Business
intelligence (BI) software vendor QlikView recently ran a contest called
the Open Data Challenge: here, BI users were to analyze data on
populations, poverty, health, education, transportation and the environment,
and create a dashboard that would inspire users to take action.
And contestants could only use open data (meaning data that is publicly
available for free download).
First-place winner of the contest is Alexandre Perrot: as a
QlikView Solution Expert for Keyrus, he has a decade
of experience developing business intelligence dashboard applications.
His
dashboard “Do You Realize?” tells the story of how great the human
impact is on our planet, and provides end users with resources for effecting
positive change. Perrot completed the project in just three weeks: an amazing
turnaround.
The “Environment” screen of Perrot’s
dashboard, displaying information about population growth for selected areas of
the world.
I
caught up with Perrot to discover the dashboard creation tips that helped win
him first place in this contest. Here are some of his best practices for
creating an effective and actionable application:
Think About the Data
Before the Dashboard
Before
you even begin to build your dashboard, you must think carefully about the data
and how it will be used, says Perrot. It’s absolutely imperative to use
accurate, relevant and reliable data: if your numbers aren’t sound, the insight
they provide is misleading, and the dashboard isn’t useful.
You
also need to organize this data and present it in a way that makes sense.
Perrot commonly uses star or snowflake schemas, and this is what he suggests
others default to. For most datasets, these models provide the best
infrastructure for organizing data and enabling the end user to view it in a
variety of different ways.
Here
is the snowflake schema Perrot used in the “Do You Realize?” application:
Here,
you can see that the user is given different dimensions they can interact with
depending on how they want to view data: they can choose to filter the results
by “Year” and “Country.” The field “KEY” is most important, because it provides
the link between all other data tables.
This
means that whether the user is viewing all the data at once or only looking at
a specific country or year, they’re always looking at the same information,
just from a different angle. In this example, the star schema is most
efficient. Here, instead of recalculating the information every time a user
specifies a country or year, all the data is always readily available—it simply
displays different information according to the user’s specifications.
Perrot
estimates that 60 percent of the time he spent working on this project went
towards brainstorming how the end user would need to interact with the
dashboard application. This included ensuring that he had the perfect data
model to reinforce expected user behavior and to present the right information
quickly, clearly and accurately.
It’s
also important to think about how this data will be used. Here are some of the
questions Perrot suggests dashboard creators ask business users:
- Which Key
Performance Metrics (KPIs) will be most important for the end user?
- What story
are you trying to tell through your dashboard?
- Which data
supports that story, and illustrates it most clearly for the end user?
- What are
the common dimensions of analysis between my different data sources?
- Which data
is the user going to click on and interact with?
The “Take Action” screen, which provides users
with links to charities they can donate to in order to make a positive impact.
Develop the Overview
Screen First
When
looking at the overview screen—the high-level screen that summarizes the
dashboard’s most important information—the end user should instantly understand
what is wrong with his business (or, in the case of the Challenge, what is
wrong with the world).
Perrot
recommends building this screen first; when creating “Do You Realize?,” he
started with the “Overview” screen. Next, he builds additional screens that
will help the user better understand the significance of a given KPI: here, the
“Environment” and “Health” screens.
Finally,
he builds out a screen that allows the user to compare different data points
and build their own analysis. It’s important to work in this order, Perrot
says, for a few reasons:
- The
overview screen is essentially a high-level summary of the your most
important KPIs. If this screen isn’t clear, it means that you don’t know
what you want to explore and demonstrate with your dashboard, says Perrot.
- Since it
presents all the most important information in one place, the overview
screen is where the majority of your end user’s business questions will
arise from. You have to imagine all the questions the end user might have
after seeing the dashboard, and develop the other views to help him answer
those potential questions.
- Once the
overview is clear, Perrot says, the rest of your application and the story
it tells will come naturally. Without having this first, you’re likely to
develop overly complex, inconsistent screens—or worse, screens without
relevance to the end user’s business needs.
The Overview screen for “Do You Realize?”
provides a snapshot of the most important KPIs for the world as a whole and how
they have changed over time.
Create a Dashboard
That’s Both Attractive and Informative
As
with so many things in life, first impressions are crucial when it comes to
your dashboard. A well-designed, attractive dashboard is more compelling to the
end user—as long as it’s also functional. Your design choices can also inform
your data, indicating important things to the reader that help them understand
the greater message.
→ Use
color to convey change. Change
over time is a key component of almost any dashboard. According to Perrot, the
best way to help the end user quickly visualize this is to use color coding.
Assigning colors to trend lines or data points can tell you something about the
data: for example, green means the measure is moving in a positive direction;
red means it’s not.
→ Use
a four-column table to express change. Another effective way to represent change over time,
Perrot says, is to use a table with a time breakdown (usually by month) and at
least four columns, which contain: the actual value; the comparative value; the
percentage or value of change; and a color indicator (red or green).
→ Choose
function over form. If you have to
choose between form and function, lean towards function. A dashboard that’s
pretty but that doesn’t work right isn’t worth much to the end user. For
example, Perrot’s dashboard includes both clickable countries and a search
field to make sure users are able to quickly drill down to country-level
specificity.
Make Your App Easy
Enough for Anyone to Understand
Make
sure that the analysis your dashboard conveys can be easily understood by all
end users. Without having any prior knowledge of the subject, the user needs to
be able to look at your app and instantly understand what message the data is
trying to convey and what it means to them personally.
- Since
dashboards help business users make decisions quicker and more easily, you
need to create applications that lead the uninitiated user to the same
conclusions you came to. Perrot says his ultimate test is to show his
completed dashboards to his wife: “If she doesn’t understand everything,
it means something is wrong and I need to restart.”
- In this
Challenge, Perrot’s primary goal was to keep it simple. If you try to
include too much information in your app or clutter it with graphs and
tables, he says, the end user will get lost. Here, it was also important
to clearly and directly tie the KPI data to specific actions the user can
take to make a difference.
- When
building dashboards, Perrot recommends getting straight to the point.
Display the most interesting data the most prominently—and, perhaps most
importantly, use visual elements that help users understand where the
numbers come from and how they evolve over time.
Perrot’s dashboard is simple and to-the-point:
important information is displayed prominently, charts and graphs convey change
over time and pop-ups provide a summary of relevant data.
While
these are good general guidelines, don’t be afraid to break your own rules: in
the end, always let the data dictate your dashboard. For example, when
developing QlikView applications, Perrot usually puts the dimensional filters
of the left side of the screen.
For
the Challenge, however, he wanted to use the left side to navigate the
different sub-menus as well as to access outside sources like video. There’s
always room for innovation.
Following
these best practices helped Perrot win first place in the QlikView Open Data
Challenge—and you, too, can use them to create more effective and actionable
dashboards that inspire your users to take action.
