News

How to Escape the Data Bottleneck : Part 2 of 3


BUILDING A HIERARCHY OF DETAIL

In the last article (click here to view), we discussed how to begin the process of escaping the Data Bottleneck. Again, what we mean by the Data Bottleneck is the twofold challenge of:

  • Timeliness – Not having access to data fast enough – the data request process is slow and drawn out
  • Completeness – Not having enough of the data you need – or, incomplete datasets

Part 1 explained how to build the foundation. Clear communication and collaboration are integral to developing a common understanding of the data needs of your business. Taking the time to define your data environment is crucial to creating a proactive data analysis environment, rather than one that is reactive. Finally, we discussed how being strategic in how the data is modeled will increase efficiency and allow for cross referencing.

Now that the foundation has been set, what can you do with your data?

You can start by building a:

Hierarchy of Detail

This means, your data should be presented in a logical, hierarchical manner.

Think of a data set comprised of geographical data. At the highest logical level, you have countries. Underneath, states or regions. Next, cities… and so on. These data points are all related and fit into a hierarchical, or tree-like structure.

This hierarchy of detail applies to both how the data should be modeled, and how it is presented. In this article we’ll focus on the latter.

 

High Level Detail with Interactive Dashboards

Probably the best way to display high level detail is in the form of dashboards.

Dashboards are typically graphical user interfaces with groupings of visuals and reports displaying high level company details. For example, a sales manager’s dashboard displaying several visuals of key performance indicators for their team… organized by agent, product group, geography, etc.

So, what are some characteristics of a good dashboard?

  1. Multiple “big picture” views and graphics displaying top areas of interest
  2. Clean and simple design – low to no clutter
  3. Pre-attentive attributes used to show outliers, trends, or important data points
    • Pre-attentive attributes are methods of conveying meaning without the user needing to think, read, or analyze much further
    • They can be thoughtful use of color, size, length, spatial grouping etc.
    • So, green = positive, larger shape = larger quantity, and so on
  4. Interactivity through parameters, filters, or visuals with buttons etc.
  5. Ability to drill down to lower levels (see next section)

Dashboards are not graphics without a purpose… but are used to simplify the data analysis process. They are incredibly helpful from a logical standpoint because high level details are shown first and the user can start at the top of the hierarchy of detail and then drill down.

Pre-attentive attributes draw the user’s attention to important data points, and areas to investigate. This in itself increases efficiency because it highlights those outliers and trends.

Through the interactive elements of the dashboard, self-service is possible for users, thus greatly improving time spent on analysis. Customized parameters and filters can be applied to all visuals shown on the page, so there is no need to put in a new report request. Set the exact date range, or region, or product types you wish to analyze, and the data is instantly filtered.

Typically, there are multiple reports all in one place which can be related to the same subject area but showing different views. This allows for a more complete picture and comparison of the different visuals.

There are also multiple uses of the dashboards since the interactive elements can change the purpose of the report. In one instance you could be looking at numbers on a quarterly basis, and by changing one parameter you could change it to look at numbers annually for example.

 

Specific Detail in Drill Down Reports

From the dashboard, we move down the hierarchy of detail to more specific reports. The user identifies an area needing further investigation and chooses to drill down through that visual to view more specific data.

To illustrate:

  • Let’s say you have a dashboard showing the activity of a customer service team
  • One of the visuals on the dashboard shows a large spike in activity and open cases last week
  • By clicking on the visual, a more detailed report is shown on the screen
  • This drilled down report has more information on specific clients that required more attention from the customer service team

On the top level of the dashboard it would be difficult to see the specifics because, with hundreds or thousands of clients, the visuals would become too crowded to make any sense. Instead, the spike in activity was identified, and the visual was selected, and the drill down report displayed with the details.

You could continue investigating multiple levels of detail… drilling through the hierarchy gradually getting more granular. Usually the lowest level available would be the raw data itself. Through this structure you can investigate to the exact level of preciseness needed to answer your business question.

This hierarchical structure will certainly have a positive effect on the speed of analysis. Having the ability to drill down to lower levels of detail instantly, and not need to wait on additional report requests makes a difference. Also, from a logical perspective, it makes sense to start with the more general and work down to specificity, rather than the other way around.

Because the levels in the hierarchy are all related reports/visuals, you as the user can cross reference the data points and get a more complete picture of the business question at hand.

With raw data available at the lowest levels, you can validate the specifics that caused certain outliers or spikes in a trend. If presenting your findings to executives or your internal team, you can show the full story.

 

HOW DOES THIS HELP?

Let’s summarize how building a hierarchy of detail helps you address the challenges of Timeliness and Completeness.

Timeliness

  • Dashboards simplify the analysis process by showing high level information first
  • Through the use of pre-attentive attributes, dashboard visuals can instantly draw attention to outliers, trends, and other areas to investigate
  • Dashboards with interactive features allow for a certain level of self-service, instead of waiting on individual report requests
  • The hierarchy provides users with instant drill down/up capabilities, again with no need for additional report requests

Completeness

  • With a well-designed dashboard, you have access to multiple reports all in one place
  • Since there are different filters and interactive elements available, there are multiple uses of these reports and dashboards – in other words, you can change the purpose depending on your analysis needs
  • The reports and underlying data are all related which allows for cross referencing analysis
  • At the lowest level, you also have access to the raw data and so can truly see the complete picture

 

WHAT’S NEXT?

The hierarchy of detail and the use of dashboards are an excellent way to handle your high volume, frequent, repeatable data needs.

What about more complex questions, or less frequently used parameters that come up in the future? What if there are non- traditional data sources? There could also be unique business questions related to particular initiatives. These would likely not be covered in your traditional reports and dashboards.

In Part 3, I will share thoughts on how to provide business users the versatility to analyze data on an ad-hoc basis further eliminating the data bottleneck.