Skip to main content
cancel
Showing results for 
Search instead for 
Did you mean: 

The ultimate Microsoft Fabric, Power BI, Azure AI & SQL learning event! Join us in Las Vegas from March 26-28, 2024. Use code MSCUST for a $100 discount. Register Now

Greg_Deckler

Power BI Usage Models in Pictures!

Introduction

Power BI provides a huge amount of flexibility when it comes to how you implement and adopt Power BI within an organization. There are nearly limitless ways to organize and utilize the various components of Power BI, including workspaces, apps and dataflows as well as how information is shared and distributed. So, I have been using the term "usage model" of late to refer to this overall adoption and governance architecture. To be clear, I don't know what the proper term for this is or even if one exists and; big surpise, I don't really care, but since a data model organizes your business information into a usable format for reporting, it made sense a usage model would organize the various  components of Power BI into a usable form for your business. Think of it this way, a usage model encapsulates how a business adopts and utilizes the various components of Power BI in order to provide governance and process around the adoption of Power BI within an organization.

With the laborious explanations out of the way, let's dig in and take a look at some various usage models. And, since everyone loved Power BI Licensing in Pictures, lets do it with pictures! And finally, to be clear, what I am showing in no way reflects all of the possible usage models that exist, because there are nearly limitless variations. However below are some of the base ways that Power BI can be adopted and used within an organization.

Anarchy

Many Power BI implementations probably start quite simply as near chaos or anarchy, the grass roots adoption of Power BI by end users.

anarchy.png

In the Anarchy model, end users just start connecting to data sources, building datasets and reports and sharing those reports willy nilly with one another. It's grass roots adoption at its finest but likely gives professional data analysts and Chief Data Officers (CDO's) the heebie-jeebies and keeps them up at night. CDO's and their ilk worry endlessly about trivial things like data lineage and stewardship, single source of truth and other equally unimportant matters (I'm joking) and Anarchy is pretty much the antithesis of those ideals.

But, the important thing to keep in mind here is that the chaos existed long before Power BI, it was just done in Excel and via email and file shares. So, the chaos has always been there and will always be there. Power BI just enables this chaos to swirl in a more refined state with better tools and visibility. The dirty secret is that swirling under the veneer of almost every Power BI implementation lies the ever present Anarchy usage model and that's a good thing. Anarchy represents the business analyst, citizen data analyst and citizen data scientist. It is a key component to building a data culture within an organization.

So, embrace the chaos, leverage Power BI to provide your users methods and procedures by which they can surface their discoveries within the organization. This usage model is good for very small businesses although even the largest enterprises will also have this usage model occurring within their organizations (whether in Power BI or somewhere else).

Centralized

Perhaps at the other end of the scale is the Centralized usage model. The Centralized usage model is more in line with having a central data analysis and reporting team (read IT) that develops all reports within the organization. It is the centralized reporting structure most organizations had prior to the advent, or even thought, of self-service business intelligence.

centralized.png

Under the Centralized usage model, a central team (or teams) of professional data modelers and report builders create the data and report assets using Power BI Desktop and publish these to a workspace or multiple workspaces. In this version of the Centralized usage model, the central team then shares out the reports and dashboards with end users and groups of end users. This usage model is good for small and mid-sized businesses that have dedicated data and report creation personnel but are too small to have data marts or anything like a true data warehouse.

Centralized - App Sharing

A variation on the Centralized usage model is to utilize Power BI Apps.

centralized-app.png

Not much needs to be said here, this is the Centralized model but Apps are used for sharing with end users instead of direct sharing of reports and dashboards. Apps provide a method of bundling dashboards and reports together for distribution as a single link for end users and thus provide a different user experience than individual dashboard and report sharing. In addition, Apps provide a central point for security rights, which can assist in governance. The downside is that Apps can lead to Workspace bloat as there can only be a single App per Workspace. This usage model is suited for the same organizations as the Centralized usage model, small and mid-sized businesses that have dedicated data and report creation personnel but are too small to have data marts or anything like a true data warehouse.

Golden Dataset

While the Anachy usage model represents a business-centric usage model of Power BI and Centralized represents an IT-centric usage model of Power BI, the Golden Dataset usage model represents a partnership between IT and the business where IT is responsible for those boring, tedious topics of data lineage and stewardship, single source of the truth and yawn and on while the business is responsible for report development and data analysis.

goldendataset.png

The Golden Dataset usage model represents a nice partnership between IT and the business. IT worries about all the boring, tedious stuff and the business creates the reports they need because they actually, you know, know their business versus are just technology wonks. It's a nice partnership that leverages the various strengths within organizational teams. This model has IT deploy a dataset or datasets to a centralized workspace or workspaces. These datsets have curated data as well as calculated columns and measures that adhere and encapsulate business rules. Business users can connect to these datesets to create their own reports and then publish these reports back to the workspace. Who controls the sharing of these reports with end users can be either IT or the busines or both. This usage model is best suited for upper middle market businesses and small enterprises.

