On a daily basis, we read multiple articles about data science and data analytics and how these practices are driving the next generation of innovation in retail. Insights that can be drawn from data assets are moving the needle”at businesses in ways such as maximizing sales, improving margins, identifying strengths and weaknesses in marketing campaigns, and improving employee training and turnover—all with the goal of increasing profit and shareholder value.
In loss prevention, we are accustomed to using data to make informed decisions: from complicated queries to find POS exceptions, to using the inventory shrink figure to identify focus stores. But the pool of available data is expanding rapidly, and the proven methods of the past can’t keep up. To find success, it is imperative that the LP practitioner become more knowledgeable about the universe of data science and analytics.
This post is the first in a series where we will dive into some of the major components of data science and analytics, helping you to build a foundation that will allow you to better understand the topic and recognize some of the ways that these practices are relevant to our industry and your job.
We’ll begin by laying out background information, setting the stage for a deeper discussion moving forward.
Let’s Start with our Data; It’s Big!
Data science is a practice focused on data, so first let’s understand the nature of the data we’re working with. This brings us to another buzzword: Big Data. Our industry had been dealing with big data long before the term became popular. Let’s take a closer look.
Big data is data that is characterized by the “3 V’s”: Volume, Velocity, and Variety. Volume and Velocity are apparent by looking at the POS data we process. It’s generated quickly, and there’s a lot of it. When we add in supporting data and metadata about a transaction, such as the items in a basket and their attributes, along with the attributes of the terminal and cashier ringing the purchase, we can see that we are generating and capturing a tremendous amount of information.
Taking a broader look at our systems lets us see the Variety. In addition to the structured data from the POS terminals, we have unstructured data such as video feeds and customer service phone call recordings. Adding structured or semi-structured data from other systems such as EAS, burg and RFID, audit scores, all of the additional sales channels and processes, inventory management systems, and even telematics data from your supply chain logistics makes the data even more diverse. Storing and analyzing such a variety of data is the challenge associated with the third V of big data.
It’s important for us to recognize that we have big data. Trying to manage our data with approaches designed for “small data,” while potentially useful for some problems, can be limiting for others. When we recognize the big data nature of our work, we can include the tools and techniques used to manage big data as we figure out how to get more value from the information our operational systems are producing. This is important because we need to first manage our data properly before we can begin to analyze it. Without having the right data management tools in place, including processes for validation and data cleansing, we cannot begin to do anything useful with our data.
Add Machine Learning, Traditional Analytics, and Business Intelligence
With the right data management systems in place, we can begin to work with our data. While machine learning (ML) tools such as neural networks have been available for quite some time, the computing power to apply them at the scale we need is not available on-premise at most companies. But with cloud computing, all companies can access enough computing power to experiment with and apply machine learning to their problems.
With data management and the availability of large-scale computing power to form a backdrop, we can begin to see advances in anomaly detection, fraud, improved recommendation engines, sales forecasting, and operational optimization.
Even traditional analytics have become more powerful. In fact, now that we have better data management tools, we can more easily work with data that was otherwise difficult to extract or integrate into our systems. Business Intelligence (BI) is also more important than ever, with many BI tools having become integrated with many big data management platforms, running both on-premise and in the cloud. Modern BI tools can do an amazing job of letting us view insights generated by ML and traditional models and analytics. Indeed, using business intelligence tools as a delivery platform is an extremely efficient way of viewing information and getting interactive insights to us.
Visualization to Bring it Home
With the added complexity created by big data, and after applying tools such as machine learning and cognitive analytics, understanding what our models and systems are telling us becomes more challenging. This is where visualization becomes critical. Visualizations complete the last mile of delivering insight from analytics. Visualizations are the primary way in which we interact with and derive value from the models and analytics we have built. They can show us new information and help us identify new areas to explore. Identifying the right visualization tools and visuals takes on even more significance in the presence of the large volumes of data and more complex analytical tools that we now have.
The Right People
To make this all work, it’s important to have the right staff with the right skillsets and domain expertise. There are many tools and platforms available for each of the areas discussed: Data management, analytics, and visualization. Knowing which tools to choose for a given problem, and then having the expertise to use those tools properly, is not trivial. This is where dedicated data science, analytics, and data management teams can excel, letting us take advantage of what’s available to get the most value out of our data.
It’s also critical to pair these quantitative experts with subject matter experts who understand the business problems we’re trying to solve. There’s a tremendous amount of information that can be gained from data, and subject matter experts can help guide the analysis towards the areas that matter most. Building a team of quantitative and subject matter experts is critical to getting value from our data.
Over the next few articles, we will discuss the topics of data management, analytics and machine learning, and visualization. We’ll see how, with the right people, we can work with these tools to give us new insights from our data, and we’ll explore some best practices for making them work for us.