Data Visualization and Dashboard Projects

Quickstart Guide
Introduction
So, you’re launching a data visualization and dashboards project? Not to worry, this quick
guide will help you understand the best ways to visualize your data, plan your dashboards
and meet your audience’s needs. We’ve based this content on the extensive experience of
Hitachi Vantara’s custom UI and visualization design team in collaborating with customers
on a wide variety of projects. Here’s a brief agenda of what we’ll cover:

Getting Started and Defining The items you need to consider before you start creating analytics — project
the Project objectives, core KPIs, user needs, data readiness and more

How to identify visualizations that best illustrate different types of metrics, plus
Data Visualization Selection
useful tips for grouping data by geography, time period and other categories

Best practices for placing different metrics and visual analytics on a dashboard
Dashboard Layout and Design
for effective presentation to your audience

What to consider as you look to make your dashboard more engaging for users
User Interaction
— including dynamic filters, drilldown capabilities and more

Getting Started and Defining But how do you focus in on the correct KPIs to commu-
nicate? This requires coordinating with strategic project
the Project sponsors that are driving the business initiative, as well as

Behind every dashboard project is some type of organiza- working closely with the intended day-to-day users of the

tional need — whether it’s as simple as keeping salespeople dashboards. Determining the KPIs to show (and executing

up-to-date on weekly performance or as sophisticated as on the entire project) should be an iterative process where

providing physicians with better treatment recommendations you continuously circle back with those stakeholder groups

via predictive modeling. Addressing these needs with concrete for feedback. As you decide on your first set of requirements

measurable objectives provides the necessary framework for and begin the design process, you should schedule regular

delivering the right information to the analytics consumer in the update discussions with users and sponsors, continuing

right fashion. If the project goal is to increase customer service these meetings until the final project roll-out.

efficiency by at least 10%, for example, that specific objective

While we won’t get into detail on all of the project items that
will help shape and prioritize the type of key performance indi-
should be addressed, you can use the following checklist to
cators (KPIs) that are presented to the service representatives
make sure you are considering everything you need to before
using the dashboard.
laying out your visual analytics.

Data Visualization and Dashboard Projects 2

 Question Sample Answer

What’s the organizational need you Understand return on investment (ROI) of different marketing campaigns in order
are trying to address? to better allocate budget going forward

Have you agreed on objectives with Yes, the Marketing VP is a driving sponsor and the whole team knows about
business/executive sponsors? the project

Have you met with your potential I’ve met with the marketing program managers and gathered requirements, but I
dashboard users to understand their still need to meet with our social media and event marketing coordinators
needs, preferences and context?

What are the main data sources you Data from a CRM system, marketing automation system, finance databases and
need to access? social media accounts

Will you be able to receive updated data Yes, this analysis will mostly be on a week-to-week basis and my information will
in the frequency you need it? be refreshed at least every day

Is there a measurable goal you Reduce the average marketing spend per qualified lead by at least 10% in the
want to achieve? next year

What are the core KPIs you want Cost per lead, cost per sales opportunity, return on campaign investment — all
your users to understand? broken down by marketing channel

What’s the analytical sophistication Medium sophistication — they will sometimes want to drill own into the granular
of your audience? information behind the dashboard

Is the data integrated, accurate It will be soon — our IT operations team are leveraging a data integration solution
and clean? to cleanse the data and schedule proper data updates

When gathering user requirements, it is important to understand databases, big data stores, or other information systems. You
your audience and their level of data fluency. Are they executives also need to ensure that your data can be properly integrated
looking at the big picture or business analysts who will want to and blended together for your project — this includes join-
filter and customize the data? Maybe they are customer-facing ing data sets and ensuring metadata reflects business rules.
employees who really just want to see a few key operational At the same time, you need to be sure that the information is
numbers with simple recommendations on what actions to take clean, accurate and available in the frequency you need it to be
next. Your audience could also be external to your organization, updated (weekly, hourly, real-time, etc).
in which case the dashboard theming, branding and context
may need to fit with existing outward facing applications.
Data Visualization Selection
At the same time, it may help to think of your goal as telling a story In the large majority of dashboard projects, we’ve found that
with the analytics content, and as such make sure to consider KPIs fall into a handful of core categories, which have specific
the “what”, “who”, “when”, “where” and “why” of your dashboard implications for visual analytics. As such, we’ve mapped visu-
project. Overall, it’s an exercise in meeting your users’ needs and alization guidelines and tips to the different varieties of KPIs,
creating a good experience for them — one that will promote along with examples.
adoption and boost usage of the analytic content. Considering
multiple dimensions of your analytics will help with this. As you While this section starts out with the basics, it should help guide
start to gather requirements, creating sketches of possible visu- you as you determine which data visuals are right for you.
alizations and dashboards with your users will help you begin to
determine optimal ways to meet their needs.

