Reviewer

Dr. Scott Downey, Director and Professor, Purdue University

Article

Effective Implementation of Predictive Sales Analytics by Johannes Habel (University of Houston) and Sascha Alavi and Nicolas Heinitz (University of Bochum Germany)

Source

Habel, J., Alavi, S., & Heinitz, N. (2024). Effective Implementation of Predictive Sales Analytics. Journal of Marketing Research, 61(4), 718-741. https://doi.org/10.1177/00222437221151039

Summary

In the face of an increasingly complex sales environment, salespeople may need to adopt more sophisticated approaches to analyze customer information than their predecessors (who relied primarily on “belly-to-belly” transactional relationships). With an increasing number of decision-makers and influencers in a crowded, competitive marketplace with scientifically complex products, salespeople now need analytical tools to understand their customers better. This is a hot topic and will grow in importance as new tools emerge that use AI and other technologies. 

Predictive analytics use information about customers to anticipate their behavior. The question of whether salespeople dismiss information because they “know better” or embrace these insights is one that is closely related to the adoption of customer relationship management (CRM). Salespeople who are skeptical of data’s usefulness are less likely to track customer information or adopt technologies to analyze it. 

A recent paper examined some of the factors that drive salespeople’s use of predictive analytics, particularly in forecasting customer churn, or the likelihood of reduced spending. Salespeople were not asked to do calculations; they were simply provided with information that came from analysis of their customers’ behaviors. The experiment was conducted in the construction industry, working with B2B retail salespeople. The construction industry is a parallel industry to agriculture but experiences more frequent transactions over the course of a year. 

In the study, regional sales teams were divided into three groups:

  1. The first group received predictions on the likelihood of their customer churn.
  2. The second group received the same prediction data, plus training on the model’s accuracy.
  3. The third group was a control group that received no information as a comparison. A follow-up study with a different group was used to confirm the observed effects.

Some of the relevant conclusions from this research include:

  1. Information usage depends on the customer. Data use was higher with customers who had a very high likelihood of churn (lowering their spending) and large customers. This accounted for 38% of the use of the data.
  2. Attitudes influence data usage. The rest (62%) of the information usage depended on other factors, specifically attitudes toward the usefulness of the information, how attentive the seller was to customers as opposed to transactions, learning orientation, and others.
  3. Explaining prediction accuracy can hinder adoption. One of the bigger questions answered in the study was whether explaining the accuracy of the prediction led to more use. While it provided a deeper understanding, it also highlighted model imperfections, which initially reduced data adoption.

What does this mean for food and agricultural business?

Other research on salespeople in food and agriculture has shown that planning for how to approach customers is quite limited. Most salespeople identify what they want to sell and territories but rarely plan beyond the next customer interaction. Even then, it’s usually not a formalized planning process. 

Prioritizing sales activities based on predictive data may be a new behavior for many sellers. They often resist being directed to take specific action, dismissing it as uninformed compared to their personal understanding of what should take priority. When their income depends on personal judgment, they may be reluctant to change strategies that have consistently proven effective, regardless of the information source.

When predictive models contain inaccuracies, this resistance grows, as any errors in third-party data can validate their doubts. For instance, a sales rep might reject data estimating a farm’s acreage if it doesn’t account for all entities under the farm’s decision-making umbrella. Third-party data may have shown that a farm was 2,000 acres, but the seller knew that the farm made decisions for multiple entities totaling 3,500 acres. Sometimes all the data was dismissed because of those inaccuracies, even if it was largely correct for the market as a whole.

CRM adoption by food and agribusiness sales forces has been varied. Perhaps one of the reasons for this is that salespeople didn’t see the usefulness of the data from these systems – even if salespeople themselves were the ones who are responsible for its low quality.  

Food and agribusiness salespeople are very good at using information to make product recommendations and helping their customers achieve yields, production, or revenue goals. But it may be worth considering how good salespeople are at using information about their customers to improve their own sales outcomes.