“Powerful?” Yes, you read that correctly. This article centers around powerful BI, and although we’ll discuss Microsoft’s Power BI platform in this larger context, our focus here will go beyond just the tooling.
The point of this discussion is to understand the following fundamental principle that will help you get the most out of business intelligence:
Making better, data-driven decisions requires more than just pretty dashboards or a very robust BI tool such as Power BI.
Truly empowering a proper business intelligence function in your organization requires the mindset, culture, education, and thought leadership to actually allow a platform like Power BI to have its full impact.
If you take time to fully grasp that principle, it will ultimately change your outcomes in a powerful way.
Additionally, we’ll address the unsung heroes that make great dashboards possible – fundamental data management with, in this particular case, an emphasis on data engineering and modeling.
Let’s talk about a typical scenario. If you’re like many folks who are beginning their journey down the business intelligence (BI) path, you’ve heard of Power BI.
Given Microsoft’s position in the Gartner Magic Quadrant for Analytics and Business Intelligence Platforms (No. 1 in both Ability to Execute and Completeness of Vision), it’s not a surprise that this would seem like the logical place to start.
You do some Googling, you see you can download Power BI for free, or if you are like many shops that use Office365, you see that it’s already available to you as part of the Office365 SKU.
Your inclination might be to download it, connect to some of your more commonly used data sources, maybe take some training from Pluralsight or just watch some videos on YouTube to see what you can figure out.
Also likely as you travel farther down this road, at some point you’ll realize that dabbling in design work only gets you so far. You'll choose to reach out to a professional services (consulting) organization to better understand the right way to build dashboards or roll out BI at the departmental or enterprise level.
When you reach out to a consulting firm, you’ll probably receive one of two types of assistance.
The first type is where a firm will simply do as you ask -- perhaps build some dashboards or help improve the ones that you have developed. This is basically “answering the mail,” which is OK, but misses a much larger and very important opportunity for you and your organization.
The second and preferable type will provide you with more consultative advice that should go something like this:
The key to building powerful dashboards is to really think through how you make decisions, where data could positively influence the outcomes of those decisions, and let the information needs required to answer key questions or validate key assumptions drive your BI implementation.
This is the road you want to take, simply because this is where you’ll see results. So let’s talk through this more mature path with this idea as the starting point.
The Mature Path to POWERful BI
As stated, key questions and information needs are the factors that should drive your business intelligence. Typically, these will take the form of a user story.
Here’s an example: “As a [role name], I need to understand [answer to key question] as measured by [key metrics] by [various dimensions - could be area, department, month, etc.].”
Facts: Measurements and or metrics that can be aggregated by which dimensional attributes are analyzed.
Dimension: “A category that can be used to arrange data by facts and measures for data dicing (grouping) and slicing (filtering) purposes. Commonly used dimensions are people, products, places, and time.” Ultimate Glossary of BI Terms
In lay-person terms, facts are something that can be measured and summed up for analysis by different variables. The variables by which they are summarized are call dimensions. If you look at a credit card statement (at the transactional grain), you aggregate dollars and number of transactions (facts) BY various dimensions (by store, by category, by date, by geography).
It’s at this point where some of the biggest (and often not understood) challenges will lie: mapping information needs to data needs and the provisioning, preparation, modeling, and supplying of that data for consumption by the BI tool.
Modeling and data engineering comprise the vast majority of “work” when it comes to producing powerful BI. In fact, without that, your ability to produce meaningful dashboards is extremely limited.
Over the years we’ve seen multiple legitimate, well-respected BI platforms completely shelved because they were seen as low value due to the fact that the underlying data was never properly prepared. Rather than acknowledging this was a data engineering and modeling issue, it was easier to disparage the tool itself as a piece of junk rather than do the hard and honest work to prepare the data.
In this sense, I have referred to the BI tool as the “shiny red caboose” on the train. If you aren’t old enough to catch this reference, the caboose on older trains tended to be the only train car with real color and would stand out when looking at a train; but if the engine, coal car, dining car, and transport cars were not in place, the caboose would just sit, providing little value relative to the real purpose of the train itself.
These other cars can be likened to the business case, user stories, data quality provisions, data modeling, and data engineering needed to make your BI powerful and of high value for your organization.
What Do We Mean by Powerful?
Let’s turn our focus to mindset. BI, for better or worse, grew in large part out of more traditional reporting needs. Early on, reporting and BI were fairly synonymous. Reporting was BI. BI was reporting. But not anymore.
The result was that early attempts at dashboards typically looked to simply re-create static columns and rows reporting in the new BI platform.
That approach obviously doesn’t take advantage of the purpose or features of a BI tool to allow the end user to truly interrogate the data. Interrogate may seem like a strong word, but let’s consider it for a moment.
To merely report on data and expose it in a static way does not lend itself to active interaction with the data itself. Properly modeled data exposed in the right types and combinations of interactive visualizations allows a user to cross-filter and drill into data to truly ask questions of the data.
So maybe interrogation is too strong of a word, but the key here is to take a more aggressive stance towards changing the mindset of your organization when it comes to really understanding what is behind your data.