Finally, you’ll want to have your data in order before you identify
the key metrics to showcase. This means identifying all of
the data sources you are using, whether they are relational

Data Visualization and Dashboard Projects 3

Quantities the plot. If identifying extreme data points or outliers that don’t
You can think of these as your bread and butter counts or mea- conform to a normal relationship is important, this could also
sures, such as the number of units sold, time taken to complete be a useful way to visualize data for your users.
an examination, or tons of freight carried by a rail car. Currency,
whether Dollars, Euros, Yen or otherwise, is also an example. We will touch on more options for visualizing basic quantities as
we navigate through the other types of KPIs and groupings below.
There are understandably a wide variety of options for visual-
izing basic quantities. Bar or column charts are a good place Trends and Changes Over Time
to start, where each bar represents a quantity of something Time series data is a deep analytics topic, but the crucial point
(like units sold) for a specific grouping (like product lines). If for most visuals it that it adds a new dimension to the analysis,
there are a limited number of groupings, these types of charts normally where time is displayed on the horizontal axis of your
provide a visually intuitive way to compare quantities, as repre- chart. The more lines or series you add to a line graph, the more
sented by length or height. Line graphs are also a good choice difficult it becomes to understand and interpret at a glance. So
if the quantities are tracked over time. it’s a best practice to keep it to less than 4 series, else you risk
confusing your users.
INTERACTIONS

1,732

931

Basic time series graph plotting sales and gross profit

Simple bar charts are great if the user needs to rapidly over a month.
compare different quantities.
These types of charts give users an intuitive feel for changes
If you are more interested in the direct relationship between over time — i.e. growth rates and decline rates. When looking at
two different quantities, you will probably want to use some rates of change for a specific measure over time, it can be useful
type of scatter plot visualization. When a substantial correla- to simply indicate the rate with clear text near the appropriate
tion between the variables exists, a linear pattern will form on series on the chart. Showing a directional arrow icon (up, down,
or flat) may also provide the user with a clear sense of the
change rate.

Displaying a growth rate and icon next to a KPI — in this case

entrants in a race.

Scatter plots help show correlation between variables — this

example shows the relationship between total store sales
(vertical) and sales from a specific promotion (horizontal).
The colors represent different product types.

Data Visualization and Dashboard Projects 4

Relative Shares and Proportions A unique alternative for showing proportions could also be a
Shares and proportions display a relationship between the parts radar chart (also known as a spider or web chart). In this visual,
and the whole, rather than differences over time. Examples here the proximity of the data point to a corner indicates the relative
are shares of an investment portfolio allocated to stocks vs. importance of that corner, while several ‘webs’ can be overlaid
bonds as well as website conversion rates — i.e. the portion of to illustrate relative results for different periods or other group-
people who opened an email, clicked it, or didn’t open it. ings. Indeed each ‘web’ creates a different shape representing a
different scenario to the viewer.
If the primary interest is to display the composition of an aggre-
gate quantity at one point in time — say the breakdown of the
investment portfolio by asset class today — then a pie chart
would make sense, with slices representing different portions
that add to 100%.

Basic pie chart showing proportions of medical patients

who spent time in different types of hospital rooms.

Pie charts on their own, however, do not convey a tremendous

amount of information. If we wanted to show proportions over
time, we would use a stacked bar or area chart that can be
configured to display total quantities over time broken down by
categories in either absolute or percentage terms.
Radar chart captures the relative importance of different
customer churn drivers for a telecom company both in
2013 and in 2012 in a quickly consumable format.

Area chart comparing the uptake over time for different

software versions following their initial release.

Data Visualization and Dashboard Projects 5

Ranked Lists Location can get fairly complex, but there are a few core
Presenting a ranked list of the highest or lowest items from a concepts to remember. First and foremost, don’t be tempted
data set is a common dashboard approach. Though not really to include overlays like streets, terrain, or traffic if they are not
data visualization, this is a good way to provide the ‘need to relevant to the insight you are trying to provide. If overlays are
know’ information, assuming there is a way for the user to important, give users the option to turn them on or off accord-
access greater detail. For instance, a sales management dash- ing to their preferences. In general, you will want to limit the
board may include quota attainment for the 5 best and worst colors, lines and other detail in your default map view prior
performing sales representatives on the team, with the ability to overlaying your specific KPIs on the map (your metrics are
to click through to a view that shows all sales reps with a wider what should stand out!).
array of metrics for each rep.
Second, there are different options for coding regions — like
states, countries, zip codes, etc. For example, you could fill
within borders with colors representing additional categories
or metrics. Alternatively, you could map specific granular
data points and look for clusters and patterns independent of
regional totals.

