Concepts to know before reading this article: Business Intelligence, Data Model
Concepts covered in this article: Why, How, What, Analysis, Synthesis
New Here? Read the Post that Started it All, and check out the Learning Map
We’ve lived for a while now in the ‘how’ of business intelligence. How do we go about organizing our data into a model? What are the pitfalls to avoid, and opportunities to take advantage of? Those are good questions to ask, but it’s actually the ‘why’ of learning that feeds me. Why make a data model? Why do any of this? I want to start with an illustrative story, and then we’ll get to the lesson.
A Personal Story About Alligators – Why Build a Data Model
Remember, when I started at Source to Share I had just left the nonprofit sector, where I had run operations at an arts nonprofit for 7 years. During my time there, I learned many lessons, some of them the hard way. Simply put, our staff was overworked and underpaid. Furthermore, no matter how clearly we dreamed about renovations, updates, and expansions, the day-to-day grind seemed to always get in the way.
I remember when we wanted to switch internet providers to save money, and I had to figure out which of the cables criss-crossing nightmarishly around the building were part of the alarm system, the old internet system, or the CCTV monitoring system. What followed was a herculean effort to categorize, cull, and label the remaining cables. I created a corresponding cable map with the hope that some future arts-lover-turned-administrator would silently thank whoever had come before them.
I watched this story play out again and again while I was there. Dozens of kids were arriving to fill out audition forms for an upcoming show, and there were no pencils. I would run to the store or, if I was too busy with some other emergency, I would call someone to run to the store for me. Then, later, after a long day, we’d try to get down to the business of making sure we never ran out of pencils again. Where do we keep pencils? Who tracks them? Who purchases them?
I remember in the early days when we would put on a show, we’d put a call out to the community for musical directors, choreographers, and stage managers, and when people would ask what the pay was for the job, we’d have to tell them we don’t know. We didn’t know how many students would enroll, and we didn’t have a savings, and we’d only know how much money was left when the stage was struck and the dust settled. We’d try to give the prospective theatre crew an estimate or a range, but we knew in the back of our minds that we couldn’t pay professionals with money we didn’t have.
To quote a saying that’s been used many times in many contexts:
“It’s hard, when you’re up to your armpits in alligators, to remember you came here to drain the swamp.”
We used to talk about this idea a lot at the nonprofit. If only we could get enough help to fend off the alligators, we could finally drain the swamp, eliminating the problems at their source. Rather than constantly dressing alligator-bite wounds, we could build a better arts nonprofit. We eventually did, and I’ll explain how we did that at the end.
Dan’s Book and Stephen’s Blog and the Idea of Synthesis
When I was first starting this article series, I was deep in learning about data visualization. One of the giants of that field is Stephen Few. In my research I came across his blog, and a very powerful idea he was writing about – ‘upstream thinking’. He had just read Dan Heath’s book Upstream, and his review of it turned on a light for me. It’s a concise review, well worth the read. Here’s one quote from Stephen:
“Essentially, Upstream explains and promotes the importance of systems thinking (a.k.a. systems science). Systems thinking involves synthesis, seeing the whole and how its many parts interact to form a system. It’s the flipside of analysis, which focuses on the parts independently and too often gets lost in the trees with little understanding of the forest.”
Much of our work at Source to Share is analysis. We’re analysts after all. However, analysis isn’t everything. What we do is more of an analysis sandwich, with thick slices of synthesis on either end. We’re taking in as much information as we can about the client’s needs and goals, and all the factors that are hindering them. Then we dig in and analyze how to fix each piece. At the end, it’s back to synthesis – seeing how our solutions are fitting their needs in real time.
How to Drain a Swamp
At the arts nonprofit, we were constantly bogged down trying to handle problems when a simple upstream system fix could have eliminated the downstream work altogether. We were running programs as fast as we could in order to make enough of a margin to keep the lights on. We didn’t have time to stop, look at past trends (assuming we even had the data), and build a budget with which to project earnings. If we did, we could tell people exactly what we could pay them, knowing that even if we undersold the house for a particular show, it would balance out in the long run.
Last night, my partner and I were watching an episode of Queer Eye, and I had the realization that Tan, Karamo, Antoni, Bobby, and Jonathan Van Ness are systems analysts. They may not self-identify as such, but they each work on systems, identifying those that are causing downstream problems and creating new supportive systems for the guests on the show. Overhauling systems is a big lift, and the energy that the Fab 5 invest creates efficiencies that result in a massive long-term return.
