Staying Ahead of the Competition with Predictive Analytics
Most product professionals are competitive, well-researched and typically grounded in their thinking and expectations. They’re adept at influencing without having direct power. It’s the things outside of their control that keep them awake at night, and the primary “thing” that’s out of their control is competitive disruption.
Changes in customer behavior, the industry and competitors’ offerings are why products routinely go out of favor—particularly in the digital space. For example, a digital enterprise product that was well-received when it launched in 2015 may be on the “reinvent or die” list this year after a series of startups enter the market with highly competitive offerings.
But this is the typical tale of many software solutions that failed to read changes in both customer behaviors and the market. Companies that see these changes early in their innovation journey have an edge, while the rest find out the hard way. The trick is to use your business data to figure out how to reach and keep that competitive edge.
Leveraging Predictive Insights
Today, most applications, such as apps or dashboards you use in your daily job, tend to talk about what happened in the recent past. This is useful for the business to understand the health of the company, but it doesn’t help shape the future. With a gold mine of business data, you have an opportunity to get an edge by predicting what is likely to happen as well as recommendations for the best action to take.
Predictive analytics allows us to uncover what will happen—and subsequently incorporate those insights into current and future digital products. The following are examples of refactoring current applications to add predictive insights for significant differentiation from the competition.
Adding Predictive Insights into Your Existing Application
Following are four simple baseline areas product professionals can explore to create the right frame of mind for adding predictive insights. Each includes example questions to put your thinking on the right path. And remember: Every question you are trying to answer should have a good ROI.
- Predicting outcomes:
- Will this customer churn?
- Will this customer default?
- Forecasting metrics:
- Number of orders?
- Number of calls to call center?
- Identifying anomalies:
- Is this a fraudulent transaction?
- Is this a fraudulent claim?
- Creating segments:
- Segmenting customers by demographics / sales patterns
- Segmenting patients by demographics, health history and medications taken
Predictive insights provide an opportunity to innovate and add new capabilities to existing products or help create new product offerings. For the digital products you’re currently managing, look at your existing product and ask what future outcomes can be predicted that will create an opportunity for you to add significant value for your customers. It can be as simple as taking your current key operational metrics, using them to predict their future value and deciding whether that prediction, along with some recommended actions, will help positively affect your business.
Getting a Handle on the Technical Side
To truly understand how you’re able to acquire and leverage the predictive insights you need to make your products successful, it’s important to understand the behind-the-scenes work it takes to get the analytics you want and the information you’ll need to get it done.
Traditional programming involves someone (a programmer) coding a program (building rules) that uses available data to produce an output. To keep the rules updated, the programmer writes more rules:
In machine learning, the data that is input and output are fed to an algorithm that automatically creates the rules. In turn, this program can predict on future data:
For example, to build a customer churn application, you would feed the machine-learning algorithm with data (e.g., demographics, product use, transactions) that includes samples of customers who have churned as well as those who have not. The algorithm mines this data and formulates the program to predict whether someone will churn in the future:
This enables the creation of higher-order complex rules that can be updated with changing customer behaviors—something that is impossible if manually written by programmers. It can also identify multi-factor rules that can be hard for humans to identify.
This automated process provides you with an opportunity to leverage business data as a financial asset and add new predictive features to your products.
Putting Analytics into Action
After identifying a key predictive insight, it’s time to test and check in with your customer base. It’s important to constantly experiment and engage customers with new ideas—that’s what leads to innovation and the ability to remain relevant in the competitive market. This work leads to enhancements of the existing product and possibly results in a new product for a cross-sell/upsell opportunity.
A Call to Action
Identifying the predictive problem you’re trying to solve is your starting point. Define the benefits this solution will have for both customers and the business. And don’t be afraid to experiment to get early customer feedback. Capture enough data to engage with stakeholders and, ultimately, navigate your way in the right direction.
Looking for the latest in product and data science? Get our articles, webinars and podcasts.