#14 Kickstarting Your Organizational Transformation to Become Data-Centric
Subscribe to get the latest
on 2020-08-20 00:00:00 +0000
with Darren W Pulsipher, Sarah Kalicin,
Today’s episode is all about kick-starting your organization to become data centric and the value this can bring. Darren’s special guest is Sarah Kalicin, lead data scientist for data centers at Intel.
Keywords
#analytics #data #datamanagement #multicloud #people #process
Operationalizing Analytics System
The first thing to understand about the process of operationalizing analytics is that it is complex. Some people think data is the new electricity and works in a similar system. Consider this example: In a simple electrical system, there is a power source and some wires that go to a light bulb. As long as there is the power source and the wires are high integrity, you can turn off the light, and then six months later come back and flip the switch and know the light bulb will come back on.
An analytical system does not work this way. There are many types of “short circuits” and variables that can cause failure, such as entropy. It’s important to have the right people on your teams who can work through and understand all the different types of faults that can occur or you will not get correct or useful insight.
Your data must be formatted in a certain way for machine learning algorithms. The algorithm must understand what the data represents to look at the pattern appropriately. Unless the data is carefully formatted, there is a garbage in/garbage out risk. This is where your engineers, data scientists, and other supporting data architects are important.
Asking the Right Questions
Another part of the system is making sure you have the right business questions. An organization must supply foundational questions about what they are trying get from the data: What insights does our business need? How do we paint a better picture of what our landscape actually looks like? What data will allow us to do that?
Many executives love the idea of doing analytics, AI, and big data, but don’t know how to get started. The first step is for executives to take ownership of what they want the data to do for the organization. What business value is going to come from the data? What type of data do we already have? What data can we create that will paint a more complete picture of what is happening? What does success look like? The focus should be first on the organizational foundation.
Data Collection, Preparation, and Insights
After those questions are answered, the next step is to outline what needs to be accomplished and start breaking it down into smaller parts. Organizations should be looking at their current data sources, determining their reliability, and identifying what is missing, followed by data collection and preparation.
With the data in hand, it is time to create the story. This involves not only data scientists, but domain experts and business and marketing people. Input from these different groups will ensure the data will add value to the business rather than becoming just an academic exercise or a one-time experiment. From the data, you ultimately want valuable, reliable insight that will help improve your business rather than simply data collection that is convenient but not actionable.
Now that you’ve collected the data, analyzed it, and have some insight, how do you operationalize it?
Operational Analytics Example of Success
An example from manufacturing is a good way to show how the data successfully works through the pipeline. Start with the foundational business value of maximizing your product yield. What data sources could tell you if you have broken or disfigured products? Cameras at the end of assembly lines could create a deep learning model that detect or predict if a product is broken or disfigured. A data scientist could then train a model based on good product and bad product to improve yield.
This type of operationalized data could be applied to several key areas that add value to an organization. Quality control, such as making sure raw material meets specifications is one example. Predictive maintenance of machinery and regulatory compliance, such as in a weight requirement for a food product, are other examples.
By tapping into the data in real time about product, quality, or regulatory issues, you are not just creating a higher product yield, but minimizing costly human tasks, such as inspection at the end of an assembly line. These employees can be redirected to higher value tasks. You are also minimizing waste and creating a more consistent consumer experience. In addition, you are protecting the business by minimizing regulatory risks.
This is a win-win scenario. The more the organization knows what they want to gain from data, what data to collect, and how to utilize it based on a business strategy, the more successful they will be.
More to Come
This is the first of six episodes that talk about the data-centric organization. We hope you will join us for those future podcasts. Want more? Check out Sarah’s website,