Golden Dataset - Keep 'Em Separated

If the close commingling of IT and business is perhaps just a tad too unsavory and vomit-in-the-mouth inducing for either side to bear, a minor variation can perhaps alay all fears of unholy unions being formed. This variation of the Golden Dataset usage model keeps IT and business at arms distance while still preserving the essential partnership between the two.

goldendataset-variation.png

In this variation, datasets are allowed to be shared across workspaces. Thus, IT can keep a death grip of imperial totalitarianism on their workspace while the business is freed from the tyranny of evil men within their workspace. It's a virtual fist bump versus a hug between IT and the business, the COVID-19 greeting equivalent version of IT and business partnership. As with the Golden Dataset, this usage model is best suited for upper middle market businesses and small enterprises.

Data Warehouse

Within larger mid-market businesses and enterprises, Power BI ceases to be the only game in town and is often coupled with a data warehouse. For businesses that; even after 19+ years, have failed to have their IT systems make it into 21st centry (just saying), that means an on-premises data warehouse.

datawarehouse.png

This is very similar to the Golden Dataset approach except that the development of the "golden dataset" does not occur within Power BI but rather within more tranditional systems like SQL Server, SQL Server Integration Services (SSIS) and SQL Server Analysis Services (SSAS) or Teradata, Oracle, etc. Otherwise, the concept and approach is pretty similar. This usage model is best suited for large mid-market and enterprise organizations that have ancient, almost archaic technology stacks that include mainframes, Facits and Commodore 64's. Don't get your shorts in a bunch, take a joke people.

Data Warehouse - Cloud

For those IT organizations that are not mired in a "job preservation" mentality and actually want to help the business succeed versus clinging to lifeless, soul-sucking, monotonous jobs patching servers (still joking people), their data warehouses are in the cloud these days.

datawarehouse-cloud.png

Same concept as the Data Warehouse, just post Abacus.

Reality

Ah man, reality? Who let that guy in? Major buzz kill. Well, in fact, we must address reality because there are literally endless variations and hybrids of the usage models listed in this article. And we haven't even covered the use of dataflows or row level security for example, which provide variations on our usage models. But, given the sheer insane flexibility of Power BI, it is a futile effort to even attempt to list all of the possible variations. I mean 7! is already up to 5,040 different possible combinations/variations. So, too long for a blog article to cover them all. But, here is one particular hybrid usage model that I have found to be a successful formula within real world businesses, specifically within the mid-market and enterprise spaces.

hybrid.png

This usage model is a hybrid of several of the usage models already discussed but at it's heart is the Golden Dataset, Keep 'Em Separated usage model. There are a number of highlights, including:

  • Certified, “golden”, datasets are created and published to a centralized workspace. These datasets allow for the standardization of KPI’s and measures to ensure that these are standard across the company. This promotes good data governance and stewardship. Only data modelers will have access to this workspace.
  • There will be a main workspace for each department, such as Finance. Official reports will be built from the certified dataset shared across workspaces. Multiple official reports can be published here and packaged up and displayed via a published App.
  • Each department or sub-department will have an additional workspace created as a “playground” area. In this area, reports can be created from the certified datasets shared across workspaces or new datasets created by the business.
  • Workspace access will be controlled via Office 365 groups
  • Datasets can employ row level security (RLS). The RLS model utilizes roles whose membership will be Office 365 groups.
  • Column level security can be implemented by having a separate, related table that only contains the columns that need to be secured. RLS rules can then be implemented on this table.

This is a fairly common real-world usage model that I have seen used or I have recommended for Power BI in a wide array of organizations. The key here is that most real-world implementations of Power BI will have a usage model that is a hybrid of the usage model archetype discussed within this article.

Conclusion

Power BI is fantastically flexible and adaptable when it comes to how organizations can adopt and use the wide array of functionality that comes with Power BI. How organizations deploy, utilize and combine these features is called a usage model. There are nearly endless variations of usage models and real world implementations will often be a hybrid of different usage models. Having a good, defined usage model is a key component to the governance and adoption of Power BI within an organization.

Comments

Hi Greg, excellent images and exposition of scenarios, very very useful. Thanks!!

Anonymous

Great article. Thank you

We started using PowerBI in a Self Service mode only a few years ago and our Model for that completely resembles the Anarchy model.  Now we are moving to PowerBI for Enterprise and our overall model we want looks very similar to what you have described as Reality.  Our challenge will be trying to move our current Self Service Anarchy model more in line to what we want in the Reality Model for both self service and enterprise.

@Greg_Deckler  Fantastic! OK with you if I share this with my customers and partners, as long as I give you all the credit?

@Greg_Deckler great post!  I completely buy into the Golden dataset setup and I've recently been considering a Golden dataset consisting of multiple mini Golden datasets joined together, using direct query to the PBI dataset, thus allowing further separation of models.  Would be interested to hear your thoughts on this approach if you have any or have pointers you could share? 

The current pain point I'm finding in this approach is unifying the Dimension tables common to each model and subsequently joining the related dimensions - a Golden dataset of these dimension tables is something i'm considering to help overcome the issue.