If you want to get your analysts off the dime to truly move them from a reporting mindset to interactive data mining, using a stronger word can be useful to get them to think differently.
So, what do we mean by powerful? It’s a combination of producing well-designed dashboards against properly modeled data to allow your end users to:
Many organizations fail to take advantage of these new features either due to a lack of understanding or appreciation for the data prep work that is required. As a true consultant in this space, Onebridge wants to ensure you are best positioned to get the most out of your BI platform, be it Power BI or any other.
“Powerful BI” is not the result of just the technical preparation of the data and dashboard design. Your organization’s culture and attitude toward data management as a whole can make or break the best of your technical efforts to roll out BI in a meaningful way. So don’t brush aside this critical component.
Culturally, there needs to be an appreciation for the power of data. At the same time, the power cannot be held by an elite few, but rather democratized to a larger group of knowledge workers who truly understand the data.
Having the right leadership who appreciates this fact will go a long way to ensure the appropriate conversations are taking place between the business and IT.
This, in turn, helps you achieve the proper balance between governance and democratization to responsibly develop data assets needed for powerful BI.
From an education standpoint, covering some basics and achieving mutual understanding around terms and “data thinking” should all be part of what is often referred to now as data literacy campaigns within your organization.
This goes beyond even your analysts and IT custodians. This should include groups like legal. It is hard to determine the ethical implications of building and applying a certain model if there is no understanding of the concepts of feature selection and feature engineering, particularly when informed by BI activity to help build or develop said model.
Education will go a long way to ensure both proper data-driven decision making and managing the risk and legal implications of how those decisions were made.
OK, we’ve covered the fact that there are a lot more considerations than just building a Power BI dashboard when unleashing the true power of BI. (See what I did there?)
There are definitely best practices for visualization and dashboard design, and we’ll share a few tips and techniques here. But this is part of a much deeper discussion around how your brain processes data and the most effective way to communicate through the use of different types of visualizations and their combination in a single dashboard.
Subject your visualizations to a simple initial test I call the “back-of-the-room” test. A visualization should guide your eyes and brain to the right bit of information that requires further exploration.
Rows and columns spreadsheets communicate nothing when viewed from the back of the room, where you cannot read the actual data values. Visualizations should be designed such that when viewed from across the room, your brain knows where to start looking. Outliers, clusters, trending, the use of color for “good” or “bad” – all of these can allow you to get you focused on the right things to interrogate.
Use of color is an interesting point as well – ensure that the same colors mean the same things on a given dashboard to help minimize confusion when trying to tell a story with data.
One other concept I’d like to introduce here is the idea of what I call “overloading” of dimensions or facts. To truly perform analysis, it is often helpful to devote dashboards to multiple visualizations that all share the same fact or measure to allow you to interrogate that fact by multiple dimensions. I call this “overloading” a fact.
The same can be done with dimensions – pick one dimension to interrogate by various facts or measures. This, too, can be very powerful when doing deeper level analysis.
There is no question that storytelling is the most powerful aspect of true power BI.
As humans, our brains naturally gravitate and lean in when compelling stories are being told. They provide context and help us develop a greater appreciation of the story elements that come up along the way.
Using dashboards and visualizations to actually tell a story that explains the “so what” to the end user or ultimate audience allows you to use data to influence and persuade when trying to make a compelling case or call to action.
Many BI platforms include the ability to create stories, leveraging visualizations and dashboards along with callouts to lead your audience through a narrative.
Power BI actually has a narrative “visualization” that creates a narrative for you from your data. Very powerful.
Tableau has three levels of content – visualizations, dashboards, and stories; but sadly, stories are often not used by organizations. This is a mistake. Becoming a good storyteller with data should be an aspirational goal for any individual, department, or organization as a whole when it comes to truly unleashing the power of BI.
Power BI is indeed a compelling platform. Weekly releases of Power BI as a service have led to a rapid expansion of the capabilities.
While Power BI played catchup with various other vendors in the space for a period of time, it has been recognized in Gartner’s Magic Quadrant for the last 13 years now and offers some of the most advanced capabilities of any BI tool.
The availability of new visualizations and augmented capabilities in the marketplace gives you the opportunity to go beyond traditional graphs and charts to incorporate visuals around “key influencers” to highlight correlations, utilize more advanced techniques, such as PCA for clustering of data with higher dimensionality data, and pull in natural language narratives to direct action.
Power BI also has data engineering capabilities that allow you to pull and model your data in the platform itself, while tracking any data manipulation steps in an easy-to-understand interface to better understand how your data has been transformed. The DAX language is very powerful, allowing you to create custom measures, while in general many of the types of calculations you used to have to do in DAX are now just part of the standard features of the tool.
While there are some limitations working with the desktop/local version, and Power BI as a service is only available on Microsoft Azure, the platform offers a great way to get started and apply powerful BI thinking to a very robust BI tool.
At Onebridge, we want to help with your journey regardless of where you are. Whether conducting training, doing development work, assisting with dashboard design and data preparation, or focusing on larger data management and organizational change management challenges, we want to be your partner. Please reach out us today with questions about how you can get the most of Microsoft Power BI or to have a conversation about your goals and challenges.
Have more questions about Power BI? Contact us