Finally, a must-have for geo maps is the ability to zoom in or

out to different regions, tracking city vs. state/province vs.
country level aggregates in the process. Your users are likely
used to navigating a service like Google Maps for their own
personal geographical needs — so their standard for ease of
use will be high.

Ranked list of top 10 music tracks by number of streams

over a time period.

Geography and Location

As you might expect, in this type of visual you are looking at data
on a map or physical representation of a real location, which
gives the user a richer information experience based on known
spatial and physical relationships.

Simplified geographic map with no overlays makes it easier

to see data points.

Data Visualization and Dashboard Projects 6

Dealing With Multiple Dimensions at Once
As you might expect, some of the richest visualizations display
more than one type of quantity and one type of grouping in
the same space. However, as the variety of information you
are trying to communicate increases, so does the risk that
your users will miss the point. We’ll walk through two types of
multidimensional visualizations here — bubble-scatter charts
and heat grids.

In a bubble-scatter visualization, the scatter plot discussed

earlier is augmented with one or more additional dimensions.
First would be the size of the data point (hence the ‘bubble’)
— where the area of the bubble represents a numeric value
associated with that point. This would be followed by color
of the data point, to represent another factor.

Heat grid displaying average revenue per user (color

scheme) across customers, by both geographic market
(vertical) and type of mobile device used (horizontal).

Dashboard Layout and Design

You could have the best data visualizations in the world, but
if they aren’t organized in a logical and orderly fashion, your
dashboard project will miss the mark. While not every dash-
board should be designed the same way, there are several
Bubble-scatter chart showing marketing campaigns by value
guiding principals. As you follow these guidelines, we’d sug-
generated (horizontal) and invested cost (vertical), as well as
gest sketching out different possible layouts on paper.
duration the campaign ran for (bubble size).

Amount of Content
In a heat grid, we can leverage color to convey values or ‘tem-
You will usually want to include no more than 3-4 actual charts
peratures’ of data points on a horizontal and vertical access.
or graphs for your audience. Any more than that, and things
This type of visual is more appropriate where we expect mul-
will quickly get crowded for the user as you start to add head-
tiple Y axis values for every X axis value, and thus makes sense
line metrics, titles, text and labels. By the same token, limit-
for representing values sliced and diced by different general
ing the actual number of visualizations will also minimize the
or neutral categories. Put another way, a heat grid can help
amount of scrolling the user will have to do. If you must have
where a stacked bar chart might fall short — rather than trying
more visuals, then make the others accessible via a button
to display values by size stacked on one another, the size of
click or toggle that replaces the content shown to the user. It
the data points can be left constant with the colors represent-
is better to have contained groups of information presented to
ing the value scale. That said, the size of a data point could be
the user at all times than to try to cram every last data point
added in as another dimension in some situations.
into one screen.

You could have the best data visualizations in the world,

but if they aren’t organized in a logical and orderly fashion, your
dashboard project will miss the mark.

Data Visualization and Dashboard Projects 7

This telecom dashboard follows best practices by including simply presented aggregate metrics at the top and drill-down visuals
beneath. Note also that churn rate and average revenue per user appear first from the left, as they are the most important
metrics across the full dashboard.

What Goes Where Themes and Formats

A good general pattern to follow in your layout is to have higher We couldn’t complete our key guidelines on design without
level or ‘headline’ metrics displayed near the top part of the discussing theming and formatting. In choosing color schemes
dashboard — sometimes as simple as a few core KPIs in large for your dashboard, it is crucial to select a palette with good
clear font, perhaps with adjacent spark lines to show trends. contrast and balance. For instance, avoid reds overlapping
Just below should be the data visualizations that show over- with blues or dark grays. As you want to display more and
view data like aggregate KPIs over time. As you move down, more different groupings in the same data set (high cardinality),
more granular visuals should be included — for example, if you need to use more colors — and it thus becomes harder to
overall customer growth rate is a key metric you highlight near tell two shades apart. Ways to deal with this include balancing
the top of your dashboard, then a breakdown of new custom- values for hue, saturation and lightness.
ers by segment or demographic that displays actual account
names should go further down. Rather than trying to reinvent the wheel, a good place to start
might be to look at your organization’s current branding for its
Not surprisingly, you should also follow the general rule of corporate website and collateral. Following color schemes,
placing the most important information at the top of the screen font types and other visual elements will provide you a good
and to the left, as people start reading in this direction — at template for the content you are producing.
least in English-speaking countries. If you are designing for
an audience that reads right to left, layout should be adjusted Finally, we’d note that text such as titles, labels and user
accordingly. It may be obvious, but you should remember that instructions should always be present but not in an over-
larger visuals and larger text indicate the relative importance of whelming manner. While it makes sense to illustrate the most
information as well. important KPIs in large clear text, other types of text should
appear more for reference than anything else.

