Dashboard: - An autopilot for decision making, make sure you get it right.

Typical Monday morning boardroom scenario....

Suit 1: I would like to see the performance results for this quarter and the possible trends.

Suit 2: Yeah, for sure! We do have few data guys we can call and get that done.... Then on Thursday... Suit 1: Oh these visuals are incredible :-) Thanks! "Clearly" Let's make the target on campaign XYZ next quarter.

Well, above is a very common scenario. In many organizations, data is driving decisions and the best way to look at the data is through the dashboards. If you ask the question, what exactly is the most important word in above conversation. Its clearly 'Clearly'. We want our viz to be clear- We want them to provide clear vision.

What goes behind to get there is a very complex procedure.

Following are the few guidelines which will lead you the dashboard which can give a huge round of applause in the next meeting.

1. Who are you presenting to? ( The dashboard User)

The audience for your dashboard is the first and most important aspect of any dashboard design. If your audience is the board room director, the space and complexity of the dashboard should be minimal. While if you are designing the dashboard for operations, for example, an engineer in the production environment, the dashboard requirement, and complexity changes significantly.

The complexity of the dashboard is inversely proportional to the position in the organization.

For example, CEO doesn't time to adjust the parameters or filters to reach to the specific data goals. The dashboard, in this case, is going to be most static with the exact data points indicated, which leads to the decisions. This leads to our next point, business question.

2. What are the business questions? The overall and point in time.

Always before looking at the data or before preparing the dashboard, it's a good idea to write the business questions you are trying to address in the dashboard. It's just like the software development. Your business questions are requirements for your dashboard. Indicative of the knowledge that can be highlighted on the dashboard.

There are two types of business questions depending on the situation in the organization

Organizational Goals business questions: These are overall goals of the organization e.g. what is my revenue this year? What was the scenario for next year looking like? Which campaigns did the best?
Short term business questions: how did my XYZ campaign do last year? What were the changes in sales over last month?

3. What are the sources?

It is true that 80 percent of the time in setting up the data for the dashboard and only 20 percent of the time goes in developing the visualization. After figuring out the business questions to answer, always go for sources. It is very important to understand that sources can be very deceptive.

Always consider only those sources which are answering business questions ( Write down Pseudo logic using data fields)
Avoid overload in data loading for the tool ( Do not import unnecessary fields)
Obviously, include all the necessary fields.
I prefer preparing my data calculations using the views, before loading into the dashboard.

There are two ways to configure the source,

Perform pre-processing of the data on the source data itself and then load in the dashboard. Dashboard won't has to perform many of the calculations. Like mapping of the county code to the county, the name can be done in source data. This will reduce the load time for the dashboard. This is used only when you do not expect to replace the data source much often.
Perform the processing of the calculations on post data (Inside the dashboard development tool may be Tableau or Qlikview) This is recommended if you replace the data source pretty often. This will reduce the overhead of writing all those queries in the staging again and you can easily replace data source with desired year.

4. What is the best visualization to use?

This can be very tricky, you should always know which visualization to use based on the requirement of the client or business questions you are trying to answer.

Comparing data side to side - Bar chart
Combine absolute and relative values - Combination chart
Gauge - indicate ratio or proportion
Pie - ratio to the total amount
Scatter Plot- Display correlation of the measures
Tree Map -Display hierarchical data
Geo Map- Geographical location or regions
Line Graph - Trend over time

5. Validations and obvious unknown

Always perform basic validations using excel or any other device after implementation of the visualization. Random selection validation trick: The trick to validation is to take one or two random data point on the dashboard and trace it back to the source sheets or database. There is no greater sin than showing wrong values in the graphs and charts. Sometimes while giving conclusions it is very important to consider all the parameters and obvious unknowns. For examples, You see the sales rise by one of the campaign started on specific data but ignore the fact that the day was the holiday and average sales on that day are lesser than any other holiday.

6. Controlled parameters/filters

Depending on the user you are going to discuss the parameters/filters which are controlled by the users. If its executive user, the number of parameters should be very less. If its operational user, he/she can use more parameters to get to the very specific data points on the dashboard.

Make sure that your parameters/filters are controlling all the graphs. The database are connected accordingly with proper key matching. For example, Your sales have year and so does your profit table. Now both sheets have a different year. You have not connected those internally. You are using the parameter as the year from sales table. Make sure that you connect sales and profit using year as a key so that all the data is filtered accordingly.

7. How do you like it?

Dashboard development cannot be one simplified standalone process. It takes iterations. It's always good practice to define a prototype with visualizations and take the opinion of the audience at every stage as if they would like to see what is shown in the graph. Understand the data constraints and feel free to communicate the same to the clients. Sometimes clients might not have the clear understanding of the data constraints.

Follow these 7 rules and you can be the next hero in the strategic decision support modeling.