Power BI puts a veritable army of powerful data visualization and reporting tools at your disposal. But how do you get this army marching in the same direction, working in concert towards the same unifying goals?
When it comes to business intelligence (BI), growing your data architecture is easy. It’s growing your data architecture intelligently that is the real challenge. This means setting yourself up for long-run prosperity and tranquility-of-inbox—while avoiding a frantic cacophony of short-term solutions piled atop each other till they reach the very heavens (also known as “where letting inertia steer the ship will get you”).
Good data architecture is about setting up a vast system of systems, allowing data to flow from the snowpack of data sources through a slew of cleaning and organizing tools. Until finally, data trickles out in the form of well-crafted reports that actually help the people around you make decisions.
This system of systems is going to include things like:
- Source Systems (ERPs, CRMs, and other abbreviations with exactly three letters)
- Master Data Management (MDM) Tools
- Information Logistic Tools (Data Lakes, Vaults, Stores, and Warehouses)
- Enhanced Semantic Querying (Various Cubes and Tabular Models)
- Reporting Layers (Power BI Service, Tableau, SSRS, and of course…Excel)
- Types of reports (Strategic, Operational, and Ad-Hocs)
- End-User Personas (CFO, Department Head, Sales Lead)
That’s a lot of interrelated things to understand at once. However, any successful data strategy is built on top of a rock-solid foundation of deeply understanding your data domain. The first step towards getting there? Funny you ask…that would be a data assessment.
Why You Need a Data Assessment
Put simply, a data assessment is the first step in any successful data strategy because it provides the required 30,000-foot view. The assessment allows you to see how these already complex elements are either working or not working together so you put clean and useful information into the hands of decision-makers.
Building this 30,000-foot view helps you understand individual problems in terms of how they fall within the larger system. With a data assessment, you will spot—rather than guess—downstream effects and also get a feeling for when the problem you’re thinking about is better served if it is solved at a higher or lower level.
In short, with a data assessment, you get perspective.
How To Set Up a Data Assessment Framework
Lists are great for linear, well abstracted, task-oriented thinking. This is the opposite of the water we are about to wade into with a data assessment framework.
We’re going to harness the human brain’s incredible geospatial processing abilities by representing our system as a map. This map will start as a tool you can bring up on a screen to understand your data as a whole and locate both problems and opportunities within it.
Over time your mind will start to recreate this map internally, so you carry it with you wherever you go and pull it up in any conversation to see how potential changes or issues might ripple across its surface.
Resist the temptation to start with Visio. Instead, grab a sharpie, find some multi-colored post-it notes, and get to work on a large wall (or a very big whiteboard).
Go through this exercise even if you already have a data map—the tactile knowledge is as valuable as the final artifact. Then, follow these data assessment steps to make sure that nothing falls through the cracks:
Score your sources (on a scale of 0-5) by estimating the value and complexity of the data they contain. Identify the type of database for each dataset and whether it’s on-premise or in the cloud. List them all on the far left side of your map.
Write down existing master data management systems, each on one post-it. Then, list out software applications that use the mastered data (these will be duplicates of the source system post-its) and note how well-mastered the data is. If the answer is none, that’s OK.
Take inventory of existing information logistic structures (i.e. data lakes, vaults, stores, and warehouses). If multiple systems of the same type exist, list out those considered formal to a certain degree (i.e. IT and sales may both have a data lake) and note which one(s) are more trusted.
Map out existing enhanced semantic querying layers (i.e. SQL Server Cubes, SQL Server Tabular Models, Business Objects, Hana Cubes). Give each a score for its current usefulness, trustworthiness, and maintainability to identify those that are no longer relevant.
List all the existing reporting systems. At a minimum, you’d have Excel. Other tools may include Power BI, SQL Server Reporting Services, Tableau, Qlik, Cognos, etc. Score each system based on how many people are using it, whether the users trust it, and if they like using it.
Write down 5-15 report types currently in use, such as C-suite finance reports, one-off specific question reports, operational reports (i.e. list of customers to check in with), and data science reports (i.e. information on fraudulent transactions).
Map out decision-maker personas and place them near (but not at) the very far right of your map. You should have at least 3, but to start, try not to have more than about 10.
Add post-it notes listing the pain points of the current reporting infrastructure, such as missing information, untrusted information, stale information, conflicting information, and slow or ugly reports. Tie each one to the persona it most affects (even if it impacts multiple personas).
Include post-it notes for near-future pain points and write “future” at the top. Then, place all the post-it notes with pain points at the far end, on the right side of the personas. Start with as many pain points as you can come up with and then widdle the list down to 10-20.
List items (i.e. analytics software) that you don’t have but think you might want to add. If you don’t know what a product is, just write “something that does X.” If you don’t have any, that’s fine.
Take a break. Not a “five minutes of Outlook and Twitter break” but a real “90-minute walk around the park with your phone on silent” type of break.
Ideas will come to you over this break and that’s fine. Don’t try to record them or sort them here—just consider them and let them float back into the ocean of your consciousness. If they’re important, they will be there tomorrow. If space allows, go home and get a good night’s sleep to let all the concepts your mind is juggling start to organize themselves.
Map out your top three data strengths and top three data weaknesses, including a brief description of each. “Brief” is the word here, no need to try your hand at James Joyce-style prose even if you can write small enough to fit it on the post-it.
If you can’t think of specific data strengths or weaknesses, that’s fine. But in that scenario, ponder the question over the next couple of days and see if any ideas bubble up. My guess is you’ll have plenty—and the tricky bit will be getting the list down to three and three.
At this point, take some high-res pictures of your data map. Discuss the discovery with your team and transfer the information to Visio so you can reinforce your mental model while creating something more durable and sharable.
Congrats..now you have a data map on which to build your data strategy! This will be your 30,000-foot view that will both help you understand where you are, understand where you need to go, and see how the impact of given choices might ripple across your organization.
Need a hand with your data assessment and data strategy? The CSG Pro team will help you evaluate the maturity of your data landscape and perform a thorough gap analysis to identify weaknesses and opportunities. Check out our data strategy services.