Predicting the Unpredictable: How Sales Managers Can Get Better ROI from Predictive Sales Analytics Tools

Journal of Marketing Research Scholarly Insights are produced in partnership with the AMA Doctoral Students SIG – a shared interest network for Marketing PhD students across the world.

As the importance of data in decision making continues to grow, predictive sales analytics have gained immense popularity in business. Companies are currently using predictive analytics for numerous functions such as predicting conversion likelihood to prioritize clients and improving consumer retention by predicting consumer churn. However, salespeople may be averse to using predictive analytics, as they may mistrust technology or lack the knowledge to use it. Still, it is unclear under what circumstances predictive analytics can be effective.

In a recent Journal of Marketing Research study, researchers Johannes Habel, Sascha Alavi, and Nicolas Heinitz address this challenge by focusing on customer churn prediction in a business-to-business context. Their extensive study, which analyzed 9.7 million transactions, revealed that the success of predictive analytics tools hinges significantly on the characteristics of both customers (like churn probability and previous revenue) and salespeople (such as their perceptions of technology and abilities to use it).

The research further demonstrates that the effectiveness of these tools is enhanced when users have realistic expectations about the algorithm’s accuracy, though this was found to be effective only in specific circumstances. These findings, derived from a meticulously conducted field experiment conducted by collaborating with 12 companies for more than two years and reinforced by two follow-up experiments, provide critical insights into the nuanced interplay between predictive models and their real-world application in sales.

For practitioners, policymakers, consumers, and other stakeholders, the key takeaways from this research are:

The success of predictive analytics in sales, specifically for customer churn prediction, is highly dependent on both customer characteristics (like churn probability and prior revenue) and salesperson traits (such as technology perceptions and abilities).
Implementing strategies like fostering realistic expectations about the algorithm’s accuracy can improve the effectiveness of predictive analytics tools, but this is highly situational.
This study underscores the importance of understanding the nuances of predictive models in sales, encouraging a more informed and strategic application of these tools in various business contexts. These insights are crucial for managers looking to optimize their use of analytics in decision-making processes and for stakeholders interested in the practical implications of predictive analytics in the business sector.

We had a chance to contact the authors to learn more about their study and gain additional insights.

Q: What inspired you to delve into the topic of predictive analytics in sales, particularly in the context of customer churn?

A: The origin of this research project arose from two observations of our team in machine learning and firm practice. Regarding machine learning, we noted that the capabilities and access to machine learning models through the statistics software R vastly improved at that time. At the same time, firms grew increasingly excited about the potential of using such ML for their analytics in sales. In the workshop, we decided to cooperate with a B2B firm to explore the opportunities, potentials, and dangers of using predictive sales analytics. We discussed various use cases with the management of the company for predictive analytics applications in their sales force. Since customer churn often constitutes a key issue in B2B industries, we decided to focus on a predictive tool that could assist the sales force in this respect.

Q: Could you discuss the significance of your research in the broader context of business and technology? How does understanding customer churn through predictive analytics impact the field at large?

A: In the broader context of business and technology, our study points to a key tension: AI-based technologies exhibit tremendous potential to improve productivity and diverse other outcomes. But these are only promises of potential, which are not automatically and easily realized. Rather, all too often, implementations of such new technologies fail and may even backfire. Our study underlines the enormous nuances and contingencies that need to be considered when implementing such new AI-based tools. Such implementations never constitute low-effort quick fixes but major change management projects that require massive attention, effort and resources of management to succeed. Findings of our study contribute to marketing research by underlining the fundamental role of contingent effects of such technology implementations and uncovering the most important categories of contingencies.

Q: Are there potential extensions of your work that you find particularly exciting? How could other researchers build on your findings to explore new dimensions in predictive analytics or customer behavior?

A: Our study uncovered that salespeople deal very differently and develop different strategies in response to the implementation of predictive sales analytics tools. At the same time, the firm communication strategy to foster effective tool use did not unconditionally provide the wished for results. Consequently, an intriguing endeavor for future research, as we see it, is to develop a taxonomy of AI implementation strategies in sales and examine the differential impacts of different strategies to foster adoption.

Q: What were the biggest challenges and the most significant lessons you learned from doing a large-scale field study collaborating with multiple companies?

A: The first—and most important—lesson, in our view, is to first identify the right company, that is, the right and suitable cooperation partner. This lesson may sound trivial, but its significance cannot be overemphasized. This is because many companies initially intend to cooperate and join a research project on such an important and promising topic. Yet, after the initial enthusiasm, we noted that many companies realize the high effort and necessary resources which establish barriers. Moreover, sometimes companies are not suitable for compelling field study due to various issues such as lack of data, insufficient commitment or resistance by employees, or organization structural aspects. Therefore, it is important first to ensure the firm partner is eligible and committed. Otherwise, one may face enormous losses of time and energy.

Q: Do you think that predictive analytics also have negative effects in the long term? If so, how could managers mitigate this?

A: Yes, predictive analytics might exhibit negative long-term effects for companies and their employees. Such harmful effects may occur, for instance, if the new technology is initially implemented in an ineffective way such that employees’ fears and aversion had not been properly addressed and resolved. Then, such an implementation might cause disruption in teams, harming interpersonal trust and triggering negative long term social dynamics in the organization. The seed for positive long-term use and development of predictive analytics needs to be set with the initial implementation and introduction. Managers need to:

accompany the implementation with a deliberate change management project,
select an appropriate tool which indeed provides value and usability to the employees,
give them freedom and time to learn and experiment with the tool, and
make sure that the tool is employed in the right customer and market situations.

Q: Beyond asking people to have realistic expectations toward predictive analytics tools, what other strategies could be used to trigger people’s trust in predictive analytics even when they can fail sometimes?

A: Employees’ fears and anxieties when asked to work with the new predictive analytics tools often arise from deep-seated uncertainties regarding the potential consequences of the tools. For instance, they might feel uncertain about how to effectively work with the tools, perceive them as an intrusion into their work practices, fear being outperformed by more tech-savvy peers, or, most prominently, might dread being replaced by effective AI tools. Manager strategies to foster trust in predictive analytics must tackle and resolve those fears and uncertainties. A major prerequisite for such strategies to be effective is that a healthy relationship exists between the manager and their employees. For that, managers need to consistently demonstrate to employees that the purpose of the tool is to assist them and help them, and not replace them or monitor them.

Managers need to consistently demonstrate to employees that the purpose of the tool is to assist them and help them, and not replace them or monitor them.

Ideally, managers succeed in establishing a trust-based company culture that prioritizes employee well-being. In such a culture, fears of AI should be less pronounced, and employees should be more open to effectively leverage the opportunities provided by the new technology.