Predictive Analytics: Foretelling Successful Sales and Marketing

The concept of predictive analytics is not new. Predictive analytics has been around for well over a decade. Despite its age, it has mainly been the purview of large organizations for most of its existence because of the technical complexities of building the predictive models. In its simplest definition, predictive analytics leverages techniques from data mining, predictive modeling, and machine learning to create models that, in some fashion, can be used to predict future outcomes.

In the world of CRMs, the benefits of applying predictive analytics can range from the obvious to the esoteric. All organizations perform “predictive analytics” each time their sales teams engage in a forecasting meeting. However, moving from gut-feel-based predictions to data-driven predictions can dramatically impact company performance.

It's All About Your Data

If you've ever attempted to use Salesforce’s built-in Einstein-based AI tools, you may have run into one of the first hurdles of AI and machine learning: data of insufficient or inadequate quality. Regardless of your CRM, data quality (or lack thereof) can be one of the biggest challenges to implementing a predictive analytics process. A 2020 report by data quality company Validity found that 27 percent of marketers reported bad data cost them 10 percent or more in lost revenue annually. A 2020 report by Dunn & Bradstreet found that 100 percent of companies that invested in data quality saw performance improvements. So data quality is critical in building effective predictive models. However, even with better quality data, building successful predictive analytics practices often comes down to skills and resources.

Monolithic or Home-Built?

So you've decided that predictive analytics is of strategic importance to your business. Great! The next step is to determine how to do the work necessary. While there are several ways of building predictive analytics into your CRM workflow, a fundamental decision is about platforms. Will you leverage monolithic, built-in tools that (often) come with your CRM? Or will you expand your predictive analytics modeling to include data outside of your CRM? The side you land on usually has more to do with human resources, skill, and available data than an actual strategy. Traditionally, building a predictive model that leverages data from your CRM and other data sources can produce a far more accurate and useful model. However, creating the infrastructure to develop these complex multi-data-source models has traditionally been expensive, time-consuming, and has required expert programming talent. The answer ultimately will still come down to resources. Organizations, however, need to be aware of new technology that has become available in the past few years that is suddenly putting the power of sophisticated, multi-data-source models into the hands of just about any size business.

Think Through Your Use Cases

Before you consider AI-powered automation and how you can leverage it for more sophisticated modeling, you need to think through possible use cases. Identifying high-quality use cases is often one of the biggest failures in building predictive models. It is a mistake that can turn a predictive model from mission-critical to just an experiment. When ideating your use cases, always look at it from the perspective of your business users. Sales and marketing leaders are often concerned with maximizing ROI or accelerating pipeline. Your marketing leadership may be worried about their lead rotting rate, while your chief revenue officer may worry about accelerating sales velocity. Your use cases may vary, but there are some common, high-value use cases for sales and marketing teams:

1. Lead scoring. Most sales and marketing teams engage in simple lead scoring to measure lead quality. However, predictive lead scoring can significantly benefit your sales team by prioritizing higher-quality leads for faster follow-up by using historical lead data to identify higher-quality leads.
2. Predicting buying propensity. What if your sales team could reliably predict which opportunities are more likely to close? That's the concept behind sales-cycle predictions. You can leverage historical data to provide accurate predictions of which opportunities are more likely to advance.
3. Customer retention and upselling. These areas can benefit significantly from consolidating data from multiple sources. For example, you could combine CRM data with finance data and service to understand how payment history, service performance, and perhaps product usage can impact customer churn and identify upselling opportunities.
4. Managing your sales quota. This is always a challenge. How do you know that your sales quotas are fair, achievable, and beneficial to your business? Predictive analytics can help you model sales quotas to build more effective, suitable quota systems and help identify problem areas before they happen.
5. Territory management. This is a critical yet often mismanaged part of keeping any CRM system up to date. Once again, a predictive model can help businesses manage territories more efficiently and for maximum growth.

These are just some classic examples of how predictive analytics can provide concrete benefits to your sales and marketing teams. As you work through your use cases, always ask yourself a simple question: How will this provide value to the business?

Explainability: Black-Box vs. White-Box AI

Regardless of how your business builds its AI models, explainability is a simple but essential step in the process. One of the biggest challenges in implementing a successful predictive analytics program is to gain buy-in from line-of-business users. Getting buy-in is especially critical when dealing with sales and marketing teams. The challenge of developing predictive analytics is that some of the algorithms typically used in prebuilt predictive systems tend to be “black box” in nature—very accurate but impossible to explain. The lack of explainability of your predictive solution can create a trust problem where predictions are ignored, despite being accurate—all because your users don’t have a high degree of confidence in how the predictions originated.

AI Automation: Making Predictive Analytics Easier

Eventually, your organization will want to build its predictive models using custom algorithms. The challenge your business faces at this final stage is one of talent and skills. There is a high likelihood that you will not have the right mix of data scientists and programmers needed to build your own in-house AI models. That's where no-code AI development solutions will come in handy. New platforms available today put the power of AI automation into business intelligence professionals’ hands. Finally, everyone from business analysts to data engineers can create sophisticated, custom predictive analytics solutions without employing an army of data science professionals.

Ryohei Fujimaki is the CEO and cofounder of dotData. Prior to founding dotData, he was the youngest research fellow ever in NEC Corporation’s 119-year history; the title was honored for only six individuals among 1000+ researchers. During his tenure at NEC, Ryohei was heavily involved in developing many cutting-edge data science solutions with NEC’s global business clients and was instrumental in the successful delivery of several high-profile analytical solutions that are now widely used in the industry. Ryohei received his Ph.D. degree from the University of Tokyo in the field of machine learning and artificial intelligence.