You’ll learn the results of a field experiment conducted by Yael Karlinsky-Shichor (School of Business at Northeastern) and Oded Netzer (Columbia University) to explore “Who makes better pricing decisions in B2B settings – humans or machines?”
We now know that algorithms and artificial intelligence are part of our daily lives. The general trust in systems like Google Maps is relatively high when looking at the number of users utilizing the platform.
Nevertheless, it often happens that we see a route suggestion that causes us to frown and think, “That seems weird to me. I don’t think that’s right.”
The same can be said if we think that the predicted travel time displayed on the screen is incorrect based on our gut feeling. Whatever the reason, justified or unjustified, it happens. What is going on in our mind?
Yael Karlinsky-Shichor has made it her mission to find out. Karlinsky-Shichor’s research focuses on the automation of decision making and its application to marketing and sales. She also explores the psychological aspects of using automation and artificial intelligence models.
This post is about one study by the researcher which examines the effectiveness of artificial intelligence to make pricing decisions, as opposed to humans.
Let’s take a closer look at it.
Field experiment human versus machine – what have they done?
The experiment took place in a B2B wholesale company that sells aluminium in the United States. The purpose was to investigate what could generate a higher profit for the company – prices set by humans or machines.
The researchers provided a group of salespeople with a tool that makes automatic pricing recommendations according to customer and product demands in real-time. The company uses historical sales transaction data to train the tool. The method behind it is the “Random Forest” machine learning model. In simpler terms, the model calculates probabilities for price acceptance and then feeds specific price recommendations into the CRM system.
Thus, the group of salespeople in the experimental group adjusted their pricing based on historical data. The control group consisted of another group of salespeople without utilizing this tool.
Each incoming inquiry was randomly assigned to one of the two groups. During the experiment, they processed a total of 2000 price quote requests and over 4000 product requests.
The sales group was not bound to accept the price recommendations. The salespeople have the decision-making power, i.e. a right to veto the algorithm’s suggestion. Why do we think this is good? It’s realistic. An AI tool, like Predictive Sales Analytics, is a helper, not a dictator.
The results of the experiment
The group of salespeople with the pricing tool made 10% more profit than the control group. It has shown that the model leads to higher profitability in most cases. However, when it comes to pricing or quoting decisions for customers with unique or complex characteristics, the salespeople make much higher profits than the algorithms. The salespeople often have information about their customers, products, competitive situations, etc. that is difficult to code. Since it is difficult to code, the model does not have as much information as the salesperson has. This is because the model was trained using past sales transaction data alone (i.e., who bought when, how much, at what intervals, etc).
So, what now? When should automated pricing be adopted, and when should humans be responsible for making pricing decisions? Based on the results, the study was expanded.
The researcher developed a hybrid model. A machine learning system calculated when which pricing decision should be applied. “We use machine learning to decide who should make the pricing decision automatically – the salesperson or the model,” Karlinsky-Shichor explained.
The results speak for themselves: “We found that a hybrid structure, where the model calculates most of the incoming quotes while the expert salesperson takes the cases that are unique or exceptional, performs even better.”
We are happy with the study results, as they confirm our experience in the field on a scientific level. We are sure that humans and machines achieve the best results together.
CALCULATE NOW THE ROI OF QYMATIX PREDICTIVE SALES SOFTWARE
Conclusion for B2B companies: “Who makes better pricing decisions in B2B – humans or machines?”
What to do now? Science proved that machine learning methods create precise forecasts and thus support sales decisions enormously. The technical term in sales for this is “predictive sales analytics.” In this case, pricing forecasts alone increased profitability by 10%. If you also forecast customer churn and cross-sell to engage your customers effectively, you will improve your margins.
Even though the B2B market is valued at trillions of dollars, it still lags behind the business-to-consumer (B2C) market in adopting automation technologies. It’s time to change that. Wholesale in Germany is suffering from ever-shrinking margins and could benefit the most from such models.
Use your sales data and provide your sales team with a sales forecasting tool (predictive sales analytics software), which supports them with concrete, data-based recommendations to take action. Your salespeople don’t lose their jobs; as a result, they get better at what they do.
Is it time for you to become data-driven? Then talk to us or read the free whitepaper “How to get started with Predictive Analytics“.
I WANT PREDICTIVE ANALYTICS FOR B2B SALES.
Further Read:
Karlinsky-Shichor, Yael, and Oded Netzer. “Automating the B2B Salesperson Pricing Decisions: Can Machines Replace Humans and When?” Columbia Business School, April 8, 2019.
Khalida Sarwari (2019): Who makes better decisions: Humans or Robots?