For a small business, nonprofit, or a research operation, it can be difficult to implement systems overhauls. Training staff (or adding a position) to learn business intelligence or a new program is expensive. You might not have the funds to stop operations for a week and create new systems that will multiply your impact down the road. In reality, you can still make changes even if you don’t have the Fab 5. At the arts nonprofit, with every extra bit of funding or time, we worked on systems. We figured out a supply schedule and got office supplies delivered. We started tracking data and after a year, we built a budget. It wasn’t great, but the next year we had time to build a better budget. The year after that we had enough time to upgrade our budget again, and still had resources left over to make an annual report. There was no single magic fix except diligence. Small changes snowballed over time.
This is exactly how we work at Source to Share. With clients, we meet them where they’re at, with whatever budget they have, and implement as many systems overhauls as we can. I’ve been amazed by what Bo and Derek can accomplish in even an hour or two (I’ll get there someday!). Those overhauls increase efficiency and free up capital, allowing businesses to reinvest in further efficiencies and systems improvements.
Speaking of nonprofits, we do pro bono work at Source to Share as one way to give back to the community. If you’re a nonprofit professional or know of a nonprofit that would benefit from business intelligence services, send us an email or give us a call!
BI Applications
The upstream solutions ideal applies not only to where and when we work, but how we work at Source to Share. We often encounter obstacles while creating reports for clients. By addressing those challenges upstream at the data prep level or with our data model, we’re able to avoid downstream problems in measuring or visualizing our data at the report level. Let’s look at a generalized example of how this plays out…
Having better data models means less complex DAX code (or whatever code you’re using to measure your data), less DAX overall, and less processing load on your computer when you evaluate expressions. Plus, complicated DAX is just harder to maintain. We’ll get into the technical details in our upcoming series on DAX fundamentals. For now, know that measures create CPU load whenever the system is queried (which can happen often as an end user explores a data model), and calculated columns take up space in RAM while also increasing the file size – both finite resources.
By building complexity into the data that’s loaded into the model, rather than into measures performed on the data, the work is being done when Power BI (or whatever BI service you use) refreshes the model, ideally at 3 or 4am each day. Then, when the user is exploring the data, the simplified code is evaluated quickly; the user won’t need to wait as the engine does complicated calculations on 600 million rows. Performance increases and the user can get back to doing the work that they intended to do.
Think of it like trying to build a home with the screwdriver in your pocket rather than going out and buying (or borrowing from the library!) an electric drill. They both work, and even though the screwdriver is immediately at hand, you’re going to save yourself hours and hours of work if you invest a little bit of time or money into the electric drill.
In Conclusion – A Summary of Season One
This brings us to the end of Season 1! In this season, we’ve taken a broad view of business intelligence in five articles:
1.1 – The Post that Started it All: What is Business Intelligence?
1.2 – What is a Data Model?
1.3 – The Challenges of a Single Table, or: A Star is Born
1.4 – A Key is Key: Making Relationships Between Data
1.5 – Upstream Solutions to Downstream Problems
These articles covered the upstream concepts and principles that will remove downstream problems for you as you grow in this field. I’ve included a summary below, telling the story in brief while highlighting the important concepts in bold:
We talked about the four broad disciplines of BI: Data Prep, Modeling, Measurement, and Visualization. Then, we learned how data fits into tables which evolve into data models, starting with basic denormalized tables. These single-table models could be all you need to answer your questions, but they can also grow past limitations to become star schemas–this process is called normalization. The process separates data into fact tables and dimension tables. In order to understand why tables can grow long (but not wide), we learned about cardinality and columnar storage. Our new star schemas rely on relationships between tables in order to filter the data, and so far we’ve covered directional relationships and many-to-one relationships. These relationships connect tables using keys. We learned the different names for keys we might encounter out in the wild. Finally, we spent some time discussing the why, how, and what of BI, pairing these tools with real-world needs. In doing so, we learned a new way of understanding BI work: as a combination of analysis and synthesis.
We’ll be diving into more of the technical bits in future seasons, and as we do, I encourage you to keep the underlying purpose of BI in your mind:
“Business intelligence is about having the right information at the right time in order to make informed decisions for your business.”
-Me, from “The Post That Started it All”
It’s easy to get lost in maximizing a formula or creating the ‘perfect’ data model (if you ever see one, please send it to me). What’s harder is keeping the purpose of BI close at heart, and doing no more and no less than is necessary to move your endeavor forward (or help your client do so).
Thank you for taking the time to invest in yourself by reading these articles. By participating, you not only broaden and enrich the field, you also honor my contribution. You’ve taken the first steps along a profession that’s having a profound impact on the world. We’re in the business of efficiency, of eliminating wasted effort and resources, and creating clarity. In a literal sense, we’re on this journey together. I’m writing as I learn, and I’ve already started work on season two. I hope you’ll join me.
Sign up to receive email notifications when we release new learning content
Sources
- Stephen Few’s blog: http://www.stephen-few.com/blog/2020/04/14/upstream-thinking/
- Data Modeling for Power BI – Online Course from https://www.sqlbi.com/
Leave a Reply