I believe that predictive analytics is poised to enable businesses in ways we didn’t think were possible several years ago. It’s starting to play critical roles in solving inventory problems, loan prediction, user personalization, customer segmentation, propensity to churn, product pricing, and many other areas, with the goal of enhancing customer experience and the bottom line. And we’re starting to see applications being deployed in Healthcare, Manufacturing, Finance, and Retail.
While senior management may recognize the importance and value that predictive analytics can generate, they may equally be swamped by other issues and put predictive analytics projects on the back burner. So here are three strategies to help you and your team transform data into actionable insights and drive real value to the bottom line.
1. A Good Business Problem with Clear ROI
The key to getting started with predictive analytics is to identify a business problem that is meaningful, well understood, and has a clear return on investment (ROI) with a time horizon in mind. Business problems with high ROI will make it easy to get management, and possibly the whole company, aligned quickly.
Business problem possibilities can range from very large to very small. To get a better idea of the possibilities, here are a few use cases (internal and external) that can get you started.
Each of the above business problem use cases have clear and timely ROI attached to them. The progress of each can be easily measured. For example, a company has $50 million annual revenue and is experiencing $3-million churn (or 6%) annually. The goal of your predictive model could be to reduce the churn to 3% annually within the next 24 to 36 months. That is a clear $1.5 million-dollar saving every year, which will resonate with senior leadership.
2. Empowering a Team through Prediction, Distribution, and Action
Once the business problem is identified, it’s essential to align with stakeholders, and offer them the opportunity to provide guidance/feedback. This will help make your predictive analytics project more successful. Having the right council members will also help navigate internal politics as the data being used as the foundation for the predictive model could be coming from different departments. Having key leaders as part of the council can help source the data needed.
For a project of this nature to be successful, the focus should not be solely on creating the predictive model. There should be equal importance given to using the insights and taking action.
Let’s stick with the customer churn use case mentioned above, for example. Let’s say the model identified some customers could potentially churn in the next few months. For that information to be useful, it has to be distributed to the right people internally, enabling them to take action. When you are developing your initial business problem to answer with a predictive model, ask yourself, if we see X as the outcome, who needs to know in order to take action. For the customer churn example, you’d most likely want to share the insights with the customer support and sales teams. Based on your business model, the following questions could be helpful:
- What information should be provided to the support team if we know a customer is going to churn? What will be most helpful?
- What recommended actions should be provided in order to address the predictive model’s outcome?
3. Scale Early Small Wins and Continuously Improve
When getting started with a predictive model, start with something simple that you’re able to beta test with your trusted council or team of stakeholders. Incorporate their feedback and scale the production of the model. As your initial team gains confidence in the model, consider this a small win and slowly expand your team to others in the organization. These small wins will add credibility to your project and will help more people in your organization become engaged.
As your model evolves, continuously engage with the users to review progress and to incorporate feedback. This will help to improve the overall, end-to-end experience.
Some examples of continuous evolution include:
- Taking new data to retrain the model and to stay in sync with changing customer behaviors
- Adding more capabilities to the model to help end users gain new insights
- Building additional models or application features to help end users consume the predictions and take action
The key to getting started with predictive analytics is to actually get started – don’t wait for all of the stars to align.
- The first step is to have an open mind, acknowledging that some steps will go smoothly and some to have roadblocks.
- Start small, divide the problem into actionable pieces and constantly ask for feedback to get more confidence and demonstrate progress.
- Prioritize your features and goals, and work to make continuous improvements to add innovative features to your solution over time.