Analytics is like a Cake!...
...and everybody loves cake!
Simply stated, analytics is the science of analyzing raw data to make conclusions about the information collected. Even more simply stated—analytics turns data into answers.
Today, it’s hard to think of any industry that doesn’t use analytics in some manner to assist in real-time decision making, conduct predictive analysis, or even automate complex, multi-step processes (think self-driving cars). Enabled by the Internet of Things, smart building technologies, and cloud computing, the building industry is no exception, leveraging data and analytics to improve building performance and energy efficiency with the ultimate goal (and inevitable outcome) to follow the automotive industry’s lead into the realm of automated command-and-control or, what we like to call, smart automation.
Seems pretty straight-forward, right? But how is analytics like a cake? It’s simple, really; both analytics and cakes contain layers that build upon each other to produce a finished product that achieves the intended outcome. In the case of a cake, the intended outcome is a delicious dessert. Regarding analytics, it’s smart automation.
Here’s how I breakdown the layers of the analytics cake:
The Bottom "Base" Layer: Fault Detection and Diagnostics
In analytics, we refer to faults as anomalies, and these anomalies identify anything that does not meet an expected result. These are basic rules based on binary “Yes” or “No” answers. The more binary questions to which we can get answers, the better will be the ultimate analytical results. In other words, more data, more answers.
Here are some examples of these basic, binary rules (using a cake analogy, of course):
Is the oven turned on or off? (Are we preheating the oven as needed?)
Is the oven exceeding a threshold? (Did we set the temperature too high?)
Is the temperature indicator displaying correctly? (Is the oven heating as indicated?)
Is the timer set properly? (Did we bake it too long and burn our cake?)
The list of potential anomalies is exhaustive. Consequently, this layer can get a bit complicated with the amount of raw data being collected and processed. But let’s be clear, the objective here isn’t the amount of data that can be collected but, rather, the amount of reliable data that can be collected. Garbage-in-garbage-out definitely applies here, so the importance of getting this layer right cannot be overstated. Good, reliable data is essential in building a base strong enough to support the remaining layers of the cake.
The Second Layer: Root Cause and Correlation
This layer of our analytics cake is all about analyzing the data in a way that provides answers that go beyond simply detecting a problem. Here, we correlate the data against known, real-world problems to automatically identify the likely cause of the problem and its impact across the system. For instance, this one problem maybe producing thousands of faults. Resolve this one issue, and you’ll eliminate thousands of anomalies instantaneously. This is a critical layer of the cake. While the analytics won’t tell you how to fix the problem, they will indicate the exact cause of the problem and the universal impact that problem is having on your building’s performance, potentially saving countless hours troubleshooting the issue and then trying to find and fix it everywhere it has occurred.
The Top Layer: Artificial Intelligence & Machine Learning
Our analytics cake is really starting to take form, and the possibilities are pretty exciting. This top layer is all about predictive analysis and proactive resolution. So, let’s recap. We’ve filled our base layer with good, reliable data and an extensive rule set that enables us to detect systems or equipment that aren’t operating as expected (faults or anomalies). In our second layer, we built out a broad (as broad as possible) data model and can correlate the faults discovered in our base layer against this model to indicate root cause and proliferation. In other words, we can identify the problems, understand the likely cause and how extensive the problem is. Now, since we are storing all of our historical data in the Cloud, our top layer pulls it all together and correlates past and current data sets with our existing data models that have been augmented over time with real-world learnings and discoveries. Now, instead of being alerted after a problem occurs, our analytics can, based on a statistical likelihood, inform us before a problem occurs, enabling us to resolve it before it potentially impacts building and business operations.
The Icing: Automated Command and Control
Our analytics cake is fundamentally complete. We can find problems and their root causes for rapid resolution and can even reliably predict issues and resolve them proactively. But no analytics cake can truly be considered done until the icing is applied. In our case, that means smart automation—the ability to take all we’ve learned and know based on the data we collected, analyzed and correlated to go beyond simply identifying the problem and its cause but, instead, to enable the system itself to fix the problem automatically without the need for human
intervention. For instance, to extend the example from earlier, our analytics have identified that our oven temperature is set too high (fault detection), has indicated human error as the likely culprit (root cause) and based on past experience can predict the likely outcome will be a badly burned cake (AI / machine learning). But because our analytics system is connected to our smart oven, instead of sending an alert to the chef telling him all this, the system can simply set the temperature to the proper setting needed to bake the perfect cake (smart automation). While smart automation won’t replace the human element from the mix anytime soon if ever (I’ve never seen an analytic replace a faulty chiller), the promise of the technology is compelling and, as I stated earlier, inevitable at some level.
Let Them Eat Cake!
While the building industry in general continues to lag behind other industries in the use of analytics to improve both efficiency and effectiveness, analytics have nonetheless arrived and are gaining traction. Why is our industry so slow to adopt analytics despite the monumental benefits of doing so? Of course, any answer given would only be speculative, but I suspect it involves the fundamental question associated with any type of profound transformation—Is it evolutionary or revolutionary?
I’ll address this meaty subject in my next post. In the meantime, weigh in on my “analytics is a cake” concept or anything else related to the use of analytics to improve building performance. I’d love to get your thoughts.