes in population over time while area charts are useful for representing the total population over time.
When working with area charts, make it easy to read by making sure transparency is in place when multiple area layers overlap. Use a stacked area chart if you have multiple sets of data and want to emphasize the relationships between parts of a whole. Avoid comparing too many categories though - in that case opt for a Line Chart instead.
In general, Stacked Graphs are useful for examining totals that are made up of several categories. Use them to compare sub-categories within a whole against each other and how each contributes to the total. For example, if you are interested in line of business (LOB) contribution to sales by territory, a stacked graph will quickly help to pick out the relative strengths and weaknesses of each LOB and their impact on the overall performance of each region.
When working with stacked graphs, make sure there are not too many categories that make up the whole, as viewers may find it difficult to distinguish the contribution of each.
Dual Axis Graphs allow you to correlate two measures and their variation. This could be performed as two synchronized axes (Combo Dual Axis) or 2 unrelated axes. It is an important graph type when looking at measures that are at different scales, to identify how they vary and evolve independently.
There are different types of multiple axis graphs:
- Histogram and curve. This dual axis chart combines a bar chart and a line chart.
- Curve and curve. This dual axis chart compares two line charts. It is possible to have more than two lines if necessary.
- Histogram and histogram. This dual axis bar chart displays two sets of data side by side.
They are also great for looking at variables that are at the same scale when looking at how their trends vary but within the data visualization community many are skeptical about using this type of chart because it can be confusing to the viewer. In fact, users need to be aware of the graph legend when dual axis are used to identify to which axis each graphs relates to, they need to be aware of whether the 2 axis are at the same scale or different; all of which are some of the elements that could lead to the wrong interpretation of the graph.
Treemaps help to visualize hierarchies or complex structures. They are a good option for viewing a large repertoire of data and variables, without taking up too much screen space. They display hierarchical information as a nesting of rectangles whose size and color vary according to the associated value. The size of each rectangle represents a quantity, while the color can represent the value of a number or a category. The treemap is a compact and efficient option for displaying hierarchies and gives a quick overview of a structure. It is also a relevant means of comparison between different dimensions.
When working with treemaps, start with clear data and a clear message; they often represent a lot of data so it's important to know what you want to highlight. Use contrasting bright colors to make each region stand out but don't use too many colors. Label each region appropriately with text or numbers to make reading and interpreting the graph easier and avoid overloading your treemap chart with too many boxes.
Heatmaps allow you to provide a global view of several indicators and measures. They are ideal for cross-examining data through placing variables in the rows and columns and using the color of the cells to highlight important data. A Heatmap visualizes your data through variations in coloring using a warm-to-cool color spectrum. The chart is useful for analyzing the relationship between two variables, placing the different possible values of each in rows and columns and coloring cells in the table based on a third common attribute. It is useful for revealing all relationships, showing whether variables are similar to each other, and for detecting if correlations exist between them.
Maps allow you to visualize data on a map. They are the perfect type of graph for representing geographical data combined with performance data. Mapping is used to display divided geographic areas, colored regions, or location as small as a building's plan, relative to a data variable. This makes it possible to visualize values over a geographical area and display variations or geographical trends within the data.
In Incorta, you can rely on either fully colored or bubble maps, You can utilize geospatial information like country, state, county names and full addresses or lat/long information to represent your geospatial information. Tooltips can be utilised to represent additional information and measures that would otherwise make the map busy.
Scatter Plots are ideal for showing the correlation between two measures. This correlation is shown by placing data points between an X and Y axis. Essentially, each of the data points is “scattered” around the graph. It is an ideal graph for identifying outliers within your data.
Tag Clouds (also called word clouds) are a type of weighted list that display text in different font sizes, weights, or colors to show frequencies or categories. They help people identify trends and recurrences that might have been difficult to see otherwise
When working with these charts, make sure to provide context - Word clouds are visually catchy and provide information about frequency, but they often do not give the viewer any context. Use word clouds to show frequency and avoid using them to display complex topics like budgets or the healthcare crisis. Watch the length of your word; words that are too long take up more space and can be misleading. Also, avoid making your words too similar in size and when trying to differentiate between them using color avoid using color tones that are too similar.
Bubble Charts help to compare and find relations between categories by using the bubbles' positioning and size. They can be used to identify relationships and correlations within the data.
Bubble Charts are ideal for analyzing complex groups but can be difficult to interpret when there are too many bubbles displayed. They are ideal for interacting with (filtering or drilling through) and can be a good first step to a more granular analysis of your data.
Clear text and comments help your audience understand what to expect from a certain dashboard and how to utilize a certain graph. This allows you to emphasize certain aspects of a visualization and to draw attention to what is important about it.
Spending time to label your dashboards and each of your graphs accurately provides viewers with context that makes understanding the story that the visualizations are telling more impactful.
There are other types of graphs available within Incorta.
- Percent Column/Bar/Area Charts allow you to show how elements that make up a part a whole compare to one another and how they each compare to the whole.
- Spider Charts compare multiple measures across categorical values.
- Funnel & Pyramid charts allow you to represent categories that need to be arranged in a way that shows hierarchical structure, as well as quantity or size.
- Gauges provide a comparison of actuals against targets. They are mostly used when analyzing performance and are good for identifying progress of an indicator towards a goal.
Incorta provides the possibility to map any data you want to connect to into a graphic representation. This mapping determines which attributes within the data and what element variation you want to focus on (length of a bar, strength of a color, size of a bubble...). Because both design skills and statistical and computing skills are required to visualize effectively, visual analytics are considered both an Art and a Science.
Always bear in mind how people utilise analytical content you're providing, and therefore take in consideration how they can read and misread the various graphs you're including. Ensure the visualisations utilised are the most understandable and effective in conveying your results.