Generating Business Insights Using Advanced Analytics
Learn the 4 keys to using AI and machine learning to make better decisions and gain deeper insights.
Join the DZone community and get the full member experience.Join For Free
Machine learning-based applications have seen significant commercial success in several mainstream consumer applications in the recent past. Self-driving cars, stock-trading bots, robo-advisors, Amazon’s Alexa, and Apple’s Deep Fusion and Siri are some of the renowned examples of commercial success with artificial intelligence and machine learning. AI has also made our lives easier by improving the customer experience of the products we use. Google’s text generation software, Netflix’s recommendation engine, and Facebook and Twitter’s fake news detection are other prime examples. In fact, every single technology company uses AI in its mainstream applications either directly or indirectly. Non-technology companies are also using AI to improve customer experience, improve efficiency, and generate new revenue streams. Chatbots, robo-advisors, systems that predict system failures, and products that generate efficient supply chain routes are some of the prominent ways in which non-technology companies use AI. This is leads to a popular belief that AI and ML are primarily used by technology companies or they are being used by non-tech companies to build AI-based products.
This popular perception is not true. There are plenty of avenues in which AI/ ML is being used or can be used by non-tech and non-product-based groups to generate insights. In this article, I am going to share with you four ways in which you can augment advanced analytics into your analytics strategy to generate insights.
I am going to first discuss what an insight is. The terms “insights using analytics” and “data storytelling” have become popular buzzwords. You would be surprised how many times these phrases appear in job descriptions and how often leaders mention them in their all-hands meetings. Webster defines “insight” as :
"The power or oct of seeing into a situation."
Generating insights helps us to understand the reasons behind an outcome. Below are four different ways in which you can generate insights using advanced analytics.
1. Explain a Phenomena
A study conducted by MIT Sloan Management Review and IBM Institute of business value conducted a study in 2011. According to the study:
Organizational leaders want analytics to exploit their growing data and computational power to get smart and get innovative, in ways they never could before.
Though it has been almost a decade since the above study was conducted, the majority of business organizations still use traditional data analytics methods. Traditional data-driven ways to understand an outcome involve creating reports, dashboards, and charts and understanding the causal factors by creating multiple line graphs. While the graphs show trends, we can’t draw causal inferences from them. Traditional data visualization techniques will also fail to explain how multiple variables interact with one another and how they impact the response variable. Instead of creating multiple graphs, we can consider performing simple statistical procedures such as significance tests, correlation analysis, and linear regression. Conducting statistical analysis even in its rawest form will give us significant insight into the factors that can effectively explain a phenomenon.
2. Create Personas
If you are in the business of analyzing sales, marketing, HR, or customer data, the concept of creating personas will not be new to you. Creating personas is the process of understanding the attributes or traits that represent certain groups’ behavior. Most often than not, these personas are defined by several “rules.” These rules are a bunch of if-else statements that were defined by few subject matter experts.
The problem with the rules-based approach is that the rules are defined by people and that means that they will have inherent opinions and biases built into them. Another common problem we observe with the rules-based method is that they change the rules whenever there is a new sheriff is in town. Machine learning can come to your rescue if you are experiencing these problems. Using simple unsupervised clustering algorithms, one can easily set up an optimization problem to understand the traits of people’s behavior. What’s even more interesting is that we can explain the behavior and create the personas optimized for different metrics. For instance, the personas of the users optimized to measure the site visits can be different than the personas of the users who are buying your products.
A persona represents an aggregate of target users who share common behavioral characteristics (i.e. is a hypothetical archetype of real users).
3. Define Strategy
When is the right time to release your product? How should you craft your job description to attract top talent? When is the right time for my workforce to start coming back to the office? How many leads should I generate to meet my sales goals? These are all very important business questions. Most modern companies define their business objective ahead of time and break them into multi-year plans to execute them. Machine learning techniques can help you to define your strategy in a data-driven way. One can deploy techniques such as scenario analysis to understand how many/what kind of resources you need to accomplish a certain task and how would that vary if the scenarios vary from one extreme to the other. Machine learning models can generate actionable insights. For instance, by building a good regression model, one can explain what you need to do if you have to increase your sales by 10% next year. The models can tell you how much you have to spend on marketing, what kind of marketing, and in which locations. By taking data-driven approaches, business leaders can define strategy in a clear and unbiased way.
4. Test and Learn
Oftentimes, business leaders rely on their intuition, hunches, and gut feelings to make decisions. There is nothing wrong with this approach as their entire career thus far has prepared them to make those decisions. But how machine learning can be used here is by measuring the efficacy of these decisions. For example, is sending marketing emails once a month to your customers better than sending them biweekly? Decisions made in situations like this can be validated and measured by using techniques such as A/B testing (split testing). A/B testing will help us to understand which variant of your strategy is yielding us better performance for a given goal and more importantly, these findings can be scaled across broader projects to achieve your targets.
Opinions expressed by DZone contributors are their own.