Data Visualization and Dashboard Projects 8

User Interaction
While static visualizations and dashboards have their place, Normally, you will want the time selector to drive all relevant
interactive analytics content provides a more engaging experi- metrics and visuals in the dashboard — and if it doesn’t, you will
ence, providing richer informational context and increasing user have to clearly explain this to your users. At the same time, this
adoption and satisfaction with the tools. We’ll touch on a few key selector should be placed somewhere near the top-level titles or
categories below. menus and be clearly labeled but not obtrusive.

Filters and Selectors For time series charts, a dynamic period selector that can be
Flexibility is key when it comes to filters and selectors. For expanded and contracted by clicking and dragging an area of
instance, when providing a time range selector, you’d want to the chart can be an easy way for the user to adjust the time-
provide a balanced mix of pre-formatted date ranges (i.e. this frame as they see fit.
week, this quarter, full year) and a full calendar widget for
defining a custom range.

Unobtrusive drop-down selector for time frame with both pre-defined period like ‘last
week’ and a small calendar for ad-hoc date ranges.

A period selector on an overview time series chart drives the timing of the KPIs
displayed below.

Data Visualization and Dashboard Projects 9

Hover and Drill Down Interaction
Mouse overs and clicks by users can provide great opportunities for interaction. For instance, when the user hovers over an item on
the dashboard, such as an interactive menu or a data point on a chart, it can be useful to trigger some sort of highlighting or shading
to make that point stand out. At the same time, providing additional information about the data point can be helpful — this could be
as simple as labeling the precise value of a single data point on a line graph when the user hovers over it. You could also choose to
provide tips and hints upon hover as well — for example, indicating that certain variables can be sorted.

Simple illustration of data point information displayed on hover.

Allowing users to drill down to more granular underlying information can also be helpful, especially for business analysts and other
more sophisticated data consumers. The crucial thing is to remember that the data and context presented upon click must be clearly
distinct from the original view. This could either be in the form of a ‘pop-out’ window that displays the drill down information or a zoom-
in from the initial view, depending on your requirements. Once the user has accessed the more granular view, there must be an easy
way to return to the original visualization, such as a ‘back’ button.

Finally, users often want to be able to export the dashboard data, most often to Excel, PDF, or CSV format. This is most often enabled
by a button or menu for the task.

Going Further
We hope this piece has given you some practical tips to enhance the success of your project. As you finalize your designs and move to
implementation, remember that the project doesn’t end with roll-out to your users. Requirements will continue to change over time and
so will your analytics content. As such, it is absolutely critical for you to anticipate the longer-term requirements of your users and other
strategic stakeholders inside and outside the organization. You’ll also want to make sure that the data and analytics software you use is
future-proofed — that it allows you to easily customize and iterate on both the end user analytics and the types of data you want to incor-
porate in your dashboards.

Data Visualization and Dashboard Projects 10

 Learn more about Pentaho Analytics at HitachiVantara.com

About Hitachi Vantara

Hitachi Vantara, a wholly owned subsidiary of Hitachi, Ltd., helps data-driven leaders find and use the value in their data to innovate
intelligently and reach outcomes that matter for business and society. We combine technology, intellectual property and industry
knowledge to deliver data-managing solutions that help enterprises improve their customers’ experiences, develop new revenue
streams, and lower business costs. Only Hitachi Vantara elevates your innovation advantage by combining deep information
technology (IT), operational technology (OT) and domain expertise. We work with organizations everywhere to drive data to
meaningful outcomes. Visit us at HitachiVantara.com.

Hitachi Vantara
Corporate Headquarters Regional Contact Information
2845 Lafayette Street Americas: +1 866 374 5822 or [email protected]
Santa Clara, CA 95050-2639 USA Europe, Middle East and Africa: +44 (0) 1753 618000 or [email protected]
www.HitachiVantara.com | community.HitachiVantara.com Asia Pacific: +852 3189 7900 or [email protected]

HITACHI is a registered trademark of Hitachi, Ltd. Pentaho is a trademark or registered trademark of Hitachi Vantara Corporation. All other trademarks, service marks and company names
are properties of their respective owners.
P-026-A DG April 2018