Predictive Analytics Value Map: Identifying Business Value on Predictive projects
One of the key ingredients to success for any project is how much business value it offers in comparison to the cost. Lately, some of the predictive projects are taken up because of the hype and those are the projects that not only fail to deliver success but also tend to cast a shadow on the AI industry in general. That problem is not related to predictive technologies but more related to the parties involved in not considering the business value of the predictive question that is being solved.
Before one moves ahead with a predictive project, it’s extremely important to clearly articulate the value that the project will deliver. When the value is clear, it’s easy to align people and resources
A Simple Framework to identify Business Value
Business Value for a predictive project could be monetary in nature or could also be non-monetary as well. The following Value quadrant illustrates a simple approach to identify Business value for a predictive project.
The key goal of this framework is to provide a simple and straightforward way to identify Value / Return of investment for a predictive question / use case being considered.
A. Monetary factors
1. Predictive Question helps to increase revenue
This could be identifying the right product for a customer to buy as part of cross-sell / up-sell campaign using recommendation engines. The full life-cycle of a customer can play a role here starting from will this customer be interested/have a need for this product, will this customer respond to the campaign and finally will this customer buy the product. All are possible ways to enhance the engagement with the customer.
2. Predictive question helps to reduce cost / reduce revenue loss
This is where most customers tend to navigate to get started on their predictive journey. Here the goals are to reduce the cost, for example, identify how many customers will show up in my shop tomorrow to have the right staff. Or to reduce revenue loss by proactively identifying customers risk of churning.
B. Non-Monetary
3. Make customers/employees happier
This goes after customer satisfaction. For example, improving the wait times in the call center or identifying the best touch-point/times to engage with the customer or reducing time to respond to a customer question.
4. Make a process better
Could be to identify inventory shortages, roadblocks in the process, an API causing a problem in application performance, number of calls a call center will receive to staff correctly. Here we are essentially looking at various processes/building blocks that can potentially be improved.
Most times a use case typically falls under 1 or 2 buckets or possibly more.
The use cases with good potential for success include ingredients from at least 2 boxes in the Value framework quadrant above
Let’s look at 2 examples below to use this Value Quadrant.
Example 1: Value Quadrant for Customer Churn
Let’s take the example of Customer churn problem we discussed in the previous article. There are three possible strategies companies use to add more revenue include acquiring more customers, cross-sell / up-sell to existing customers, or increase customer retention. The cost involved in retaining is a lot easier than the other two areas and hence companies tend to focus on retention strategies to get started. The first step in getting started is to identify the customers who have a high probability to churn.
The two areas we can see it adds value includes:
- Reducing churn helps reduce the loss of revenue
2. In the process of reengaging with the customers for reducing churn, it helps to make some of these customers happy.
One of the great aspects of this scenario is, it applies to all customers in every industry. It’s also easy to measure as the goal is to decrease customer churn. Once the customers are identified then one needs to ask (or potentially predict) what message/offer to send & what is the best time to send. Once this action is taken, the last step is monitor to make sure there is an improvement in the churn rate
Example 2: Value Quadrant for Propensity to buy a product
Another example is to identify if a customer would buy this product or not. The propensity to buy scenario uses data from previously executed campaigns. The data includes customer demographics, the type of the campaign and if the customer bought or responded to the campaign.
An example is a company selling cleaning supplies as a subscription service for the past 10 yrs and customers reorder on a quarterly basis. Most small business simply gives an x % discount to all customers. Not all customers are alike and have different order schedules. By predicting the right discount % customized for each customer, it will help this company to achieve higher sales and retention without increasing costs.
This one we can see has the following 3 ingredients
- Helps add new revenue by engaging the right people
2. Helps make customers happy as the offers they get is relevant
3. Helps improve the process by eliminating wasted resources by right targeting
Note that the second reason for this use case is not as strong as the 2nd reason for customer churn use case.
Framework Summary
The following 4 important questions provide a simple rule of thumb to identify Business Value for general predictive problems to solve.
- Is the predictive question going to increase my revenue?
- Is the predictive question going to decrease my cost? Is it going to decrease my revenue loss?
- Is the predictive question going to improve customer satisfaction?
- Is the predictive question going to improve my current processes?
Though there are a number of nuances to this framework, this provides a good starting point.
Exercises to practice using this framework
To help practice the above framework, here is a simple worksheet that can be used to practice with this framework. This simple worksheet has 3 use case examples to apply & practice the framework covered in the article.
